Transcripts

This Week in Tech Episode 914 Transcript

Please be advised this transcript is AI-generated and may not be word for word.
Time codes refer to the approximate times in the ad-supported version of the show

Leo Laporte (00:00:00):
It's time for twit, this Weekend Tech. Sam Guam. My car guy is here. So is Brian McCullough from the Tech Meme Ride home. But it's not just cars. We're also gonna talk about floating object. Shelly. Shelly Brisbane is here from six colors.com. We'll talk about the Chinese balloons that tech payload they might have had. And why Six Chinese companies now banned from doing business in the us. We'll talk about Sergey Brynn's Blimp. Or is it a Zeppelin? I know it's a Dirigible. And then chat. G p t. Microsoft's all in on Bing and Google. Well, they might have fumbled their launch. That and a lot more. Coming up next on Twitter pod podcasts you love

TWiT Intro (00:00:42):
From people you trust.

Leo Laporte (00:00:44):
This is twit, is Twit.

(00:00:53):
This is twit this week at Tech. Episode 914 recorded Saturday, February 11th, 2023. O Washington. This week at Tech is brought to you by Bit Warden. Get the password manager that offers a robust and cost effective solution that could drastically increase your chances of staying safe online. Get started with a free trial of a teams or enterprise plan, or get started for free across all devices as an individual user at bit warden.com/twit. And by ZipRecruiter, are you hiring for your team? Despite current headlines? Several industries like hospitality and healthcare are heading for a hiring boom. No matter what industry you're in, if you need to hire, go to ziprecruiter.com/twit and try it for free. Thanks for listening to this show. As an ad supported network, we are always looking for new partners with products and services that will benefit our qualified audience. Are you ready to grow your business? Reach out to advertise at twit tv and launch your campaign. Now

(00:02:07):
It's time for tw This week at Tech, the show, we talk about the weeks tech news. I have assembled as always, a panel of excellent people who are willing to get together on a Saturday because apparently there's something going on tomorrow. Some big football game. Samal Sam is here. He's rooting for the Michigan <laugh> Steam Boilers or something. I don't know. No. A Detroit Lions. Are you a Lions fan? I am not a Lions fan. No, I I I'm not into pain. I I don't like the sufferers. That would be suffering, wouldn't it? Yes. Yes. Sam is our car guy. I think they last won a championship about 60 years ago. I, you know, I was, I was rooting for him. I l I liked him. I like an Underdog Wheel Bearings podcast@wheelings.media regular on our Tech Guys show. And now on the new Ask the Tech guys and on Twit Social as Sam Ab Bull. Sam. Hello, Sam.

Sam Abuelsamid (00:03:01):
Hello everybody.

Leo Laporte (00:03:02):
Hello. Got lots of car questions for you, but, okay, we'll get to that in a moment. When, first, I wanna say hello to Brian McCullough, host of the Tech Meme Ride Home podcast. You probably listened to him in your car. Hello Brian. Good to see you.

Brian McCullough (00:03:14):
Hello, Leo, as always. Good to see you.

Leo Laporte (00:03:17):
And I guess I'm gonna ask you if anything happened this week, <laugh>, because I don't have much. That's okay. That's okay.

Brian McCullough (00:03:25):
You, I, I've been, I've been filling the good spreadsheet with some, fill it up, some

Leo Laporte (00:03:29):
Things. Fill it up. Hey, I'm thrilled to see Shelly Brisbane here. She is a producer and reporter at Texas Standard. You see her@sixcolors.com and on many incomparable podcasts. Hello, Shelly.

Shelly Brisbin (00:03:40):
Hi Leo. Thanks for having me back.

Leo Laporte (00:03:42):
Hmm. I've been meaning to get you back for a long time, so I'm glad you could be on the show. I'm glad to be here today and all three of you'll be watching the Super Bowl tomorrow. Well, along with me, I've been informed that we have 10 people coming over us, <laugh>, and that I will be barbecuing. So I'm glad I'm doing this show on Saturday for those of you watching. Yes. A rare Saturday edition of this week in tech. If something happens Saturday night, if there's another Chinese balloon discovered, we don't know about it yet, cuz you know, we're off by a day. Actually, there was another Chinese balloon discovered <laugh>.

Shelly Brisbin (00:04:16):
Right? They're everywhere.

Leo Laporte (00:04:19):
How many <laugh>? I don't, I don't really understand what's going on. And of course we haven't heard from the, the Department of Defense what they found. They did recover the balloon they shot down over North Carolina. I don't think we've heard anything about what was on that balloon. Another one was shot down yesterday in Alaska. And I guess there's another one. <Laugh>, we're they're attacking us with balloons. What kind of technology could you put on the balloon that your 26 satellites wouldn't, wouldn't tell you what's going on?

Brian McCullough (00:04:54):
Is it harder to do listening of some kind maybe from outer space? It

Sam Abuelsamid (00:04:59):
Would, it would seem like it would be easier. I mean, cuz you know, a lot of the, the signals intelligence, you know, that stuff's going through satellites anyway. Esp you know, especially, and

Leo Laporte (00:05:11):
It's all encrypted.

Sam Abuelsamid (00:05:12):
Yeah. Mm-hmm. <Affirmative>. So it, it would seem like the satellites should be able to pick that up without too much difficulty. And, and probably more consistently than a balloon that, you know, the balloon, you don't really know precisely what its, what it's trajectory is gonna be. Right. You know, where the satellites are gonna be. And if it's geostationary, you know, it's not going anywhere.

Brian McCullough (00:05:33):
But that's also like the poker tell in this. There must be something incredibly valuable because it's so easy to have been discovered. You know what I mean? Like they, you know, and if it loses control and, and cuz the wind blows it, where it it can be seen and stuff like that. There must be something really, really useful that they're willing to take this risk of discovery.

Leo Laporte (00:05:56):
<Laugh>

Shelly Brisbin (00:05:57):
<Laugh>. Or are we learning about how they discover,

Sam Abuelsamid (00:05:59):
Just trying to, you know, poke the bear and, you know, just trying to get us

Leo Laporte (00:06:02):
Riled up. There's also a possibility, isn't it always that, although why would you wanna do saber rattling? I mean, it's terrifying to me to think of us going to war with a nuclear power. I mean, why would you wanna stir up trouble? And this is a, by the way, it's inevitable that the balloons would be seen. I mean, it wasn't, it's not. Yeah. As you said, Brian. Well,

Shelly Brisbin (00:06:23):
And the reason that we know about it is because civilians saw them. And I don't think the defense department has known about them for a long time, but they haven't been sharing that with us. And that, ah, benefits them from a propaganda point of view as well as just from not letting China know what, what we know. Yeah. But people were seeing them, especially when it was up in Montana. And so there came a point where they were like, okay, we have to talk about it and then we have to shoot it down.

Leo Laporte (00:06:45):
I, I, I'm bringing up cuz I feel like it is, there is a tech story here, obviously, whatever that three bus long electronics panel, it was suspended from the North Carolina balloon. Was it, you know, it was high tech. And had, I mean, I don't know if it was for surveillance or weather measurements. We don't, I don't, we don't really know cuz nobody's said yet. I wish that is it. I guess it's unlikely that the Department of Defense will say, well here's, you know, the bill of goods that was on there. They have banned some Chinese companies who they say that provided technology used on the balloon. So maybe we, you know, kind of do know something. I don't know.

Brian McCullough (00:07:24):
Doesn't, the Biden administration kind of has priors for trying to out what's happening ahead of time. Like, remember before Russia invaded Ukraine, they were saying Russia was gonna invade Ukraine so that there couldn't be a false flag thing. Ah. So if your theory is, is that this has always been going on, but maybe for whatever reason the US government is letting everyone know what's going on now.

Leo Laporte (00:07:47):
Maybe that's it. So six Chinese companies were blacklisted yesterday for supporting the balloon program. <Laugh> the companies, this is from I think Reuters, the companies and organizations which have allegedly supported China's aerospace programs that develop airships and balloons for intelligence and reconnaissance efforts are banned from obtaining US items and technology. They've been added that entity list Huawei is on and, and other companies so that American companies can't buy from them. And they, more importantly probably to them, they can't buy from American companies. I don't see a list of the names of the companies cuz I, that would,

Shelly Brisbin (00:08:29):
I always wonder. Yeah, I would, but I always wonder when that happens, what tech initiative where is inadvertently affected? I'm not saying it's good or bad that they were banned. I'm just saying some companies somewhere was doing business with them, with a band company and is now going, oh, well, we're screwed now at least Right. Temporarily and in a way that you wouldn't expect or that probably wouldn't make the news. Right.

Sam Abuelsamid (00:08:52):
You know at least, you know, in the industry that I primarily cover, which is automotive, increasingly we're seeing the the auto industry trying, trying to migrate away from too much dependence on China. Because they're seeing what's been happening in these other sectors in, in chips and, and in various other things. And telecoms over the last you know, five years or so, five, six years. And they increasingly seem to want to diversify their supply chain away from or their, and, and their customers away from China and not be overly dependent on the Chinese market reason.

Leo Laporte (00:09:32):
Market market quite reasonably. Yeah. I mean, yeah. It's mm-hmm. <Affirmative>, it's probably a matter of time before we get iPhones that are not made in China. Maybe made in India or Brazil or Vietnam. But I think Apple would like to move away as promptly as possible. I did get the list of six companies from Market Watch, the Beijing Naung Aerospace technology company, China Electronics Technology Group Corporation, 48th Research Institute, the Dong, GU Ling Kong Remote Sensing Technology Company. They didn't, they didn't really try to hide what they do, do they? The Eagles Men Aviation Science and Technology Group, the Guang Jot, HAI Shung aviation Technology Company, and the sh xi Eagles Men Aviation Science and Technology Group. Wait a minute, that's the s same name. Different company. Okay. So aerospace companies. Aerospace companies

Sam Abuelsamid (00:10:24):
And, and sensors. And

Leo Laporte (00:10:26):
Sensors.

Sam Abuelsamid (00:10:26):
Sensors, yeah. Sensors is a big one. Yeah. Which

Shelly Brisbin (00:10:28):
Could affect a lot of different industries or not. I mean, I don't, I don't know what sort of sensors they are, but sensors are in everything of, you know,

Sam Abuelsamid (00:10:35):
There's, there's, there's probably a variety of imaging sensors mm-hmm. <Affirmative> in both infrared and, and vi visible light imaging. And probably, you know, maybe things like some lidar you know, some, some high, very high resolution lidar, which is often used in for aerial mapping purposes. So that, that may be one of the things that they're looking at is, is doing some aerial mapping with lidar.

Leo Laporte (00:11:03):
N B C says the Chinese balloon had multiple antennas capable of collecting signals intelligence. This was a state department statement on Thursday, the balloon maker has proven ties to the Chinese military. So we do know now, I think it wasn't a weather balloon that it, that it was some sort of surveillance along, particularly communications surveillance, which explain why you don't do a satellite. Just as you said, it's, it's, you know, that's, it's closer to the ground. Photos taken by high altitude U2 planes. See we still got those, I'm sure the Chinese have those too. Confirm the presence of the equipment, including multiple antennas according to the State Department, likely capable of collecting and geo-locating communications. Solar panels large enough to produce a requisite power to multiple active intelligence collection sensors. The equipment was inconsistent with a weather balloon.

Brian McCullough (00:12:02):
Here, here's a funny thing to think about. Cuz obviously the joke is we're using or they're using such old technology Oh, literally floating balloons. But think about, this also goes back to like, like a naval warfare. You had to have the weather gauge. Like if we wanted to do something similar, if we're like, we're gonna retaliate and send balloons over China, where would, we couldn't float them from the west coast over there. Right. We'd have to float them from Russia. Wouldn't let us do that. Like we Alaska would we,

Sam Abuelsamid (00:12:30):
You would probably launch, oh, now we have the basis somewhere in the Middle East again.

Leo Laporte (00:12:34):
Yeah. Cuz the, the jet stream is, is going west to east. Oh, yeah,

Brian McCullough (00:12:37):
Yeah. Yeah. That's my point. It might be,

Leo Laporte (00:12:39):
That's my point. It has to be somewhere west right. Or east of No, west of China.

Brian McCullough (00:12:43):
Yeah. This is so old school that we can't do it in reverse because the weather, the the wind doesn't blow that way. <Laugh>.

Leo Laporte (00:12:50):
Yeah.

Shelly Brisbin (00:12:51):
Which is why it feels to me like it's, it may not even be that they are after precise pieces of information. They're sort of beyond proof of concept. Obviously they can do it, but you're going to see what you can get because you don't care specifically about where you're targeting. You're, you're floating over Montana, then eventually you're over North Carolina and in between you're gathering a lot of stuff that you can then later assimilate. But it's more important that you be able to get here undetected or un shot down than it is that you hover over a particular place to get a particular piece of information.

Leo Laporte (00:13:28):
Yeah. It seems like a not the ideal way to, to do these kinds of spy efforts, but I guess we, it could go, it can't steer itself. It didn't have propellers. It didn't have rudders. It just could go up and down. Well

Sam Abuelsamid (00:13:40):
You know, is it possible that they have something similar to what Google was doing with their balloon, the balloons? Yeah. Where they could adjust the That's what Google did. Pressure the balloon. Yeah. you know, and, and adapt to the wind patterns, you know, so based on where the wind was, you know, they could go up or down in altitude to maybe catch some wind and have at least some modicum of control over the trajectory of the balloon. Yeah.

Leo Laporte (00:14:07):
I don't know. Anyway, I guess there's a lot more questions than there are answers. I just, it scares me that I don't want to, we don't want to get in a hot war. I don't even think we should get in a cold war with China. No. I guess we should say knock it off and I don't know what, what's the answer to this? It's not good. It's not good. Do we send balloons over China? <Laugh> from, from Turkey or where? I don't know where we'd sign from. Yeah.

Sam Abuelsamid (00:14:37):
Yeah. I mean, I have a hard time seeing why the US would need to, I mean, we've got enough satellites covering pretty much everywhere.

Leo Laporte (00:14:47):
Well, and China has TikTok. What are, don't they know? Don't they know everything? Anyway, it's, it has, by the way, it has distracted certain, in certain quarters, distracted people from TikTok. Right. More worried about the balloons now than they were about TikTok. Maybe that was the point. I

Shelly Brisbin (00:15:06):
Don't know. Well, some of the same people are very exercised about the balloons as our exercise about TikTok. And that's what's unfortunate to me that it immediately became this political thing where if you're on one side, oh, balloons are fine. Right. If you're on the other side, balloons are gonna destroy us. And they had AK 47 s that they were gonna shoot 'em down with. And it's just like, hello, this is National Security. Can we address it on a more substantive level, please?

Leo Laporte (00:15:30):
Yeah. Yeah.

Sam Abuelsamid (00:15:32):
On on the other side, after 20 years in service, the F 22 Raptor finally got its first kill.

Leo Laporte (00:15:37):
Isn't that amazing? That is an amazing plane. But yeah, we don't have anybody to shoot at

Sam Abuelsamid (00:15:41):
It. A multi hundred million dollar stealth fighter. And we use

Shelly Brisbin (00:15:44):
Shut down a balloon.

Leo Laporte (00:15:45):
<Laugh>. Yeah.

Shelly Brisbin (00:15:46):
But imagine the pilot, he's sitting there and there's like, Hey, you're going up today and you're gonna arm your weapons and you're gonna shoot down a balloon.

Leo Laporte (00:15:53):
<Laugh>. Yeah.

Sam Abuelsamid (00:15:53):
Yeah. Granddad, what did you do before? Oh, I was, I was the, I was the first F 22 Raptor pilot to ever get an air to wear kill.

Leo Laporte (00:16:01):
Does he get a little outline of a balloon then on the plane? <Laugh> blow his

Shelly Brisbin (00:16:06):
Maybe a patch that indicates that, yeah. His sl his coat.

Leo Laporte (00:16:09):
You know, you could probably, I would bet that whoever it was is trying to, right now trying to deny it. I had nothing to do with that <laugh>.

Sam Abuelsamid (00:16:17):
It's wasn't me

Shelly Brisbin (00:16:18):
Who's probably not allowed to say it.

Sam Abuelsamid (00:16:20):
Yeah.

Leo Laporte (00:16:20):
I was on leave maybe, but it was a nice shot. I mean, one, one missile boom, the thing came down. Nice shot. Well done. I don't, we don't know how they shut down the balloon yesterday. Do we?

Sam Abuelsamid (00:16:31):
I would assume probably the same. Yeah. Same

Leo Laporte (00:16:33):
Method. Now that we know that that works, I think that was actually, sure. I'm gonna guess one of the concerns of the military is we don't wanna miss, we don't wanna, we don't wanna

Shelly Brisbin (00:16:42):
That would look bad.

Leo Laporte (00:16:43):
Yeah. Like we can't, we can't get it down. That would not be good. So I imagine some of the, some of the delay was them researching it and now that we know we have to take 'em in Alaska from now on interesting. Interesting. I don't know. Yeah. And there, and as somebody here chair saying there's a, there are theories that China did it just to see if we could respond, would respond and had the capabilities shoot it down. Yeah. It was another side, winder X that was used on the balloon in Alaska. Very interesting. I don't know. I feel like we don't know anything and it's, anything we say is speculation at this point. Here's some things we do know the robots are about to take over. This week, both Microsoft and Google announced AI in search. Microsoft started it all on was that Tuesday?

(00:17:38):
Yeah, I think Tuesday they announced Microsoft is going to use open AI's chat. G P t in Bing search. I attempted use it. In order to do it, you have to use Microsoft Edge sign up for Bing to be your default search engine. You pretty much have to give, give your entire Windows machine over to Microsoft <laugh>. And, and I still didn't get access to it, but some have and we've seen some demos of it. What do you, what do you think, Shelly, you've probably, as most of us have played with chat G p T?

Shelly Brisbin (00:18:12):
Sure. And I, I haven't done the Bing thing because I am a minus one Windows machine, so we Mac users are going to be in the dark for a little while.

Leo Laporte (00:18:19):
Oh. But you know what? I think they want you to, I think they want you to

Shelly Brisbin (00:18:24):
Oh, I'm sure they do. Yeah.

Leo Laporte (00:18:25):
Like it change your search engine@bing.com.

Shelly Brisbin (00:18:28):
Sure. I mean goodbye duck dot. Go off to Bing. I I I am going. No I don't know. I I think we've sort of gone from isn't this chat sheet pity stuff Amazing to very quickly, which is what I would've expected to happen, is this is going to be what is, what is it mean to say this is gonna be the next big thing. What business application is it going to have? And the first one is search, which is why so much of the story this week has been the competition between Microsoft and Google and the stories that say Google is wrong-footed even before it had the problem with their demo. Google is wrong-footed relative to Microsoft because Microsoft, you know, has access to chatt P t and Google's trying to to counter them. And so it's sort of a, a weird mishmash of, of stories where some of us are amazed by this technology, some of us, and we run the risk of being called Luddites if we say it out loud. Some of us are concerned about it, and some of us are just writing cool business stories about chat wars and or AI wars. And it's, it's an interesting time.

Leo Laporte (00:19:29):
<Laugh>, Brian didn't, didn't Google call a red alert when they found out that mic, when Microsoft admitted, yeah, we've put another 10 billion into chat G P T and we're gonna use it in search. Didn't this scare Google?

Brian McCullough (00:19:40):
Yes, absolutely. And of course, the ironic thing is that this is all technology that, that Google created years ago and, and open sourced the, the T in chat, G P t I believe is the part of the technology that, that Google transformer. Yeah. So what I have heard from listeners that work at Google that can't be named of course, but also it, this, this was echoed in I think it was TechCrunch had an article a couple days ago about this too. That what happened was, if you remember, there was the time when there was a lot of controversy where, you know, certain AI people in Google were writing papers saying we shouldn't do certain things. Certain people were fired. There was a lot of controversy around this stuff. And also, so, so number one, you have like, this was culturally divisive within Google, so that's why maybe it didn't get productized.

(00:20:41):
Number two, you have the classic innovator's dilemma of, well, would this kill the golden goose? You know, if if you don't follow a link to get the answer you need, then we're not making as much money. But the thing that I heard that was echoed in the TechCrunch piece was that the way that you get money for projects inside Google is to affix it to another big team. Like, so for example, the, the the chatbot team or the search team or the maps team or whatever. And if you think about it, we've seen that the last three or four Google iOS have been little, little bits of AI added to maps AI added to this as opposed, essentially what I've heard is no one had the constituency or the power to say, well, let's go and make this a full product or a full division. And so because of that, they were sort of drip, drip dripping it into other little things, and then OpenAI comes in and just releases the whole product. And that's what they weren't prepared for.

Leo Laporte (00:21:41):
Yeah. Google went through you, you're talking about firing Tim Nick guru, who wrote a paper called Stochastic Parrots with other Google researchers. She was at the, in the ethics division of Google AI <laugh>. And in which she said, you know, this is gonna potentially problematic not only because of limitations and training data and so forth, but also people trust computers. And when a computer comes up with a wrong answer carries more weight than when a human comes up with a wrong answer, she got fired. Then there was Brent lemoyne the the priest, I guess you call him, who was working in with Lambda, who said, oh no, it's gone sentient <laugh>. And then Google immediately fired him because it hadn't, and it was ridiculous to say. So maybe Lamoin had a, had the right idea though. Do you think Google was too judicious, Sam? Do you think they should have productized sooner?

Sam Abuelsamid (00:22:38):
No. I, I, I don't, I don't think they should have productized sooner. And I'm, and I'm not convinced that open AI and, and Microsoft should be doing it now. You know, the, as you mentioned you know, the, the paper that Tim, Nick Gi and and her colleagues did, you know, on the dangers of stochastic parrots, you know, stochastic parrot is defined as a system of haphazardly stitching together sequences and linguistic forms that have been observed in training data according to probabilistic information about how they combine. That's

Leo Laporte (00:23:05):
How, without any how it works,

Sam Abuelsamid (00:23:07):
But without any reference to meaning. Yes. And that is the key is that, that, you know, these, these models don't really understand. They are not intelligent. They're,

Leo Laporte (00:23:17):
They're sophist, they're not even, they're not even trying to be factual in many cases,

Sam Abuelsamid (00:23:20):
Right? Right. They're, they're, they're sophisticated pattern matching mechanisms, but they don't really understand what it, and and this applies really across the board to most AI applications, you know, whether it's, you know, large language models or machine vision systems or, or anything else that none of them are, can we use, we throw around the terms ai, but none of these systems are actually, you know, anything close to human intelligence, the way the human brain works, they don't function the way a brain works. It's a very, very rough approximation. And, you know, from what I've seen, you know, I, I think the best description I've seen of Chad G p t is, you know, that you know, it, what, what was the, the phrase? It's very confidently, very confidently wrong. Yeah. You know, it's a man's pointer, <laugh>, right. It, it puts, yeah, it puts together strings of words in a, in a sequence that looks really plausible. Right.

Shelly Brisbin (00:24:22):
Well, and it's combined with very well done writing. Like, just in terms of, yeah. Not only just grammar and structure, it can make a sentence, but it can make sentences that read well, which is why so many people have applied it to to making a document, write a news story, write an academic paper, write this thing, and it, it can

Sam Abuelsamid (00:24:42):
Write code,

Brian McCullough (00:24:42):
It can write runable

Sam Abuelsamid (00:24:44):
Code. Right,

Shelly Brisbin (00:24:45):
Right. But that's, that's kind of a different issue because the, the, the syntax of the writing is al is one of the things that always sort of alerts you. You know, you've, you've seen plenty of spa emails that come to you Yeah, yeah. From a non-English speaker, or you've seen stuff that was generated by some sort of ai, some sort of computer. And you can, you, you think, you can always tell, you're like, I, I know that's not generated by a human or by a human that had a certain amount of education, but you read this chat g p D stuff, and it, it reads as if it's well constructed. And that gives the authenticity and, and authority of facts that aren't there.

Brian McCullough (00:25:19):
So I a month ago I did an experiment and I shared it on the show where I tried to create two entire YouTube videos from scratch using not only the video selected by ai, but the, the writing made by chat G P T and the actual, the, the, the voice itself, again, not my voice, it was a complete robot voice or whatever, narrowing in on the chat g p t part of it. So I picked the real story of Robinson Caruso, cuz that was based on a guy that really did get marooned. And then the the Crystal Palace in the 1851 exhibition in London, blah, blah, blah. Just two things that I had recently read about. And so I knew enough about. The interesting thing was is that, so like let's say you, you would say, give me a thousand words on whether or not Robinson Caruso was based on a true story.

(00:26:11):
And it would give you a perfect third grade level essay about it. But it would miss all the details that, again, since I had just seen a History channel thing on it, I knew there were more details. So I would have to go in and say, okay, but tell me about how he survived on the island. Okay, now tell me about how, what happened to him after he was rescued. Now tell me about how Ru, how Robert Lewis Stevenson found out about this story. So it was never wrong. And again, think about the fact that I knew enough to know if it was doing the wrong thing, but not, that wasn't even thing that was interesting to me. Cuz obviously, you know, if you don't know the facts you wouldn't know if it was wrong. But to get it to be a good enough sort of story, I also had to know the story.

(00:26:55):
Like it wasn't good enough yet to be as sophisticated as I wanted it to be, just to make a five minute long YouTube video. And so, again, we're talking about people trying to train it to to do things that are factual. It was also like, it, it, I felt like a conductor more than it, it was all prompt engineering. And if you didn't know what you were talking about to begin with, it sure helped save you time to write all the words. But it's not like it did everything without me knowing word one, you know?

Leo Laporte (00:27:25):
But why are we expecting so much of it? It, it's so funny because on the one hand we're kind of like children blown away by the capabilities of this thing. On the other hand, we're going, but it's so stupid and <laugh> and, and why are we expecting so much of it? It's it's a part,

Shelly Brisbin (00:27:43):
It's kind of what I was getting to before. It's because those of us who are interested in tech, and now even those of us who, who might not be, but who have seen this thing do something that's, you know, unadult, it's, it's amazing, right? It seems amazing. We're excited about it. But then you have Microsoft and Google and all these companies that have to productize it. Otherwise why do it? Yeah. And even if Google didn't have the history with AI that it did, they have to productize it because Microsoft is doing it, which is why, you know, we're forcing it into, to search right now. And, and whether search is the best or worst application, they've gotta find some way to, to put it into products, which I don't think is good for us as humans. I don't, you know, may, it may be successful for those companies. It may in fact lead to better ai, however we define it. I dunno whether that means accurate or, or what it means, but it doesn't necessarily mean it's a, it's a good thing for us that we have more ai, but it keeps Microsoft and Google afloat, I suppose,

Leo Laporte (00:28:39):
Even in December. This is a headline from the New York Times, from December, A new chat chatbot is a code red for Google's search business, a new wave of chatbots like chat, G p t use artificial intelligence that could reinvent or even replace the traditional internet search engine. So even back in December, this terrified Google. And of course Microsoft had the announcement on Tuesday. On Wednesday, Google threw together, hastily threw together its own event where they announced they were gonna have their own chat. G p t like AI called Bard, originally apprentice Bard be in Google's search. They didn't say when, nor did, nor did Microsoft by the way. It's not available generally yet, but they did show a animated GIF of Bard answering a question, what new discoveries from the James Web Space telescope can I tell my nine year old about?

(00:29:40):
And it got it come, two of the three answers it provided were correct. But in the third, the third bullet point was, and you could tell your nine year old, we got the first pictures of an exoplanet, a earth-like planet, and another solar system, to which many astronomers, including one astronomer from the University of California, Santa Cruz, said, I took a picture of an exoplanet 14 years ago. What are you talking about, <laugh>? So you gotta wonder, was there nobody at Google looking at these answers or checking? And that's certainly not a good way to announce that your new chat for

Sam Abuelsamid (00:30:14):
They were, they we're using the editors from cnet yeah,

Leo Laporte (00:30:18):
<Laugh>, right?

Sam Abuelsamid (00:30:18):
Yeah. But now I, I think, you know, another, another big issue with using search as the first application for this as the first mainstream application for this is the fact that, you know, AI don't really learn on the, these models don't learn on the fly. You have to run them through your training data. And every time that data set changes, you have to completely rerun that training. That's why when you play with chat G p t, the chat G p T, the G p T three 3.5 model was trained on data back in 2021. So it was current up through the time, up through the moment that it was that they ran the training in 2021. But you know, if you, if you ask it about anything that is from the last two years, it will almost certainly get it wrong. And for a search engine, you know, it has to be current.

(00:31:13):
It has to be kept current in real time all the time. And, you know, this is, you know, all again, also one of the challenges with automated driving where you're trying to use ai, that it has the, the, the models have to be continuously updated. And for something like a search engine that is so vast, there's so much data out there rerunning that training continuously to keep it fresh, you know, in, you know, within minutes or at least hours which is what we expect today from a Google or a bang or duck, duck go you know, when something pops up on the web, we expect it to be found almost immediately. That doing that with, with a a a large language model seems like it would be both impractical and use enormous amounts of energy. And, and time to do that.

Leo Laporte (00:32:06):
Yeah. People are already talking about the environmental impact of this. Mm-Hmm. <affirmative> chat. G P t costs millions of dollars a day. We know that because Sam Altman has said, so he called it an eye wateringly large number. And, and chat G p t when it was first released only was up to date through spring of last year of 20 or no, 2021, even older than that. So to keep it running constantly spidering the web as Google does with its, its spiders would cost kind of an unheard of amount of money. Chat G PT Plus and

Sam Abuelsamid (00:32:38):
Google, Google has gone to enormous efforts to make their spidering mechanisms as efficient as possible, right? To, to minimize the amount of energy that that's consumed for that. And even that consumes a lot. Doing this with a large language model would be far, far worse. Now, there are ways to make AI models more efficient in terms of the, at least you know, once they've been trained you can optimize them to make them more efficient, you know, but then what ends up happening is you tend to lose some, some resolution, some granularity from that, which then, you know, has its own issues. And there's a lot of work being done on making processors more efficient for for processing this kind of data. And I saw some, some interesting stuff at CES this year on processor, you know, AI accelerator chips that they're trying to make trying to really optimize the performance per wat on those. But that's to do that at the kind of scale that you would need for a search engine. I mean, they're nowhere near nowhere near that kind of capability.

Leo Laporte (00:33:46):
There's really two questions, or maybe more than two, but there's at least two. One is, should we even be trying to do this? Is it useful? And, and third is it, is it cost effective? A couple of researchers from a company called Semi Analysis, Dylan Patel and a Fama are quoted in Forbes saying how expensive it would be to, to give Google like, search capabilities to chat G p t. They said it, it would represent a direct transfer of 30 billion of Google's profit into the hands of the picks and shovels of the computing industry. Probably Microsoft. Cuz one of the reasons Microsoft's invested in chat G p t is it's running on Azure. They say deploying current chat G p t into every search done by Google would require half a million, a 100 Hgx servers with a total of 4 million, a 100 GPUs. The total cost of these servers and networking exceeds a hundred billion of, of capital expenditure alone. Just, just to build them, of which Nvidia would get a large portion, I guess they make the A 100 s. Yeah. <laugh>. So NVIDIA's going, they're rubbing their hands with glee. Microsoft's rubbing their hands with glee.

Brian McCullough (00:34:55):
Well, and so get ready for this. Imagine the regulatory issues. Because if, if the only way right now financially you can make this happen is to sign a deal with one of the big cloud computing platforms. And those are also the platforms that are investing in these companies. There was an article I saw this week where people were complaining, I know of academics that want to start companies, but the upfront capital cost is so huge that the only the first conversation they have to have is they have to go to one of the big cloud providers and sign a deal with them. So don't you think that regulators, the Justice Department, would look at that as a certain <laugh> pretty clear cut case of anti-competitive practice? Yeah. If this new industry, everybody has to go to the incumbents Yeah. To, to sign a deal.

Leo Laporte (00:35:46):
But at the, on the other hand, I'm sure regulators want this to come along cuz this is the kind of disruption that can create new startups and

Brian McCullough (00:35:54):
But you could, it couldn't if, if the

Leo Laporte (00:35:56):
Only way you could go to Microsoft in Amazon, right. Google to do it. Yeah. by the way I, you know, it could, could very well be, but Microsoft thought, you know, for a mere 10 billion we can get Google to spend 30 billion on infrastructure. So that I mean, who

Sam Abuelsamid (00:36:15):
Knows? And in fact, Satya Na said something essentially to that effect in interview earlier this week with with Neil Eli Patel after their announcement, you know, he, he talked about you know how what they're doing is, is forcing Google to come to the dance.

Leo Laporte (00:36:30):
Oh yeah. And who's, who's the piper? Who's the Fiddler <laugh>? Yeah,

Brian McCullough (00:36:35):
It's the old Bill Gates thing. We don't have to make a dime on the web browser. We Right. Microsoft doesn't have to make a dime on search, but Google sure does.

Leo Laporte (00:36:45):
Google employees not thrilled about this rushed launch of Bard. In fact, they called it on, there's an internal board message board in Google called Mean Gen. And according to C N B C, which saw some of the messages the board filled with criticisms of company leadership calling it rushed and botched comically shortsighted stock market didn't like it either. Alphabet shares dropped 9%. How did Envidia do this week? I <laugh> should have checked. Interesting.

Brian McCullough (00:37:16):
No, they've been up, they've been up a

Leo Laporte (00:37:17):
Lot. Yeah. Yeah. Un googly according to some of the messages. Microsoft might have pulled a little you know, in football since the big game is tomorrow sometimes you'll line up pretending that you're gonna run in a fourth down situation trying to get the opponents to false start to jump ahead so you can get a little five yards extra before the punt. Maybe Microsoft did a little, did a little a little faint to get Google to leap off the line prematurely.

Brian McCullough (00:37:49):
I think it's, it's strategically brilliant because if, again, let's assume that Google could have done this five years ago, could have done a chat G P T and released it as a standalone product, but didn't, again, for the most obvious reason, which would be they have their golden goose, which is search advertising and this obviates that. So number one, Microsoft has gotten more people to talk about being in the last week than they have in the last five years. And they forced Google to rush products out that they had previously decided. We know they previously decided this isn't probably good enough. And then number three, by they, they've, they've moved, if Google was hoping, if we have to kill the gold golden goose someday, hopefully we can mitigate that over a period of a decade or something like that. So they're forcing Google to make decisions about whether or not to burn the boats sooner than they maybe were ready for. Interesting. Yeah. And again, Microsoft gets nothing, there's no negative to Microsoft out of this other than a lot of money spent on their cloud platform. Yeah. Right.

Sam Abuelsamid (00:38:56):
Yeah. And, and I think, you know, what we saw from Google this week was another, another proof point in, in Mike Elgin's assessment of Satcha or of not me, Sudar Pacha as CEO of Alphabet and Google. Yeah. Mike has said that he's not actually very good at it. Yeah. Michael will be on Twig on the Wednesday. He'd probably take a

Shelly Brisbin (00:39:17):
Little victory. Now, I would be interested to know how the perception would've been different had Google not had this giant failed. Because the first headlines I read about this, and this is not to defend Google by, or defend or attack them for me, honestly. But had they not, the first headlines said Google loses a hundred million when the, when the first in stock, the, the first, that was the first headline. And so it was obviously on the level of, you know, great demo fails that we've seen from companies in the past, including Google a couple of years ago when they made people crazy with the AI-based voice assistant that they were demoing and didn't work out so well. And so I I'm sure that there is a lot of internal concern at Google of the kind that kept them out of chat out of AI previously because of ethical issues. But I also feel like there's nothing like a pile on and, and I think the assessment that you guys have made about Microsoft position and all this makes total sense. But I, I do think it's sort of easy to sort of pile on Google at this point and say, well, they had this terrible demo, which proves they're doing a bad job. And there may be plenty of other reasons to indicate they're doing a bad job.

Brian McCullough (00:40:23):
And, and again, I've heard from everyone that Google actually has the better technology, right? And people are like, it makes chat G P t look amateurish. It's just that they weren't ready to release it. So mm-hmm. <Affirmative>, I, I'm still un until proven otherwise, I'm still willing to believe that Google has the goods. It's just that they weren't ready to put the shine on the goods and present it to the public. I think.

Shelly Brisbin (00:40:47):
Well, there was a good Wired article that sort of gave some of the history of how, how we got chat G P T, and we've talked about some of it where you have this model from 2021 that the CEO of chat, g p t says, of Open AI says let's release it to the public. And an unsuspecting public is all of a sudden able to use this technology. And that's all anybody's been talking about for a couple of months. But it's old technology and it, but it's a great big PR balloon to go back to balloons. It's, it's an opportunity to get into the public consciousness in a way that that company and that technology hadn't been before. Certainly there was Dali and there are all the other things that have been out in 2022 that have to do with images. But nothing like this. Nothing that captured the general public's imagination. And so Google's hand was forced, and, and I don't think it's good for the future of ai and it's probably not good for all the incumbents involved with the possible exception of Microsoft.

Leo Laporte (00:41:40):
Let me ask you this. If if Bing does come out with a chat enhanced search, would you consider changing from Google to Bing, Shelly,

Shelly Brisbin (00:41:51):
Why you stuck Duck go. So what do I know?

Leo Laporte (00:41:53):
Okay. Or switching from Duck, duck go. Well,

Shelly Brisbin (00:41:56):
I guess, I guess for me, I have to figure out what it is that I want from search that AI would give me. The, a lot of the searching I do has to do with my day job. It has to do with finding sources for stories I wanna write. It has to do with finding new stories initially that I can turn into segments for the radio show or for my podcasts. And I, I have to ask myself, other than the sheer entertainment and technology enthusiast perspective, what does AI provide me in terms of search? And that case case hasn't been made to me as an individual. Would I try it? Of course, I would. Would I give it all sorts of fun experiments to do? Yeah. But I don't have a use case for it in my own life. A lot of people are probably not in the same boat I'm in. And I am a sort of a careful adopter when things like that come along. But <laugh>, but I'm not ready to

Leo Laporte (00:42:46):
Change. How about you, Brian? Would you consider changing to a bing.com? Well,

Brian McCullough (00:42:49):
I was gonna say, I don't know if anyone read Joanna Stearns hands on with the, the the Bing, or maybe it was more with the the edge browser version of it. But what she was describing was more the, she like asked it for question. She was interviewing Sacha Adela, and she asked it for questions that a Joister might asks an Adela. And she's like, I didn't, not all of them were good, but there were some good ideas in there that I used. And she said, it's a, it's, it's, maybe we have to think of search in a different way. Maybe it's not search in the way that we're used to it. She said that I'm already using this tool to come up with story ideas. It's not writing the story for me, but it's prompt again, it's prompt engineering. Mm-Hmm. <affirmative>, it's, it's giving me ideas that I'm free to take or not take.

(00:43:41):
It's me as the conductor. I'm not the one performing. I'm, I'm the, the conductor analogy is sort of like, you use this tool to, to do the things and sing in the right direction, but you are still in charge. But as opposed to just put a keyword in, get a link to a page. You can use this in sort of the Star Trek, the next generation computer thing to, to give ideas, to steer in a different direction. You know, the, the, the, the example Microsoft used for the browser was like, if you wanted to craft a response to your boss in an email and it had access to your email. So it says, all right give me the, give me the status of the t p s reports and it would know the whole history of the t p s reports cuz it would have access to your email and it would say to you, okay, do you want a hopeful response? Do you want a skeptical response? Do you want a humorous response? And then you, you, it's up to you to either cut and paste that or whatever, but it gives you thought starters and things like that. Like that to me is something that's a level, that's a type of computing that we really haven't done a lot of till now. How about

Sam Abuelsamid (00:44:55):
You, Sam? I think that's a really interesting, interesting idea. Yeah. I, I think, I think what Brian, I think Brian's onto something there, but I think that there's that brings up another problem though, and that, you know, one, one thing we've seen over the last decade especially is we're, we've become really bad at media literacy, you know, at filtering through and, and picking out what are, what are the, what's real, what's not, you know, picking through the misinformation and picking out the, the, the genuine information that we should be paying attention to. And I think what, what Brian's onto is, is, is a really good idea of, you know, if, if we can use tools like this to generate those prompts in our minds, to give us things to think about, the problem is I think it's gonna require a whole lot of education for everybody in terms of how to use it in a way that is going to be positive and and actually useful to us and not create new problems. Yes. Cause it, it, it can definitely really go off the rails very

Brian McCullough (00:46:08):
Quickly. I agree with you, but sorry, I lo I lost my point and I, and I was, I wanted to to point it to Shelly and give her ideas of where I think it might be useful for her to create podcasts and things like, so another analogy I've heard people use is why do sh TV shows have a writer's room? Because it's like, all right, for this next episode, what are the eight things that could happen to our characters? Or why does a a magazine ha or in the old days used to have an editorial meeting? What are the 10 stories that we could put in the next issue? And so, in essence, I think even Joanna might have said this, that, you know, what it becomes is your own personal writer's room where you can come up with jokes. You, it'll, it'll help you come up with idea story ideas and things like that. And again, well, I hear what you're saying in terms of getting the answers right. Like, you wouldn't want your doctor using this stuff yet, but for now, I am intrigued by that sort of level of usage where it is prompt based, it is idea based, and then it's up to you to sift the wheat from the chaff, you know? Yeah. I

Shelly Brisbin (00:47:14):
Take that point and I, I really do. And I was thinking about, well, an example that came to mind is what if I got a call tomorrow and I, and the caller said you get to interview Sonder Pacha or Sacha Noela, you could interview one of those guys and you have an hour before the interview starts and I am in my own little bubble. I mean, I know how to research. I can, I would be panicking because I, I wanna ask good questions. I don't just wanna ask, so how do you think Microsoft is going to do this quarter? Or Hey, what's up with, what's up with chat e p t I wanna ask good intelligent questions. And I can see a tool like this giving me prompt. I, I start by giving it prompts and it comes back with substance based on not only whatever publication or podcast I might be working for in the sort of the, the, the approach that we typically take, but what my own interests are that it would presumably know and what my readers slash listeners are interested in.

(00:48:10):
And I, I can see that if there's a way in which it's actually really intellectually stimulating to think about how you could ask more deeper questions than you could ask just of a search engine where you're looking for a link in response. And I think that's another aspect of media literacy, where it's not simply that you have to not trust things that are potentially lying to you or giving you bad facts, but that you actually have to change the way you think in order to get the most out of this resource. And that's actually a really interesting educational discipline that we could see down the road that, that I'd be sort of fascinated to be a part of.

Leo Laporte (00:48:48):
Google's always said that this has been somewhat controversial. People search for looking often search for an answer. Sometimes you search for the nearest pizza place, sometimes you search to find out who is somebody. And, but a lot of times they just want an answer. And Google has increasingly given its search results as answers, which has bothered a lot of publications because they're not driving traffic. This is, I did a search for who won the Grammys this year on Google. And it gives you a knowledge graph with the, there's the winners right there, obviously extracted from somewhere. I don't have to go any further than that. It's got an overview, it's got nominations. It also has a knowledge graph on the right that it pulls solely from Wikipedia. And one of the things that's gotten Google in trouble with original sources is well done, right?

(00:49:40):
I don't need to click any of the subsequent links. In fact, good luck getting to the subsequent links. That's a long scroll because Google's already extracted the answer that I want. And Google's answer is, well, that's why people search. I'm gonna show you. I use another search engine called Neva. Let me search for who won the Grammys 2023 on Neva. Neva uses an ai, it's generated this knowledge graph from four sources, B, bbc, New York Post, Huff Post, c n n, about who won, not as complete as the Google answer, but maybe for many qu queries, this would be sufficient. Again, stopping me from going further. Right. Answering the simple question. I think one thing that we're gonna see that Google has not addressed, Microsoft has not addressed is, and this is the problem you've seen with chat G P T and Stable Diffusion and other AI is by scraping original content, it's giving people no need to go any farther. And it's in effect dis intermediating sites like yours, Brian's, you know, like your podcast like tech meme, like, like Sam's podcast, like Shelly's podcast because they've extracted the information people want and they're just gonna provide it right on the search page.

Brian McCullough (00:50:57):
The, the worrying thing is that it obviates the need for search the Galaxy Brain thing. Is it obviates the need for media?

Leo Laporte (00:51:05):
Yeah. And you could see why Microsoft, Google and others might want that. I think they're gonna get a world of hurt from the original sources. And by the way, if you kill the original sources, then you have no f no fuel for your ai.

Brian McCullough (00:51:19):
Well, okay, another Galaxy Brain thing, Leo, everyone is talking about how potentially the internet could be flooded with a ton of spam from this, like me creating some dumb YouTube video.

Leo Laporte (00:51:31):
You could game it. Yeah.

Brian McCullough (00:51:32):
Right. So what if, so right now all of these tools have been based, have been fed on the internet. What if two years from now half of the internet is spam generated by the bots,

Sam Abuelsamid (00:51:45):
Right?

Brian McCullough (00:51:46):
Then what happens to the quality of the bots?

Leo Laporte (00:51:48):
Their feet, then they be gonna a vicious cycle, right? They're source

Brian McCullough (00:51:51):
Material, it's a recursive thing that could happen where the quality would go

Sam Abuelsamid (00:51:54):
Down <laugh>. Well then, then, then, then you, then you either keep going and you end up in Idiocracy or you say, okay, we're gonna, at the point when we turned on chat G p T, we're not gonna trend it on any more new content. Right?

Brian McCullough (00:52:08):
Exactly. Like, so my theory is what if it's like varietals, like the, the, the type of grape, the flavor of grape? What if the business model and or the skillset is, well you can trust our varietal of this AI because we've only trained it on the best medicine stuff. Right? Well

Leo Laporte (00:52:28):
That's, that's what Neva by the way, Neva does. Yeah. Which is interesting. If you ask it about if you say, what do Covid vaccines work? It will, it is already pre-identified certain sources is verified and trusted and will only give you results from those verified and trusted sources. Not another insight. So then, so

Shelly Brisbin (00:52:44):
Then do you end up having, do you end up creating a market for custom high quality ai AI-based

Brian McCullough (00:52:50):
Search? That's what I'm suggesting.

Leo Laporte (00:52:51):
Well, interestingly, Neva is not, but

Brian McCullough (00:52:53):
Actually

Leo Laporte (00:52:53):
Free Geneva's a $5 a month paid alternative to Google

Sam Abuelsamid (00:52:58):
Free. Something like this actually just happened this past week. There was a Twitch channel that was running a completely procedurally generated AI generated signed Seinfeld. Like

Leo Laporte (00:53:12):
Yeah, it was terrible too. Animated

Sam Abuelsamid (00:53:13):
Chill, running continuously. It's not funny. Yeah. That was generating the dialogue and generating the animation. But what happened was they, for some reason that I, I can't recall exactly right now, they had to switch to a different version of G P T from the one they were originally using. And very quickly within, you know, a, a day or so, it might have even been hours it started making transphobic and homophobic statements in the dialogue, and Twitch pulled it down because, you know, you had this varietal, you know, it went, it switched to a different varietal of the AI model that <laugh> just went completely off the rails within hours. Yeah.

Leo Laporte (00:54:00):
Yeah. Not that it's lost is mourned, but Yeah. No, but so really the question, I mean, it's so, so funny because what's happened is these companies have con almost been panicked into rushing this stuff out. They've been holding onto it for a long time. Even open AI was really slow and judicious until recently at releasing its chatbots or its art bots. But all of a sudden there's this land rush and they're panicking, and they're all releasing it at once. What do you think the outcome of this is gonna be? Are we gonna finally all just say, you know what, it's a parlor trick. It's stupid. Forget about it. Or do we buy into it as much as the big companies are buying into it and suddenly Idiocracy, what, what is the outcome of all this? Shelly, what do you think?

Shelly Brisbin (00:54:57):
Well, first of all, I, I think my biggest concern about all of this is not whatever the initial thing is, whether we get some sort of AI generated search out of it, but how, because it's generative ai, how we move forward, and whether that's the spam we were talking about, or whether it's just a new, new normal on the internet. I don't know what that is. And I don't know that we have the ability to sort of control it or, or process it at this point. Have we

Leo Laporte (00:55:24):
Already lost control? Really?

Shelly Brisbin (00:55:26):
Well, I don't know, but, but a year

Leo Laporte (00:55:27):
Ago you said somebody back in time quick, <laugh>

Shelly Brisbin (00:55:30):
Sounds like a plan. Let's write that up. <Laugh>. We'll have the AI write that up, but, but when Dolly came out, it was easy for me to ignore the art stuff. I mean, not, not that I had, I, I had great sympathy for artists who were concerned about it. Right. but it, it didn't seem practical in the same way, once they started talking about writing, then I cared all of a sudden. But it, it feels like it's gone too far in terms of putting it back in the bottle. It, and it, obviously Microsoft and Google are so completely engaged in it. I don't think it's gonna come out the way they necessarily want it to. I don't think it's gonna be like a traditional, well, we're gonna, Microsoft and Google, Google are gonna battle for a while, and then one of them is gonna have supremacy because of their superior technology or their superior marketing or whatever. I, I think it's, I think it's too early to predict it to be on. I think we're in this sort of wild west. How, how many metaphors, can I jump into one sentence? But, but I think we're kind of in the wild, the

Leo Laporte (00:56:23):
Westworld. Are we in Westworld? Is Yu Brenner coming for us? I

Shelly Brisbin (00:56:26):
Think that's it.

Leo Laporte (00:56:27):
<Laugh>

Shelly Brisbin (00:56:27):
<Laugh>, yes. The original Westworld is back. Mul Mul Brenner is after us. It's, it's

Leo Laporte (00:56:32):
Always been our fear. It's interesting cuz it's always been our fear about AI that we would lose control that Whopper or Mule Brenner would, would take over and or, you know, sky net.

Shelly Brisbin (00:56:42):
And I don't, and when I say lose control, I, I'm not even talking about some sort of all encompassing big bad. I think it's, it's easy to sort of knock those worries down because I will, will Luddite for people who don't understand or care about technology are gonna always say something like that. But I do think there are ways in which, and I I think we, we always see it with technology when some, whether it's social media or whatever it is, there are ways in which things grow and mutate in an unanticipated fashion. And I think we're at the stage right now where there's no sort of governors on how this stuff is gonna change and grow and evolve. And I don't mean regulation specifically, although that's an issue but just in terms of like how much is out there and how much technology that is controlled by different people who wanna fight each other for market supremacy.

Brian McCullough (00:57:26):
Can I can I give you a best case scenario? Yeah. Because I, we all know what the worst case scenario is. There's a reason why in the Dune universe computers were outlawed in. You had to have those guys in the tanks piloting the ships. Right? here's a best case scenario for the near term. And I have to credit Mark Andreessen. I heard him say this on a podcast months ago. But like, do you prefer an older doctor or a younger doctor? An older doctor has more experience. A younger doctor maybe is more up on the latest thing. Do you trust when you go to the doctor that they're having a good day or a bad day? Maybe she is going through a divorce and is distracted or whatever. Here's the best case scenario that 10 years from now, you go to the doctor, there's an AI ambiently in the background listening to the consultation.

(00:58:16):
You describe your symptoms, you know, it reads the, the x-rays, et cetera. And it's not gonna have a good day or a bag eight cuz it's an ai, it's la it's up on all of the latest stuff. Even the stuff that was just published 30 minutes ago. And it can give you a 99.999% accurate diagnosis, et cetera, et cetera, et cetera. In the same way that, and Sam, you can speak to this, that in theory self-driving cars are safer than humans. And so the role of the doctor becomes the counselor, okay, this is what the AI says is going on. Here are your options. Here's the option recommended by the ai, but let's think of these other things. And it becomes more of a counselor thing than a diagnostic thing. Yeah. We've

Leo Laporte (00:58:59):
Put doctors again in a terrible position because they're supposed to have perfect memories and be good at Yeah. Bedside manner and handholding. Yeah.

Brian McCullough (00:59:06):
And so again, it comes back to that sort of man computer symbiosis where you're the conductor, you're the prompt engineer, and you, you don't expect the computer to be perfect every time. But your, as a professional,

Sam Abuelsamid (00:59:20):
Your job is to sort of get the best out of that instrument.

Leo Laporte (00:59:24):
This is very Mark Zuckerberg, because of course that's the best case scenario. Sure. <laugh> does it and we're already learning that this doctor could also say, you know, that gallbladder take his foot off and that and be completely wrong, in other words. And well,

Shelly Brisbin (00:59:39):
That's why the mutation scares me more than the concept itself. Yeah. Like I'm, I'm not gonna argue AI good or AI bad, but the mutations are what

Sam Abuelsamid (00:59:48):
We're seeing

Leo Laporte (00:59:49):
That AI could be wrong more, more to the point. And you don't want a doctor to be wrong, Sam, what do you, what do you think?

Sam Abuelsamid (00:59:57):
And, and, and I, you know, I I liked where Brian was going with that, you know, a as a best case scenario, positive. Yeah. I ideal ideally, you know, having, you know, something like an AI as a co-pilot, you know, to, to augment human capability, I think is, is fantastic. I think, I think that's a a that that would be ideal. It, you know, to, because there are all, you know, human memory is, is always flawed. You know, there there are advantages to, you know, things that we gain from human experience and the nuance of the way humans think and process information that I think, you know, in that, that role as the counselor can be extremely valuable. And if you can get the AI to the point where it can filter through the erroneous information and keep that out of it, I think that would be great. But you know, I I'm, I remain unconvinced that we can get there. Yeah. 

Leo Laporte (01:01:01):
But self-driving cars are a good template for this because we do like driver assist AI technologies, right.

Sam Abuelsamid (01:01:09):
To some degree, yes. I mean, even, even those, you know, have their, have very strong limitations

Leo Laporte (01:01:15):
On Well, but the humans never gives up complete control. You always kind of, in case the AI decides to drive into a wall, you always keep some control.

Sam Abuelsamid (01:01:24):
Well, that's the, that's the theory. The, the reality is though, that when you have a system, you know, whether, whether it's just a mechanical machine doing something or a piece of software that's doing something that works really well, most of the time, the, the closer it, the closer that machine or software or me, the closer that mechanism, whatever it is, whether it's hardware or software, the closer that mechanism gets to a hundred percent, you start to get into that uncanny valley where we trust it so much that we cannot be, we as as humans cannot necessarily be trusted as the supervisors to say that, yes, this, you know, this is okay, or maybe this is a better solution. We tend to then give up control to that machine, right? We, we are not good at retaining that level of control that we need.

(01:02:21):
And what we see with with some of the assisted driving systems out there is that the better it works most of the time, then you know, the more likely we are to become complacent about it, right? And then what we actually end up needing is more mechanisms to keep us engaged in the process of supervising the original mechanism, <laugh>. So you end up with something like GM Super Crews, where you have a lane centering and, and speed control system that works really well most of the time, but it requires human supervision. So you add a driver monitor system to monitor the driver to make sure the driver is still monitoring the other system. And it, you know, I'm not sure that it's necessarily improving things, right? Because it, it creates new, new cognitive workloads for, for humans. And, you know, it gets back to what I said earlier, but it requires a whole new kind of education for the people using these tools. You know, whatever, what, you know, in whatever context to understand and monitor what they're doing and to make the best decisions about, okay, is what this is telling me good? Or is this nonsense? You know, and do I need to ignore the, the mechanism and go with my own, my own knowledge, my own intuition? Or, you know, do I just follow what it's saying blindly?

Leo Laporte (01:03:49):
It really is that dichotomy. And on the one hand, <laugh>, we wanna, we want these things to be as smart as possible, trust them as much as possible because it would really be a great thing to lift. Mm-Hmm. <affirmative> the burden <laugh> off of us and let us do things we're good at and not have to do things we're not so good at. And on the other hand it, it looks like these things have the judgment and intelligence of a four year old and nobody in their right mind would trust them to take over anything <laugh>. And we're stuck in the middle right now. And I don't know if it's because of timing or because this is actually the problem inherent in this. Let's take a little break. I wanna take a little break. We've got a great panel and we're not gonna end this discussion because there's a super Bowl ad coming up about this very topic. Sam Bull Sam is here. He is a car guy wheel Bearings podcast. He's also a regular on our shows. We love having him. Thank you for being here, Sam. I appreciate it. From the te Meme Ride Home, Brian McCullough, internet historian and podcast. Always a pleasure to have you, Brian and Shelly Brisbane, you know her from the incomparable and six colors.com. And if you live in Texas, or even if you don't, you probably know the Texas standard. And Shelly is a producer and reporter there. What is the website for Texas Standard?

Shelly Brisbin (01:05:02):
It would be Texas standard.org. Leo, that's

Leo Laporte (01:05:05):
Pretty easy. I should have known that. It

Shelly Brisbin (01:05:06):
Is. Yeah. And yes, our show is available as a podcast. So if you're a Texas ex app, we have a quite a substantial Texas expat community that listens to our show that

Leo Laporte (01:05:15):
Way. So if you're not from Texas, there's not a lot here for you. Is that what you're saying? Saying, well,

Shelly Brisbin (01:05:20):
Come on. I wouldn't say that <laugh>. I would say that if you're a Texas expat, there's more for you. However, if you would like to know why dairy queens in Texas are very different than Dairy Queens in other states, or if you would like to know what our governor has been up to this week, we have got you covered on both of those.

Leo Laporte (01:05:36):
Very Queen is just, I saw Article <laugh>, those of us not in Texas, Texas need to be keeping an eye on Texas just in case.

Shelly Brisbin (01:05:43):
Yes. I, I agree.

Leo Laporte (01:05:45):
If you think a steak finger country basket sounds good, you need to, you need to go to DQ in Dallas or a hunger buster. Hmm. Our show today. Thank you Shelly. Our show day <laugh>

Shelly Brisbin (01:05:59):
Drag queen sponsoring the show today.

Leo Laporte (01:06:01):
<Laugh>. Deq. I wish they were, you know, cuz I could use a hunger buster right now. Or maybe a, what was it? A finger basket? I know. Sounds so good. Steak finger basket steak

Shelly Brisbin (01:06:09):
Fingers. We all, there's a country basket in steak

Leo Laporte (01:06:11):
Fingers. Finger basket don'ts. Yeah.

(01:06:14):
Our sh <laugh> no, our show Today's not brought you by dq This week at Tech is brought to you by Bit Warden. Bit Warden Bit warden. We talked to so many people these days who are ready to leave those other guys after their horrific breach, their lack of concern and lack of communication for their customers. Can I recommend the password manager I've been using for the last two years? Password Manager I trust and love Bit Warden. It's the only open source cross-platform password manager you can use at home, at work, on the go. And Trusted by millions. I think a great many of us have switched over to Bit Warden. I'm happy to say I did that a few years ago. Bit Warden's easy to use it's platform, but more importantly it's open source so you know exactly what's going on. And when there are is a better way to do things, people contribute and Bit Warden gets better.

(01:07:09):
Bit Warden will let you securely store your credentials across your business and your personal worlds. If you're using a personal account free Forever, I might add across multiple platforms with unlimited passwords. You can even store your vault on your own. You can keep it out of any centralized storage, and that's pretty nice. Of course, bit Warden doesn't track data in the mobile apps. In fact, only crash reporting in even that is removed in the OID installation. If you're on Android, bit warden's open source invites. Anybody to review the library implementations at any time on GitHub to review the bit Warden privacy policies@bitwarden.com slash privacy it's the one I trust. I love these guys at Bit Warden Protect your personal data and privacy with Bit Warden by adding security your passwords with strong randomly generated unique passwords for every account. But you can go a step further because they also have a username generator so you can create unique usernames for each account.

(01:08:08):
In fact, if you use one of the five integrated email alias services, you can have unique emails for every account that kind of doubles your security. Not only does a hacker have to dis guess your password, they gotta guess your email works with Simple Login, a non ADDIE Firefox Relay, our sponsor, FastMail, and now Shelly Duck Duck Go. They will allow you to create a masked email address when you can use only once on one website, for example, forwards any email to that primary email account. When you get email at that address, it's a really great way to keep an eye on people, see if they're selling your email address off. Use a unique email address with them. You'll know if anybody else sends you email, they sold your address and it keeps your main email address out of the databases of services and sites you sign up for.

(01:08:53):
It's just an extra security layer and it's great. It works so well. Bit Warden, great for business too. Fully customizable, adapts to your business needs. The teams organization the $3 a month per user for a smaller organization. If you're a big company, you're enterprise plan $5 a month per user. It, it adds new features too, like sharing data securely with coworkers across departments, the entire company, the basic free account though for individuals free forever. For an unlimited number of passwords, you can always upgrade to the premium account. I did that just to support. Bit Warden. 10 bucks a year. Come on man. There's a family plan as well. Gives up to six users. All the premium features for just $3 and 33 cents a month. Very easy to get data into. Bit Warden if you're on another password manager. It's so simple to export that data.

(01:09:46):
Important at Bit Warden and you're good to go. Look, you know, you gotta use a password manager if you're not, you're really at risk. But it's important you choose the right password Manager. Choose Bit Warden. The only open source cross-platform password manager you can use at home. You can use on the go. You can use at work, trusted by millions of individuals, teams, and organizations worldwide. Get started with a free trial right now. Teams are enterprise or free for all devices forever as an individual user@bitwarden.com slash twi. Really there, there shouldn't be any reason not to use a password manager. Make it easy. Make it free. Just go to bit warden.com/twi, tell your friends, tell your family it's just the right thing to do. Bit warden.com/twit. We thank you for their support of this week in tech. We thank you for supporting us by using that address so they know you saw it here.

(01:10:41):
Bit warden.com/twit. Don't you don't wanna put all your passwords in one Steak Finger Country Basket, my friend. You want to spread it across or something? I don't know. That didn't, that didn't really work. I've got my mind still on dq. All right. You know the name, I'm sure. Sam Dan Odod. He is a California entrepreneur, a billionaire who has been buying ads slamming Tesla's full self-driving. He's now gonna buy a Super Bowl ad Now he's not buying the, the full ticket item, the one you see coast to coast nationwide. That's seven, do million dollars for 30 seconds. But he is gonna buy Super Bowl ads in all the major states in the state capitals in Washington DC He wants to speak to legislators. Austin, Tallahassee, Albany, Atlanta, and Sacramento. Now Sam, tell us who Dan Odod is.

Sam Abuelsamid (01:11:41):
So Dan, oh, Dan Odod is the the CEO and founder of Greenhill's Software which is a company that has been making among other things, real time operating systems since their early 1980s. Very secure, very safe system software systems. They run on everything from satellites and military hardware to vehicles a lot. You know, there's a lot of Greenhill software in most vehicles around the world.

Leo Laporte (01:12:15):
Oh, that's interesting. So maybe he has a little competitive

Sam Abuelsamid (01:12:20):
Issue. No, not really. I mean, you know what the, the software that Greenhills makes, you know, is lower level operating software operating system software. So, you know, he's, they're not making the software that runs that, that, that runs. So

Leo Laporte (01:12:37):
It's not competitive vehicles, but he does no, not competitive at all. Does he does want legislatures to ban Tesla style full self-driving. Right? Here's, here's his tweet. Yeah. With video, I guess this is 32nd ad. This is the one that's gonna run on, on Super Bowl in some local markets. Tesla, full self-driving, there's a kid on a skateboard, boom. <Laugh> got him. Mm-Hmm. <affirmative>, there's a, there <laugh>, here's a Tesla hitting a stroller and a crosswalk just driving right through it. Not stopping for a school bus. Elon says the guy's nuts. What do you think Elon's wrong? Elon's, you think this is credible? He's wrong.

Sam Abuelsamid (01:13:14):
Yeah. Yes, it's absolutely credible. You know,

Leo Laporte (01:13:19):
The head off that manager

Sam Abuelsamid (01:13:21):
<Laugh> Yeah. The, the DON project, which is o Dodd's current effort that that is doing this is, you know, they're not the first or not and not the only ones to be creating or to, to be doing this kind of testing and evaluation of different driving assist systems. Luminar, for example lidar manufacturer they've done similar demos to this showing that Tesla's camera only system. The, the pro, the thing that Odod is against is a system like what Tesla is using that is a camera only system that, you know, doesn't have some deterministic components to it. It's not, it's not safe the way Tesla has done it. And, and particularly, not only is their approach to the problem of automated driving and driver assist not sufficiently safe but also the way that they've marketed it and, and sold this capability that they call full self-driving to consumers you know, for up to $15,000 per vehicle. And it is nowhere close to being full self-driving. I've tried the, the FSD beta. It is not capable of safely driving, of safely operating a vehicle without human supervision is.

Leo Laporte (01:14:38):
So what should you call, what should he, what should Elon have called it?

Sam Abuelsamid (01:14:43):
I mean, if you wanna call it something, you know, like driver

Leo Laporte (01:14:46):
Assist?

Sam Abuelsamid (01:14:48):
Yeah. Driver assist system. You know, I mean, there's, there's lots of different brand names out there from various companies. You know, GM has super crews. Ford has blue crews. You know, they're, they're, they're drive, you know, some something that doesn't imply that the the system is more capable than it is because what, you know, what, what's, what, what he is talking about here is a phenomenon known as AANA washing which is a, a term that was coined by a woman named Liza Dixon. She's she's a, a human machine factors researcher. And she wrote a great paper on this cots on this topic a few years ago and coined that term AANA washing. And if you think back, you know, to the late 22, late two thousands, early 2010s, you, you hear about companies greenwashing where, you know, they were doing things that, you know, to try to conjure up this image of being more green, more environmentally friendly than they were AANA washing.

(01:15:49):
The idea is that you are marketing or, or talking about a system that is more capable. You're, you're making it seem as the system, as though the system is more capable of driving the vehicle safely than it is. And and that's problematic because it gets back to what I talked about earlier where as humans, you know, see the system functioning a lot of the time, the more it functions correctly, the more they're inclined to trust it and then to not supervise it correctly. And then when it does hit those inevitable limits of its capabilities, then you get into problems where, you know, it can't properly detect a pedestrian or it can't properly detect how far away you are from a firetruck that's sitting or a police vehicle that's sitting on the edge of the road and runs right into it. Or can't tell the difference between a traffic signal that is sitting on the back of a truck being transported by a crew to somewhere for installation from one that that is actually, you know, at a, at an intersection. So, so, so

Leo Laporte (01:16:55):
I mean, they even warn, here's the disclaimer in the full self-driving. And, and, and this is by the way, from the Dawn Projects website it says it may do the wrong thing at the worst time.

Sam Abuelsamid (01:17:07):
Yeah. This, this comes up, you know, in a Tesla with F S D the, you know, when you launch FSD for the first time when you enable it, that is the actual message I've seen that come up. Yeah. that, that is the message that comes up on the screen

Leo Laporte (01:17:20):
When full self-driving has enabled your vehicle make lane changes off highway select forks, all the good things it does, but you, but you have to pay attention and keep your hands on the wheel, right? Yes. Because it might do the wrong thing. If you watch some of these videos that were uploaded to YouTube of mistakes made by full self-driving, you'd know, keep your hands on the wheel. And who's liable, by the way, if you, if you do blow past a school

Sam Abuelsamid (01:17:46):
Bus? The dri the driver, the driver is, yeah. Tesla does not accept liability for the performance of their autopilot or F S D, which is different from other manufacturers. Mercedes-Benz, for example, last year launched a system in Germany that should be coming to the US later this year called Drive Pilot. It's a so-called level three system where the driver is not required to watch the road. So it's a hands off system and an eyes off system, but you still have to be behind the wheel and ready to take over within about 10 seconds if it's reaching the limits of its capability.

Leo Laporte (01:18:20):
What does Oud want? What does his goal here besides scaring the hell outta me?

Sam Abuelsamid (01:18:25):
Yeah. He, he wants, well, most specifically, he wants regulators to prevent Tesla from selling something that is claimed to be self-driving that is not, he is

Leo Laporte (01:18:37):
Tesla is banned in some states, I think California from, from calling it full self-driving, right?

Sam Abuelsamid (01:18:43):
Yeah. They, California passed a law, but they haven't, doesn't, don't seem to have actually done anything to enforce it yet.

Leo Laporte (01:18:49):
<Laugh> Okay.

Sam Abuelsamid (01:18:50):
So we'll sit

Leo Laporte (01:18:51):
Wants the National Highway Transportation, say, and Safety Administration to ban F S D ban Test

Sam Abuelsamid (01:18:57):
Fsd. Yes. yeah. And many of us have been saying that for many years. Yeah. that, you know, that they should not be allowed, you know, and both nhtsa, you know, which is the regulator to co governs automotive transportation safety. And the Federal Trade Commission should also be stepping in and saying, look, this is false advertising. This is not self-driving capability. And you, you cannot sell, you cannot sell this as self-driving if it is not actually capable of driving the vehicle safely by itself. Yeah.

Leo Laporte (01:19:27):
And Shelly I presume you don't have a Tesla?

Shelly Brisbin (01:19:31):
Do not. I was interested in this though because our, our legislature here in Austin, Texas, which means every two years for about six months, has just gone into session. Ah, and so,

Leo Laporte (01:19:42):
So they might see the Super Bowl ad tomorrow. I,

Shelly Brisbin (01:19:44):
I suspect they will. And also keep in mind that Tesla has a gigafactory here in Austin, and Elon Musk has been embraced by a lot of the Texas political leaders because of his bringing so much of his empire. Not only Tesla, but SpaceX, his here, the boring company. Most of, most of what Ilan is doing has some Texas and Austin's specific footprint. We also, the GM cruise r taxis here mm-hmm. <Affirmative> in Austin. And those have had some problems too. I understand. There, there are investigations into how those cars operate. I understand when they initially started here, they always had a driver. I believe there might be some. I, I, I don't know much about the how autonomous they are. Sam probably knows a lot more about it than I do. But there have been some specific issues here in Austin with it backing not, not staying in lanes properly and not parking properly. And there haven't been injuries that I'm aware of, but there have been a lot of concerns here in Austin on the local level. And here in the, here in the Capitol city of Texas, our local and our state leaders are often on opposite sides of things. We have a fairly progressive local government. You're in the

Leo Laporte (01:20:54):
People's Republic of Austin.

Shelly Brisbin (01:20:56):
Sure. The blueberry in the sea of red as they often say, <laugh>. And what that means is that sometimes what we're seeing in Austin, whether it be through what's going on with Tesla, with, with full self driving, or whether it be what's going on with crews it doesn't necessarily mean those concerns are gonna percolate to the legislature in the same way that they might. So, but there's not, as far as I'm understanding, and we've done a lot of coverage of what's coming up in the legislature so far, there isn't really anything filed or expected to be taken up that has to do with regulation of self autonomous driving, or specifically with Tesla and self-driving and given Musk courting a successful courting of the statewide officials. My guess is a lot won't come of it, but I

Leo Laporte (01:21:41):
Imagine they also are a fan of what Elon's doing on Twitter. He's made even more friends

Shelly Brisbin (01:21:48):
Probably. I, I think, I mean, they're, that's sort of the, that, but they don't have a direct, I mean, it's more on the sort of purely

Leo Laporte (01:21:53):
Economic political side. Yeah. As I understand, you know, jobs, baby

Shelly Brisbin (01:21:56):
Things. Yeah. So it's purely on the political side in terms of Twitter, in terms of the job situation. Spacex is very popular among elected officials, but it's not very popular in Boca Chico where there are a lot of environmental impacts. So there's a, there's a lot of back and forth in terms of how people regard Elon down here.

Leo Laporte (01:22:15):
Yeah. I had a Tesla Model X didn't have full self-driving, but I knew enough not to trust it, <laugh> even to do a lane change and frequently had to take control of the wheel. I have a Ford Mustang with Blue Crews right now, and I theoretically can take my hands off the wheel, but I'm still keeping an eye on things. I'm still <laugh>, I'm still holding onto the wheel just in case. No issues yet. How about you Brian, do you drive a, do you have a Tesla? You have experience with full self-driving?

Brian McCullough (01:22:46):
I actually rented a model y over the holidays. And, you know because I live in New York and, you know, haven't owned a car for almost 15 years. All of the things when I get in a car these days that I'm sure are common, like, oh my God, look at huge touchscreen or still wow factors to me, but just, just the, the adaptive cruise control was like, oh, yeah, that's all I need, you know? So like, really, I know I love adaptive

Sam Abuelsamid (01:23:18):
Cruise control. I'm a fan. Yeah. That's, that's all most people need. Yeah.

Brian McCullough (01:23:22):
Yeah. And, and, and, you know, frankly, Sam tell me how far away, this is all I really want. I don't need the self-driving in towns, cities or whatever, but if I could read a book, I'll still stay behind the, the driver's wheel or whatever. If I could watch a movie on that giant touchscreen while on the highway mm-hmm. <Affirmative>, that's all I need. Yeah.

Sam Abuelsamid (01:23:42):
You know, and, and that's, that's exactly what the Mercedes Drive pilot system does. And there, there are other systems coming over the next couple of years that, that will have that capability for automated highway driving because that's, you know, in, in the world of automated driving they, they have use a term called operational design domain, which is what are the limits of where the system can operate. And you know, unfortunately, you know, one of the things that Tesla has done wrong, I think is say, you know, the O D D is everywhere. You know, you can, they're not putting any restrictions on where you can use it. Things like Mercedes Drive, pilot gm Blue Crew, or GM Super Crews, Ford Blue Crews, they're limited to operating on divided highways. And so you know, we're, we're going to see systems like this that allow you to take your hands off.

(01:24:34):
You still, you can't take a nap. You know, you've gotta be still awake and, and ready to take over within a reasonable period of time, but you'll be able to read or watch a video while you're, while you're cruising down the highway. And then, you know, when it comes, you know, when you set the navigation, when it comes to the, the off ramp, you know, then it'll give you an alert, you know, 10, 15 seconds before and say, Hey, get ready to take over control, and then you'll do that so that, that is going to be in cars over the next 2, 3, 4 years. Oh, and one other thing too, what Shelly was talking about with the cruise robo taxis in in Austin. Yeah. They are, they, those are operating without safety drivers, safety drivers in them. They started doing that in San Francisco in June of last year. And then in December, this past December, they expanded that operation to both Chandler, Arizona and Austin. But currently they only operate like that at night between about 10:00 PM and 5:00 AM and only in a relatively small portion of the city as they, they gather more data and, and learn and, and get to a point where they have enough confidence to start expanding that operational area.

Shelly Brisbin (01:25:47):
They're in my neighborhood. I've, I've, in fact, we were out pretty late last, last week, and we saw one, and it was a weird sight. I was like, wait, is there somebody in that car? I don't

Sam Abuelsamid (01:25:56):
Know. I, I've, I've ridden in them, and I mean, I, I, I've, I've seen the reports of some of the issues, and yeah, they're not, they're not perfect. But you know, in, in my experience with them, they, they, they were, they worked very well. They worked better than most Lyft or Uber drivers. I was in

Brian McCullough (01:26:11):
San Francisco for the first time and about four years, two weekends ago, and at Market Street after 10:00 PM it's a little weird. And there was, there was one every other, every five minutes. Yeah. <affirmative>. They were all over the place. All over the

Leo Laporte (01:26:22):
Place. Shelly I know you have low vision. Do you use a, a cane, a white stick?

Shelly Brisbin (01:26:28):
I have one. I only use it when I'm in unfamiliar areas. And yeah. As somebody who, who does not drive, cannot drive the idea of self-driving auto autonomous cars has always been exciting to me. And to be honest, I would get in one of those cruise rob taxis because I'm, I'm just like, I would like to be driven around by a robot. It's freedom. And imagine, well, yeah. Absolutely. It's, I think

Leo Laporte (01:26:49):
Same thing as I get

Shelly Brisbin (01:26:50):
Older, and I, and the problem, you know, the problem with the problem with whatever Tesla's doing and whatever is that there are humans interacting with, you know, we talked before about the systems of talking, when we were talking about ai, we talked about how humans are involved. Well, in theory, if you had autonomous vehicles, if the, there were only autonomous vehicles on the streets and we didn't have those pesky humans, pedestrians and transit users and all that stuff, in theory, the system would be a better for all. In theory, I, I'm not saying that it's true. Do you

Leo Laporte (01:27:21):
Ever worry her as a pedestrian about getting CED

Shelly Brisbin (01:27:25):
By? Sure. I also worry, I also worry about electric cars that don't have motor noise. Right. So there's just plenty of stuff to worry about as a pedestrian. I, I worry about human drivers just as much or more, more maybe mean. The only time I ever, the only time I ever made physical contact with a car while I was walking, it was driven by a human. So yeah. I wasn't knocked down, but it was very scary.

Sam Abuelsamid (01:27:45):
Yeah. Yeah. And, you know, to, to your point Shelly, you know, I mean, cases like yours are one of the, the, the driving forces for automated driving, you know, to enable mobility, freedom of mobility for people that either can't or don't want to drive, you know, whether it's because of a physical limitation or you know, just getting older or the young or, or people who just don't want to drive. You know, having that ability to call up a vehicle to take you where you need to go not have to worry about parking and things like that. I think, you know, there's, there's a lot of potential societal benefit out of that. And if we can get it right, you know, the, the potential improvements in safety, you know, these things should be able to see pedestrians way better than humans can, because Sure.

(01:28:33):
You know, if they're, if they're done right, you know, which it means not like Tesla which relies only on simple RGB cameras visible like cameras, you know, the, those cruise vehicles, Waymo vehicles, others, they use a combination of cameras and radar imaging, radar sensors, high resolution radar sensors, LIDAR sensors, thermal imaging sensors that should be able to help, especially help detect pedestrians and, and animals. You know, all of this, this combination of these sensors is what's key to try to generate a sufficient level of safety that is hopefully at least as good, at least as good or better than human safety. Which we haven't yet fundamentally proven that, that they are even as good as humans yet.

Shelly Brisbin (01:29:19):
Right. And it's fair to point out the pedestrians and animals are as unpredictable Yes. Or more so than other vehicles on the road. So you can count on certain things with software about how other vehicles behave because there are rules governing them that doesn't always work, but it's less effective. But the pedestrian who might dart out in front of you, and I, I think that there, there have been incidents where when they traced it down the car, the automated car was at fault in terms of hitting the pedestrian that they should have seen. But the pedestrian or the bicyclist or whoever is also not necessarily guaranteed to behave in a way that the car expects that the software that's running the car expects.

Sam Abuelsamid (01:29:56):
Yeah. That's a, that's actually turns out to be one of the hardest parts of developing automated driving. You know, you can, you can kind of break down, break down the problem into four parts, sensing, you know, detecting what's around the vehicle prediction, then planning a path through that environment, and then actually controlling the hardware to do it. That prediction turns out to be the really, one of the toughest parts, especially with pedestrians and cyclists, vulnerable road users. Because, you know, physics, the physics of a, of a vehicle mean that it, it's gonna put some limitations on how fast it can change direction. A human or an animal or a cyclist can change direction on a dime. And that is really hard to figure out for software.

Leo Laporte (01:30:38):
Oh, that's disappointing. Cuz the main reason I want self-driving cars is so I can ride my bicycle more <laugh> <laugh>, well, <laugh> that dream up in smoke as well, although as, as I get older, you know, I would love it that I don't have to I guess my wife will drive me around, but until then or maybe after that self-driving vehicles be great. I guess the other, the problem really is that we have self-driving vehicles and human driven vehicles on the same road, would it?

Shelly Brisbin (01:31:04):
Right. That's fundamentally what what I see is the issue because I, I wouldn't, the rules about how you arrange streets and how traffic worked

Leo Laporte (01:31:13):
Fundamentally, cars different. Yeah. It's all cars, vehicles. Yeah. Yeah.

Shelly Brisbin (01:31:16):
We're, we're automated. Yeah. And if we, you know, I, I get on my my sort of red hat here but if we didn't have to have private car ownership, if I could just rent access to a car when I needed and send it away, I wouldn't have to have a driveway. I wouldn't have to have a parking space or garage. Right. If I wanted a pick up truck to take a load of stuff somewhere, but never wanted to see that truck again, I could do that. And as a person with low vision, that's pretty attractive.

Sam Abuelsamid (01:31:40):
Yeah. Yeah. Well, and, and for, for society as a whole, you know, there are seven and a half parking spaces for every vehicle in the United States. There's 290 million registered vehicles, and there's seven and a half parking spaces for every one of those <laugh>. In, in urban centers,

Leo Laporte (01:31:55):
Entire areas of the city are devoted.

Sam Abuelsamid (01:31:57):
Yeah. Urban centers, about 25 to 30% of an average urban center Yeah. Is de devoted to parking. If you could eliminate all or most of that, imagine what you could do with that land mass in cities to provide more housing more, more commercial real estate, more, more of everything that would be of benefit to society in addition to the, the potential for saving lives and saving energy from these vehicles. Yeah.

Leo Laporte (01:32:23):
I'm not, the

Shelly Brisbin (01:32:24):
Way I always say it is if I, I have in my past, I've been a cube dweller in, in office environments the car that I don't have, but if I had a car, it would have more space than I would've had in some of the cubes I've been in.

Sam Abuelsamid (01:32:35):
Yeah.

Leo Laporte (01:32:37):
PC shipments, ooh, this is bad. An analyst working for a company called Mercury Research. Dean McCarran has put out a report that says this, the, the X 86 processor market just endured, quote, the largest on quarter and on year declines in our 30 year history, the worst quarter ever for X 86 processors.

Brian McCullough (01:33:08):
And look up a recent article about the memory chip industry, which we know is famously boom and bust, but they've apparently had the worst bust that they've ever seen in terms of inventory, price declines, things like that. It's

Leo Laporte (01:33:22):
Funny because wasn't, it wasn't, weren't the prices going up for a while because they couldn't make enough and now they've made too many. Isn't that what always happens?

Sam Abuelsamid (01:33:32):
Pretty much, yeah. Yeah. 

Shelly Brisbin (01:33:34):
It was interesting that apparently server class chips are an exception to that. And I, I wonder in my non macroeconomic brain, whether that has to do with people replacing servers after the pandemic that they didn't during that time.

Leo Laporte (01:33:48):
<Laugh>, yeah. This was good for AM md apparently it had, again, in server CPU share one of the only segments that saw growth in the fourth quarter of 2022. Why is this, Brian? You think PE people, why aren't people buying PCs?

Brian McCullough (01:34:04):
Well, everybody did. It's,

Leo Laporte (01:34:07):
They already have

Brian McCullough (01:34:08):
<Laugh>. Right, right, right. Everybody, everybody

Sam Abuelsamid (01:34:10):
Needed 3, 4, 5 years.

Brian McCullough (01:34:12):
Yeah. Yeah. Because another one, yeah, we're it's the pull for, it's the same thing as e-commerce. It was the pull forward of demand, and then people were like, Ooh, new reality. And then the, the problem that the PC industry and the chip industry has is that then it, it's the rubber band sort of pulled back sooner than you could have a replacement cycle, a natural replacement cycle happen.

Leo Laporte (01:34:38):
So are we gonna see a recovery at some point in the future? Like, will people go back and buy PCs again? Or,

Brian McCullough (01:34:47):
I mean, the, so again, pandemic times, you've gotta have a tablet or a laptop for your kids. You've gotta have an extra one. You know, your work furnished you with one and this, this, that, and the other thing. And then think of, think of the, the snapback for if remote work doesn't happen as much as people thought, you know? I don't, there's, there's so many factors in terms of the, you know, you had the supply chain things as well, and now you've got fears of recession dampening consumer demand. It's like, it is, it is one of those perfect storm things where you could point, you, we could sit here and talk for a half an hour, about four different points that are hitting the PC industry all at once and, and certain a corners of the hardware industry all at once. But all it really is, is that this was a sort of black swan event that no one could have planned for people over-planned or got giddy about, oh, this could be a, a phase change in terms of how consumers or even enterprises behave and, and what their demand could be. And so it's just gonna take time to unwind that. And you know, in theory, like inflation is coming down and supply chains are getting back to normal, but they're, maybe it's just gonna take longer for, for tech hardware.

Leo Laporte (01:36:12):
Very interesting. All right, let's take a little break. Shelly Brisbane is here from six colors.com and the Texas standard@texasstandard.org. Great to have you here. She's also on, I didn't even know he had one. Jason Snell's maid on instance, which has the marvelous name of zeppelin.flights, and she's Shelly on there. Zeppelin flights. I have to ask Jason about that on Tuesday on night break Weekly. <Laugh> also here, Brian. Well, I

Shelly Brisbin (01:36:41):
Can tell you why it's called that if you care. We,

Leo Laporte (01:36:43):
Why, why is it called that?

Shelly Brisbin (01:36:44):
Well, so Zeppelin is the sort of mascot of the incomparable network, which is Jason's network of podcasts about pop culture and movies and TV and superheroes, all those

Leo Laporte (01:36:54):
Things. The incomparable has Zeppelin is Zeppelin for logo kind of. Yeah.

Shelly Brisbin (01:36:57):
Oh right. And so we could just decided to call his Zeppelin Flex.

Leo Laporte (01:37:00):
That makes perfect sense. Besides Zeppelins are super cool.

Shelly Brisbin (01:37:05):
They are,

Leo Laporte (01:37:06):
Yeah.

Shelly Brisbin (01:37:07):
Someday. Jason's gonna let me do a a show on this is, this is just, I'm near casting to Jason. If he's out there somewhere. I wanna do a show for the incomparable about movies featuring Zeppelins.

Leo Laporte (01:37:19):
I would listen to that.

Shelly Brisbin (01:37:20):
Alright. Right. There you go.

Leo Laporte (01:37:20):
See there's a, there's a, is that

Brian McCullough (01:37:22):
An audience?

Leo Laporte (01:37:23):
Yesterday we had science fiction Arthur Daniel Suarez on triangulation. And there's a scene in his new book, critical Mass where these billionaires are need to fly around and they don't wanna pollute the atmosphere with their private jets. They have a private Zeppelin. And it's so cool know. Is

Shelly Brisbin (01:37:43):
It a self-driving private Zeppelin? No,

Leo Laporte (01:37:45):
I think it's, I think there's a pie. There's a captain and there's white gloved stewards and there's silverware and it's very luxurious. They fly to Nigeria over a period of a day and a half from I think from Sweden. And it sounds very co, very sophisticated, very beautiful and very leisurely.

Sam Abuelsamid (01:38:03):
They, they tried that once.

Brian McCullough (01:38:04):
No, but yeah, that, that was the hin.

Leo Laporte (01:38:06):
But don't use hydrogen. That's all I'm just saying. Right.

Brian McCullough (01:38:10):
I Leo, I'm with you. I have never understood why that hasn't come back, aside from the blowing up part, but than that, how many other technologies have had lots of crashes and blowups and sinkings and yet we still went further with the technology. Why this one technology? Did we have a couple of bad <laugh> events and we just abandoned

Sam Abuelsamid (01:38:32):
It because airplanes were faster.

Leo Laporte (01:38:35):
Unfortunately that's probably too, yeah. Serge b Brin, the one of the founders of Google who has infinite money had a Zeppelin company, an airship company. Really? Yes. In fact, you know, those big, if you've ever been down by Google down in Silicon Valley, there are leftover blimp hangers, I think from, must be from World War ii. He at leased those <laugh> and was, and was building blimps in there. But unfortunately I think that was another of the, of the victims of Google's firings and, and shutdowns. Cuz I think that they, they're not doing it anymore.

Sam Abuelsamid (01:39:16):
You know, the the current, the current generation of the Goodyear blimps that, that they introduced, I dunno, five or six years ago, I think those are actually not blimps anymore. They are in fact, Zeppelins and the, the Zeppelin company is still in business. Oh. They never went away. And the, so if you see a Goodyear blimp around these days, it is in fact a Zeppelin. Cuz the difference between a Zeppelin and a blimp is a blimp is basically just a balloon with a the pod underneath it. I forget what it's called. Gondola. yeah, gondola. Yes. Thank you. And a Zeppelin actually has a rigid frame

Leo Laporte (01:39:53):
And you're inside the balloon.

Sam Abuelsamid (01:39:54):
Inside. Yeah. And well, and I mean, they still have a gondola. Oh. but they, the, the, the balloon itself has a rigid frame.

Leo Laporte (01:40:01):
Is this a Zeppelin that's, or a dirigible or a blimp? What are we looking? Well,

Sam Abuelsamid (01:40:06):
A a d a dirigible is the generic term that encompasses both, so that's actually, that looks like it's actually a Zeppelin. This

Leo Laporte (01:40:12):
Is Larry's

Brian McCullough (01:40:13):
On to learn things. This is

Leo Laporte (01:40:15):
Really, this is Larry, I mean Sergey's company. This is LTA research.com, and that is one of the big blimp hangers in Mountain View. And there is the blimp I guess they're not outta business. They, they says we're ex, we're united in the belief that next generation air ships can compliment humanitarian aid and Rus the carbon footprint of aviation. So, you know, if you got unlimited funds, do something crazy and imaginative here is a 3D model of the Pathfinder one, which is their their airship. Wow. So they, they've avoided the whole blimp ible thing by calling it an air ship. Right. But the gun going, yeah. Yeah. It's got a flyby wire system.

Brian McCullough (01:41:06):
<Laugh>, I feel like we, we saw a bunch of stories five or six years ago about also this has industrial uses for like flying things into remote areas that, you know, a heco like would be too heavy for a helicopter or even like a seat, whatever, you know, like, so like you could,

Leo Laporte (01:41:22):
It's got lidar baby.

Sam Abuelsamid (01:41:23):
Yeah. The, the pay, the payload capability is more than you could get with a he helicopter. It's, it's huge. Yeah. Yeah. They're

Leo Laporte (01:41:29):
Very long, but they're slow. Right. That's the drawback. Yeah. Mm-hmm.

Sam Abuelsamid (01:41:32):
<Affirmative>. So it, it, you know, it depends on, I mean, there's a lot of applications where speed is not necessarily of the essence. And, you know, these would be a great, these are a great solution for those applications.

Leo Laporte (01:41:42):
I think. Jason Snell should sell his his Zeppelin site <laugh> to Larry or Serge. Wow. also here, Brian McCullough, TE Meme Ride Home podcast, if you've been doing it how many years now? Five years, six years?

Brian McCullough (01:42:01):
Yeah, actually March 5th. So I gotta think about that. It will be five years

Leo Laporte (01:42:06):
Has has, has the news changed over five years?

Brian McCullough (01:42:09):
That's so funny. I have said that on recent shows where, you know, for how long, the first few years it was tech's ascendant and tech is conquering all before it, and tech is gonna take over the world. And the last six months have been so, so many episodes where I'm like, Hey, here's another narrative breaking, here's a layoff story. Yeah. Here's, you know, like it, the, the, the triumphalism of the first three to four years has really been, you know, record scratched in the last year.

Leo Laporte (01:42:43):
I, you know, I mean, I've been doing this long enough that I've been through the bust in 2000, the bust in 2008, now, the bust in, in 2023. And it always seems to come back, but it is definitely it's, it's interesting to watch the com the ebb and flow of our fortunes, isn't it? Yeah.

Brian McCullough (01:43:06):
Yeah. And like I said, I mean that maybe that makes it better for me because it's not the same narrative. Oh. But

Leo Laporte (01:43:13):
That's why I like it. It's not boring. Yeah. I mean, who knew that we would spend so much time talking about balloons and blimps on this episode, right?

Brian McCullough (01:43:21):
<Laugh>? Well, or, or like the, the, the, the you know, now we have a new horse race, like going back to the, is it gonna be Google versus Bing? Yeah. You know, like, yeah, yeah, yeah. So the, the, I, I feel like the first three to four years of the show, the pandemic, notwithstanding the narratives that I was covering it in the tech industry were sort of stuck in Amber and the chess board is sort of been thrown up in the air this year.

Leo Laporte (01:43:44):
Good thing to remember, if at any point we get a little bored, sometimes I do got sick of talking about Elon, there'll always be something new, different and interesting to talk about Tech meme Ride Home. You can get it right@techmeme.com or subscribe in your favorite podcast player and hear the news change right before your very eyes. Sam at Bull Sam is also here. Our car, my car, my personal car guy. <Laugh> Wheel Barings podcast is@wheelbearings.media. Robbie's on that too. I love him. We've been trying to get him on Twitter. And Nicole, don't forget Nicole. Nicole too. Great show for people who love vehicles, motor vehicles. Our show today brought to you by ZipRecruiter, as the world ebbs and flows, jobs come and go, you may wanna know about ZipRecruiter because there's gonna come a time when you're hiring. And of course those vacancies always come along at the worst time.

(01:44:42):
You know, for us, we're a small company, somebody leaves, that means we all have to work a little bit harder. And and we've gotta scramble to fill that position. Thank goodness ZipRecruiter here. We use ZipRecruiter all the time to do our hiring. And, and believe me, there are plenty of industries that are hiring these days. There's in many industries a shortage of talent. But don't worry because no matter what industry you're in, when you need to hire you go to ziprecruiter.com/twit. You could try it for free. It works so well. We've hired some of our best people there. When, when our wonderful Ashley got a better job, she was in our continuity department, and Lisa and I are going, oh, what are we gonna do? What are we gonna do? Lisa went to ZipRecruiter, posted the job opening right there. And it's fast because you're posting to more than a hundred job sites with one click of the mouse.

(01:45:34):
But it's also fast because ZipRecruiter looks at your requirements, then looks at all the resumes. They have more than a million current resumes on hand and will say they'll pick out people who match those needs and they'll say to you, Hey, look, here's 10 people who really fit the bill. You decide whether you want to invite them to apply or not. But I'll tell you this too. When you invite somebody to apply, they're so flatter, they're so excited. It's a great beginning to a potentially fabulous experience. When we post on ZipRecruiter, we usually get somebody great within an hour or two and least they'll post at breakfast by lunch. She's going, oh, here's another one, here's another one. And that is such a relief when you're a person down. It really is great. That's how we found Viva. Ziprecruiter uses its powerful matching technology to find candidates for practically any role.

(01:46:24):
That makes it very easy for you to process those applications. Cuz they don't go to your voicemail or email inbox. They go into the ZipRecruiter interface, they reformat all the resumes, so it's easy to scan them. You can create screening questions, true, false, multiple choice, even essay to eliminate people who just don't fit the fit the bill. You can narrow it down, rank 'em, screen 'em, rank 'em, and hire the right one fast in one place. It's so great. In fact, four out of five employers who post on ZipRecruiter get a quality candidate in the first day. That's always been our experience. It really is a relief. You go, oh, whew, we're not gonna have to operate, you know, at half speed. Find quality candidates, fast, zip recruiter. Keep your team growing strong no matter what industry, no matter what field, ZipRecruiter's great. Go to ziprecruiter.com/twi you could try it there for free.

(01:47:19):
Ziprecruiter.Com/T w I t. They have been so great for us all these years. We've used them many, many times. They've been a sponsor here for a long time. Thank you, ZipRecruiter. We really appreciate it. And thank you to all of you who listen and use those addresses because that makes a big difference. And they know you saw it here. Ziprecruiter.Com/Twit. We had a really great week this week on all the shows. I mean, not just Daniel Swarth. Well, you know what? I think we have a little video you can watch for yourself only on Club Twit

Leo Laporte (01:49:22):
Hey, there's only room for one AI host on the Twit network. That, that my voice in that whole promo was artificially generated. That's the 11 labs version of, of Jason Howell and my voice.

Brian McCullough (01:49:37):
I've done that. I did that this week as well. I

Leo Laporte (01:49:39):
Have to, yeah, it's kind of tempting, isn't it?

Brian McCullough (01:49:41):
Mm. Well, I, I have to take a trip to Ireland. Well have to. What if I did, instead of hiring somebody to Exactly fill in for me for a week? Exactly. What if I just wrote out the show and didn't have to edit it and pro just, yeah.

Shelly Brisbin (01:49:59):
So

Leo Laporte (01:49:59):
I it's not there yet though. Is it <laugh>?

Brian McCullough (01:50:02):
No, but it, it's, it's actually the 11 labs is the closest one. Cuz I did the descript one a couple years ago and it was nowhere near close. Yeah. Script.

Leo Laporte (01:50:09):
The nice thing about script was it was clearly generated. That one was close enough, it just sounded like, was Leo sick? But this,

Brian McCullough (01:50:17):
This adds like the breaths. Yeah, this adds the, it's pretty good, sort of the emotion. Like, so if they could just add the dials where I could add emphasis and, you know, with italics or something like that, we're almost there, but have, I don't know if you've played with it, but the other weird thing is if you, if you say, if you give it a paragraph and you say generate, it'll perform it in a certain way, generate it again with changing nothing, it'll perform it in a completely different way. Yes. Like this. Do it again. Completely different way.

Leo Laporte (01:50:48):
Yeah. AI's good at that. Shelly, you said you played with it too.

Shelly Brisbin (01:50:51):
Yeah, I, well, I decided, I don't know why I decided I made this terrible decision, but it, it had been going around in a podcast for Slack and a bunch of us had been posting samples, and I was due to do a q and a on the radio show, which as a producer means that I wrote the script and that I was gonna talk to the host. I wrote him some questions that he asked me, and then I answered them. And so what I did was I took that script and I trained it on both of our voices, <laugh>. And I was like, why am I putting myself out of a job? I don't know. But you can train it with multiple samples. And so the first thing I did was I trained it on myself doing something like I would do for the radio show, which is essentially reading about a serious topic.

(01:51:27):
We were talking about disability, legislation and legislature. And then I decided to give it a couple of samples, one from my podcast about tech and one from my podcast about movies. So sort of a different tone, sort of a different focus. And I noticed as, as I added samples, and this is my, was my interpretation, and maybe what Brian says is right, that you could read the same script and have it come back with different res different sort of tone or whatever. But my feeling was that once I had given it a couple of more of my voice samples, it changed and it, it got better. It still wasn't me. I could absolutely tell the breathing was right, the sort of emphasis, especially when I, cause I made a couple punctuation mistakes at first, and I changed them up and it was a little bit better, but it was sort of flatter and without personality. And I chose the clips from the movie podcast specifically where I'm sort of, you know, excited about something or I'm trying to engage a guest and I've got energy. So I'm trying to figure out whether it's gonna transfer that to the clip and it didn't. And it really did not do particularly well on my on my radio show host, where I only put one clip of him in. But yeah, I was able to fake an entire radio segment. I didn't put it on the air, but I

Leo Laporte (01:52:35):
Had it. Can I find it? Is it on the net?

Shelly Brisbin (01:52:38):
No, no, no, it's not. I can send it to you, but it's, I

Leo Laporte (01:52:40):
<Laugh>, I wanna hear it.

Shelly Brisbin (01:52:41):
It was, it was entertaining.

Leo Laporte (01:52:43):
Yeah. I don't think we have to worry, but in fact, Brian, you put this story in my inbox voice artists are worried about chat, G P T and other AI generous, I guess not chat, G T P G P T so much as 11 labs,

Brian McCullough (01:53:04):
Whatever generative stuff,

Leo Laporte (01:53:05):
Generative is scaring them. And in fact, more than that, they're now being asked to sign away their rights, which is interesting. This is the story from Vice Dis is disrespectful to the craft actors say they're being asked to sign away their voice to ai. You know, when we do voice work, there's always a contract. Usually it says something like, we own this recording for use in every medium conceived of to this point. And anything that has ever conceived of in the future, <laugh> we use the same release, I believe, but <laugh> in any event it's now becoming a little bit cons concern concerning you. We had this story a couple of weeks ago that folks at four Chan had used 11 labs to create celebrity voices doing appalling things.

Brian McCullough (01:53:58):
Well, the, the converse would be I, I said this recently <laugh> Morgan Friedman should sell the rights and perpetuity to his voice. He'd probably be able to get 50 million right now, right? So like, and, and, and he maybe he wouldn't have to do a thing. Oh, well, maybe he'd have to go to a studio and record for a few days. No, they have so much of it. And this stuff is getting so fast. Good, so fast. Sign away your rights and perpetuity and then you'll never lose the, the March of the Penguin's voiceover.

Leo Laporte (01:54:32):
Well, wait a minute. Now, here is an ai, James Earl Jones already did that.

Brian McCullough (01:54:35):
Oh,

Leo Laporte (01:54:36):
That's right. James Earl Jones did sell us. Yeah,

Shelly Brisbin (01:54:38):
James Earl. Well, and that's, that's great for somebody whose career is close to an end or who they be looking for a legacy for their family. But if I'm a worst voice actor and I'm 30 years old, and maybe I've even just had a hit, maybe I've had a big role in a video game or a movie or something like that, and you're asking me to sign away my rights to a specific piece of work that I'm about to create for you, that's a really different situation. Yeah.

Leo Laporte (01:55:01):
Well all of us are, I guess, affected now by ai no matter what year skill gotta,

Shelly Brisbin (01:55:10):
I'm not gonna have chat g p t write the script and then I can have 11 labs reading.

Leo Laporte (01:55:15):
Yeah,

Brian McCullough (01:55:15):
Well, and then it's gonna come for images too. You know, all the, the, the, the new Indiana Jones thing with Harrison Ford being the younger thing, like you won't need Harrison Ford. You could just do Indiana Jones and perpetuity if he signed over his rights to his visage and now his voice and

Leo Laporte (01:55:31):
Yeah. Yep. Here is another story from motherboard about chat, G P T. And for some reason, <laugh> you can put weird words in there and it responds in an even weirder way. We think we know what's going on, though. There is a subreddit called Count where people are just counting, they've been doing it for five years, they're trying to get to, what, 5,000,001 at a time. And they have a list of the top cha chap Reddit handles in Count. And those are the words. So something about the way chat, G p t absorbed the count subreddit <laugh> has, has given these people some sort of weird celebrity in GPTs token set is what Vice calls it, including solid gold magic carp, streamer, bot, and the Nitro fan with a leading space. Now you probably can't t type those into chat G B T anymore. I'm sure people have, you

Brian McCullough (01:56:38):
Know, but the funny thing was, is fixed, like, even, even, even if they fixed this, I, what I thought was interesting was like the, the way that you can <laugh> sort of glitch out the bot's sort of like Blade Runner style because one, one of the quotes in first five paragraphs that Leo is like, they, they put the name in and the the bot responded by saying, to hell with you, <laugh>.

Leo Laporte (01:57:00):
<Laugh>. It's like, it's hysterical. <Laugh>. It, it's, it's like it breaks the ai.

Brian McCullough (01:57:06):
Yeah. Yeah.

Leo Laporte (01:57:08):
And for some reason that's just wonderful <laugh>. We just love that, don't we?

Shelly Brisbin (01:57:13):
<Laugh>? Oh

Leo Laporte (01:57:13):
Yeah. Hal 9,000 K notwithstanding

Brian McCullough (01:57:16):
Gives us a sense of superiority that we're still, we're still be

Leo Laporte (01:57:19):
Relevant, we're still on top at least. Well,

Shelly Brisbin (01:57:22):
You could just still do things that are whimsical and, and weird and funny and

Leo Laporte (01:57:26):
To hell with you. Yeah, I

Shelly Brisbin (01:57:27):
Live for that sort of thing.

Leo Laporte (01:57:28):
<Laugh>. Wow. Hey everybody, it's Leo Laport, the founder and host of many of the TWIT podcasts. I don't normally talk to you about advertising, but I want to take a moment to do that right now. Our mission statement at twit, we're dedicated to building a highly engaged community of tech enthusiasts. That's our audience. And you, I guess, since you're listening, by offering them the knowledge they need to understand and use technology in today's world. To do that, we also create partnerships with trusted brands and make important introductions between them and our audience. It's how we finance our podcasts, but it's also, and our audience tells us this all the time. A part of the service we offer, it's a valued bit of information for our audience members. They wanna know about great brands like yours. So can we help you by introducing you to our highly qualified audience?

(01:58:28):
And boy, you get a lot with advertising on the TWIT podcasts. Partnering with TWIT means you're gonna get, if I may say so humbly the gold standard in podcast advertising. And we throw in a lot of valuable services. You get a full service continuity team supporting everything from copywriting to graphic design. I don't think anybody else does this or does this as well as we do. You get ads that are embedded in our content that are unique every time I read them. Our hosts read them. We always over-deliver on impressions and frankly, we're here to talk about your product. So we really give our listeners a great introduction to what you offer. We've got onboarding services, ad tech with pod websites that's free for direct clients. We give you a lot of reporting so you know who saw your advertisement. You'll even know how many responded by going to your website.

(01:59:21):
We'll also give you courtesy commercials that you can share across social media and landing pages. We think these are really valuable people like me and our other hosts talking about your product sincerely and informationally, those are incredibly valuable. You also get other free goodies mentions in our weekly newsletter that's sent out to thousands of fans. We give bonus ads to people who buy a significant amount of advertising. You'll get social media promotion too, but let me tell you, we are looking for an advertising partner that's gonna be with us long term. Visit TWI tv slash advertis, check out our partner testimonials. Tim Broom, founder of it Pro tv. They started it pro TV in 2013, immediately started advertising with us and grew that company to a, a really amazing success. Hundreds of thousands of ongoing customers. They've been in our network for more than 10 years.

(02:00:16):
And they say, and I'll quote Tim, we would not be where we are today without the twit network. That's just one example. Mark McCreary, who's the c e o of Authentic he was actually one of the first people to buy ads on our network. He's been with us for 16 years. He said, and I'm quoting, the feedback from many advertisers over those 16 years across a range of product categories is that if ads and podcasts are gonna work for a brand, they're gonna work on Twitch shows. I'm proud to say that the ads we do over-deliver. They work really well because they're honest. They have integrity. Our audience trusts us and we say, this is a great product. They believe it, they listen. Our listeners are highly intelligent. They're heavily engaged, they're tech savvy. They're dedicated to our network. And that's partly because we only work with high integrity partners that we have thoroughly and personally vetted.

(02:01:12):
I approve every single advertiser on the network. If you're ready to elevate your brand and you've got a great product, I want you to reach out to us, advertise at twit tv. So I want you to break out of the advertising norm, grow your brand with host Red authentic Ads on twit tv. Visit TWI tv slash advertise for more details, or email us advertise@twit.tv if you're ready to launch your campaign. Now AI for video is coming. The folks who did Stable Diffusion Runway have a new model. They call it Gen one, that can create videos based on an input a sample, and then they can, they can oh, let's see if I can get it to play back. It looks like it won't play back in my on Linux here, but the, the, the top animated gift gives you an example, but you could give them a video of a real person doing something and then change the background entirely.

(02:02:13):
So get rid. I mean, this is deep fakes paradise, isn't it? Transforming existing videos into into new ones. Let's see, maybe here I can get it to play from the, from the runway site. So this is the, this is the actual video on the left. It's a subway, real subway in New York City, but I can make it a cartoon using the ai. It's a pretty good cartoon, right? You can have a human being acting kind of silly in a park, but then apply it to a a magma monster and suddenly you've got a pretty good, I mean, this is a lot easier than doing rotoscoping. Here's another one. They use blocks to, or actually these are, it looks like these are runway manuals or notebooks set up to kind of simulate skyscrapers in a road and turn it into skyscrapers in a road. This is really impressing. Give a dog some dots. Research dot runway ml.com if you wanna see. This is gen one that's going to allow you to do some very interesting things.

Brian McCullough (02:03:28):
Again, varietals though, like what if you prefer the style of one bot versus the style of another bot? Or you were in, you were an architect versus an artist,

Leo Laporte (02:03:39):
You know? Well, I wonder, you know, for instance, this image that it's creating is this is some artists draw that originally and this is just kind of stealing it in effect, or is it completely generated from scratch?

Brian McCullough (02:03:51):
No, it, it's been, it's been trained on various

Leo Laporte (02:03:53):
Artists. Real, real images. Yeah, yeah, yeah, yeah. Would yeah.

Shelly Brisbin (02:03:56):
And I assume that would happen with video where you could train it on the existing TV and movie universe or any video, any, all of you YouTube for goodness sake. Right. And

Brian McCullough (02:04:07):
It has already been done. Yeah.

Leo Laporte (02:04:09):
Right. Very interesting. I think we can probably call it a day. You guys have been working hard and I think it's time to say goodbye. But thank you so much for being here. We had a wonderful time. Shelley Brisbane, you are fantastic. I'll look for you on six colors and@texasstandard.org and on Zeppelin flights. I've already followed you on Mastodon

Shelly Brisbin (02:04:36):
And I will follow you

Leo Laporte (02:04:37):
Back. <Laugh> at Shelly at, I've

Shelly Brisbin (02:04:39):
Already followed some ofWe, but <laugh>.

Leo Laporte (02:04:41):
Yeah, that's nice. Really nice. Anything else you wanna mention that you're working on?

Shelly Brisbin (02:04:47):
Well I have a, a book that I produce for each edition of iOS, iOS Access for All, which is a book about everything to do with accessibility for the iOS platform. And you can find that@iosaccessbook.com.

Leo Laporte (02:05:00):
I forgot all about that. You were on mention that last time and good, I'm glad we could give it a plug.

Shelly Brisbin (02:05:04):
Gotta plug the book. Gotta plug the book. You got a new one? Plug Micah about it for iOS today. Yes. Cause 

Leo Laporte (02:05:09):
You know, the new one for iOS 16 is out and and yeah, if you have accessibility needs the iPhone does a very good job. Ios does a very good job for accessibility, but it kind of helps to know what it's capable of. Shelly, this is a much needed, very valuable book for everybody, so thank you for doing that. Ios access book.com. Brian McCullough, when are you going to Ireland? Oh, you're muted.

Brian McCullough (02:05:43):
All right. Sorry. Second, second week of April.

Leo Laporte (02:05:45):
Are you excited?

Brian McCullough (02:05:47):
Definitely seeing a, how, that's the first time that I've personally gone and you know, like I said maybe that week, the five episodes of the tech meme ride home be

Leo Laporte (02:05:59):
AI generated.

Brian McCullough (02:06:00):
Maybe I won't have to take a microphone in my, in my carryon and, and things like that. I will literally just do it. I, I, Leo I really might

Leo Laporte (02:06:08):
<Laugh> really. I think that's an interesting idea. You could still write it and just have the AI generated. Would you use this? And I prepared, what would

Brian McCullough (02:06:15):
You use? No, I'm gonna use the 11 labs if, because I keep feeding it stuff from my show. It's getting

Leo Laporte (02:06:20):
Better and better.

Brian McCullough (02:06:20):
Yeah. I've been preparing the audience for this, like, we've done segments and I've said, okay, here, here we go. Because every, you know, when all this week it was nothing but AI news, and so I couldn't help myself. I'm like, okay, yeah, we, we gotta do this next segment with all doing AI voice we're all doing. So if, if I prepare the audience and I say, I'm not gonna do this forever, but for this one week, will you accept it? And the interesting thing was, I did one experiment with different voice. I did an Irish accent, you know, <laugh> not my voice. And people hated it. But when I did it with my voice this week, people were like, oh, that's not bad. So again, now, but again, what are the implications? The implications are I'm not paying a substitute host for a week, right. To take over my show. Right? So

Leo Laporte (02:07:11):
Ride home.info, what will Brian McCullough do? Stay tuned and find out Ride home.info. Thank you Brian for being here. Really appreciate it. And thank you Sam Abul. Sam Ed, my Car Guy Wheel Bearings podcast is@wheelbearings.media. Between you, Nicole and Robbie, I think you're probably covered for vacations.

Sam Abuelsamid (02:07:35):
Yeah. Yeah. We try

Leo Laporte (02:07:37):
Also I, we didn't even say it today. I don't know, do you want me to principal researcher at Guide House Insights?

Sam Abuelsamid (02:07:44):
Absolutely, please.

Leo Laporte (02:07:45):
Yeah. Yeah. That's his day job, folks.

Sam Abuelsamid (02:07:47):
That's, they're, they're, they're the ones that pay the bills.

Leo Laporte (02:07:49):
Yeah. So we, we

Sam Abuelsamid (02:07:50):
All, all, all the other stuff is is just a hobby

Leo Laporte (02:07:53):
<Laugh> it's good to have hobbies, you know? Yeah, yeah. Absolutely. You never know what's gonna happen. Will bearings.media for more information or to hear the podcast? You've launched a Patreon, that's great. Yeah, so people can become fans. And

Sam Abuelsamid (02:08:08):
LA la last week we had Kelly Funk Hauser from Consumer Reports on oh, to talk about they, they recently did a bunch of testing on driver assist systems and had some interesting results. And so we invited Kelly to come on and, and talk us through what their methodology was, what they were actually looking for, to, to get a better understanding of how they came to the conclusions that they did. And so that was a really interesting conversation. And tomorrow morning we'll be Nicole and Robbie and I will be recording another one. And we'll have some interesting content there as well.

Leo Laporte (02:08:44):
Join Sam, Nicole robbie@wheelbearings.media. Thank you Sam. And of course, join us. Normally we are not on a Saturday. We just thought we'd give everybody a break, cuz tomorrow is a big game, super Bowl Sunday. So we invite you to tune in next Sunday. We'll be back on our usual time, 2:00 PM Pacific, 5:00 PM Eastern. That's 2200 UTC right after. Ask the tech guys you can watch live@live.twit.tv. Chat with us live@irc.twit.tv or if you're a member of Club Twit, join us in the Fabulous Club Twit Discord. You can also get after the fact, get versions of the show audio or video at our website, twit tv. There's a YouTube, a YouTube channel dedicated to this weekend tech. And of course you can subscribe in your favorite podcast client and get it automatically the minute it's done, which we are now. Thank you for, for joining us. I'm glad you tuned in on a Saturday. We'll see you next week. Another twit is in the can. Bye-Bye. Amazing.

All Transcripts posts