Transcripts

Tech News Weekly 348 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.

0:00:00 - Mikah Sargent
Coming up on Tech News Weekly. Amanda Silberling is out, but Jennifer Pattison Tuohy actually joins me for the first half of the show to talk about all of the Google Nest news, because there's a lot of it. Google's smart home announcements include a new Nest learning thermostat, some updates to Google Gemini intelligence making their way into the smart home and the end of Chromecast, but the introduction of a pretty cool Google streaming set-top box. Then Reed Albergotti of Semafor stops by to tell us about the AI company behind many of the popular AI companies that provides human data to improve upon the AI systems we've come to know and use. Afterward I've got a quick little story of the week from Mark Gurman all about Apple's potential new Mac Mini. All of that is coming up on Tech News Weekly. Stay tuned Podcasts you love.

0:01:04 - VO
From people you trust. This. Is TWiT.

0:01:10 - Mikah Sargent
This is Tech News Weekly with me, Mikah Sargent, episode 348, recorded Thursday, august 8th 2024. Google's Nest News. Hello and welcome to Tech News Weekly, the show where, every week, we talk to and about the people making and breaking the tech news. Hello and welcome to Tech News Weekly, the show where, every week, we talk to and about the people making and breaking the tech news. I am one of your hosts, Mikah Sargent, and typically Amanda Silberling would be joining us this week Amanda Silberling of TechCrunch but Amanda is on vacation at the moment and so I have interviews planned for you. But what's funny is that my first interview that I have planned for you is a very familiar face. Joining us from the Verge is Jennifer Pattison Tuohy, to talk about all things Google. Welcome back to the show, jen.

0:01:35 - Jennifer Pattison Tuohy
Thank you, Mikah, very happy to be here, as always you, Mikah, very happy to be here, as always.

0:01:42 - Mikah Sargent
Always a pleasure to have you here and to get to talk about smart home stuff, because I no longer have that show where I strictly covered the smart home. So it's always fun to get to chat with you about this, because you dive deep into this stuff and I love it. I love how much you nerd out and really kind of get into the nitty gritty. So let's start with what just happened, because Google, which is going to have an event soon, did kind of a pre-announcement of some of its products that are maybe not, as we'll say, pixel forward and you weren't going to say exciting, right?

0:02:47 - Jennifer Pattison Tuohy
Oh no, no, no. You weren't going to say not, as we'll say pixel forward. You weren't going to say exciting right? Oh, no, no, no.

0:02:50 - Mikah Sargent
I think they're exciting, to be honest. What was funny is I just recently moved to a new place and I was getting the Nest thermostat in my home connected to, you know, reset, and then connected to my Wi-Fi, and then connected to my app-Fi, and then connected to my app, etc. Etc. And within, I think, a day of getting it set up, I opened up the app and, boom up, popped this prompt saying hey, we've got a new Nest Learning Thermostat. Would you like this one instead? Oh boy, luckily I didn't buy the one that's here. It was already here, but yeah, so what?

0:03:26 - Jennifer Pattison Tuohy
did.

0:03:27 - Mikah Sargent
Google announce.

0:03:29 - Jennifer Pattison Tuohy
Yes. So this was very much a big focus on all of the Google Home part of Google which really doesn't often get a lot of focus and hasn't had a lot of focus in a while. So this was kind of Christmas comes early for smart home reporters was. This was kind of Christmas comes early for smart home reporters. They announced three key things. The first is the Nest Learning Thermostat, which you just mentioned. This is the fourth generation of their iconic thermostat. You know that's a word that does get thrown around a lot, but I feel like it kind of applies here. I mean, this really was the beginning of the sort of DIY smart home movement about 12, 13 years ago and it was so pretty. And the new one is so pretty. If you've got to remember what thermostats used to look like, these are really pretty. Um so whole new design. This is, um, whilst it still retains a feel of the original nest, it's com, it's completely redesigned and it looks like a giant Pixel watch on your wall. But we can go on.

I will go more details, but just to flip through the next couple of things new Google TV for streamer 4k. So this is a new. Now it's a set-top box for your streaming like a Chromecast, but it's also a smart home hub, so it's basically Google's version of Apple TV so you can stream all your shows. It does act like a Chromecast and it also has smart home hub capabilities. So it's a Google home hub. It is a Matter controller, it's a Thread border router so it can do all the smart home. So that's exciting. That's $100, and then the and it's also not a dongle. So no more dongles, no chromecast dongles. It's a little box that sits under your TV. That's kind of divided people, I think it's. It's an interesting design. But new chromecast remote sorry, new remote that goes with the streamer and it looks just like the Chromecast with Google TV remote but is a little bigger. And also the new smart home feature. On Google TV we now have a pop-out smart home control panel, just like you have on Apple TV, so you can control your TV, control your smart home through your TV and view live video streams. So if someone presses your Nest doorbell it can pop up on your TV like you can do with your Apple TV. So it's very similar. It has a lot. It takes a lot of cues from the Apple TV, which is a good thing, because the Apple TV is great. So this is a good, great upgrade, I think, for Google Home users.

And then the final thing which Google kind of underplayed a little bit but I'm most excited about, is they're bringing Gemini intelligence to Google home, and I say bringing as in. This is not here. The other two products are launching in the next month. This is later this year in public, but it's a lot of generative AI powered features for Google Home You're going to get Nest cameras will gain the ability to not just send you an alert that says person detected or animal detected, but will actually be able to process, using Gemini multimodal AI, process what the camera sees and present you with a caption like a description of what the camera is seeing. And that's sort of the first step and you can kind of see where that might, where that could go. So you know you could set up automations. Where a camera sees something specific, like a car pulling into the driveway or a person opening the garage door, you could set up home automations. So it's a really interesting sort of upgrade for your Nest cameras.

They're also adding the ability to set up smart home automations with natural language, so I'd like a way for my kids to be reminded to put their bikes in the garage was the example Google used. And then Google Home can set up an automation that turns on the light at the garage, um broadcasts a message over nest speakers saying kids, put your bikes away. Um, every time kids show up in the driveway between a certain time, and Google home will just set this all up for you and all you need to do is say this is what I'd like you to do. Obviously, you also need compatible devices. It doesn't work magic like you to do. Obviously, you also need compatible devices. It doesn't work magic.

And then the final thing is Google Assistant itself is getting an AI overhaul, similar to what we've heard is going to be happening to Amazon's Alexa. It's going to be and these are their descriptions smarter, more conversational, more capable of understanding natural language and is also going to have some new voices on the smart home speakers. So all of these the new Google Assistant features are coming to this Nest smart home displays and speakers. As opposed to, this is different from what you might use on your phone or other Pixel devices, so specifically to the current Nest hubs and speakers. So we're not getting any new Nest speakers or displays, but the current ones you have are going to get more capabilities, but this is all coming later this year and in public preview, which is Google's version of like a beta program, and it does require Nest Aware subscription, so you have to pay, so that's it.

0:08:51 - Mikah Sargent
Yeah, that's everything.

And we're going to kind of get into more detail with everything, starting with the Nest thermostat. And so, when it comes to the Nest thermostat, I think it's important to kind of look at the history of what it has looked like before, what it looks like now, how it kind of there's almost an upside down bell curve right of features, and maybe you could, if you can, explain what Google was thinking in sort of taking some of what it did out. And then I also love that in your piece you talked about a competitor, which it's one of my favorite products that I have owned, which is the Ecobee thermostat. So let's start with the history. How has the Nest Learning thermostat looked in the past? And then kind of, how has it changed in terms of features over time and what was the reaction from the folks who buy these things?

0:09:53 - Jennifer Pattison Tuohy
Yeah, so the original Nest looked a lot like in that picture, the top left one which is. So there have been three versions of that, so the current one that Mikah has installed in his home, I'm assuming was the third generation one and that's what I'm holding in my hand those are my hands, by the way. Oh, that's what I'm holding in my hand on the top left and that's the most recent version and that has the. So that has a tangible dial, a physical dial, so you can see that silver band is you can actually twist it, which is great. That was kind of the feature that everyone loved about the Nest, because it was a smart device, but it also had really easy physical inputs.

A lot of smart thermostats which just have digital displays. People find a little tricky to use and a bit of a learning curve, and a thermostat is like lights in your home. It's something that everyone needs to be able to know how to use. You shouldn't have to read a manual to turn it on or turn it up. So this was a great piece feature of the Nest. It was on the original one which launched in 2000 and oh, I've forgotten off the top of my head 13 years ago someone do the math 2011. And it was a big bang when it launched. It was not Google, nest was Nest. This was before Nest was bought by Google. It was launched by Tony Fadell, who formerly was created by a number of people that used to work at Apple. It was sort by a number of people that used to work at apple. Um, it was sort of you know, the iphone of thermostats, um, so it was, and it was a. It has, was then and still is a very sort of iconic product in the heart in the smart home the original.

I have one of the. I think I have the second gen, because the second gen came out pretty fast after the first gen and was kind of the one most people would have bought, and that was a couple of years later and I still have one of those working in a house. They were built to last, which is great for a product in the smart home you don't want to spend, and they're expensive $250. These are products that need to last. And then the third gen came out about nine years ago, which was a big upgrade.

It made it a lot. It gave it a better screen, a brighter screen, more function on the screen. You could also use it as a clock. It had a feature called Farsight that as you walked towards, the device would sort of shift and show you a different view, some more detail, as opposed to further away you could just see what the temperature was. You could also set it to be a clock. It had some more sort of modern features because the original had been around for quite a while and there'd been a lot of upgrades in technology, including Thread, which is a side note that actually the original Nest Thermostat is the reason Thread was developed, which is a side note that actually the original Nest thermostat is the reason Thread was developed, which is one of the new smart home wireless protocols, and the latest Nest does not have Thread in it, but this may be a separate conversation.

And then I think the point that you were interested in here is that then, after the third gen, they came out with an entirely new Nest thermostat that had no real connection to the original and was quite disappointing in many ways and that, if you go back to that picture, is what I'm holding in my other hand, and that is just the Nest thermostat as opposed to the Nest Learning thermostat. And basically the difference with this one was it wasn't really as smart, it was just a Wi-Fi controlled thermostat. And basically the difference with this one was it wasn't really as smart, it was just a wifi controlled thermostat. It didn't have the key unique feature of the Nest, which is the ability to learn your heating and cooling habits, like how you like your home to feel temperature wise, and adjust and create a schedule for you and constantly adapt that schedule based on how your lifestyle changes, and that was a really neat feature. It could sometimes go rogue and I had a few times where you know suddenly it would be heating in the middle of a sunny afternoon, but in general it sort of manages itself, so you don't have to worry about programming your thermostat.

But the new, the Nest Learning Thermostat that came out in 2020, did not have that feature. It was just a standard programmable thermostat that you could still control remotely and set with smart home automations, and worked with Nest's big feature, which is home and away, so the thermostat could tell if you were home or if you were away and, if you're away, adjust your heating or cooling to save you money. And this was the big selling point of the Nest is that it would save you on your energy bills. And now we have the latest one, which is a complete redesign but retains the same functionality as the original. So it's physically redesigned but the software has had improvements and some additions, but overall it's the Nest Learning Thermostat that we all know and probably love. Some people might not.

0:14:58 - Mikah Sargent
Yeah, one would think. I know that the, you know, the eco features of the learning thermostat was really kind of important to a lot of people, and so the loss of that was kind of interesting. Well, you know some of the more intelligent features, right? Yeah, outside of that, what do you feel sets the fourth gen apart? Yeah, yeah, like what? What makes this something that people would cause you talk about in your piece? You know this is something that you keep in your home for a long time and, in theory, you're not upgrading every year. So someone is looking at buying one of these and they're on, you know, maybe an earlier gen, sure, but yeah, tell us about, like, what sets why you might want to upgrade.

0:15:50 - Jennifer Pattison Tuohy
Yeah, yeah, um, so it it looks. It's got a great look, um, and if if you like the nest, it's probably because you like the look of it and you will probably like the look of the new one. I mean, there are a lot of smart thermostats out there, but most of them really don't look great. Even the eco Ecobee, which has got better in its looks, still looks sort of like a smartphone app on your wall. Yeah, it does. So you know, it really does have, and I got to see it in person, although the lighting and everything in the room where I got to see it wasn't great, but it really does. It was striking. It's much bigger, which I think is probably a big upgrade for a lot of people. If you like to be able to see it clearly from further away and if you have a big open plan living room, it's gonna be more obvious. It also has matter, which the original Nest Learning Thermostat. None of the three generations of that support matter. So that means it will work natively with Apple Home, if that's something I know that might be of interest to you. Mikah, the Nest Thermostat, the cheaper one, does work with Matter as well. So they've now got two thermostats that will work with Matter. We have not heard and I have asked, but I've not had an update as to whether they may one day upgrade the original, the third gen, to Matter. They've never said never, so we're waiting to hear. So, yeah, so those are a few features. There are also some upgraded software features that they say are only going to be on this thermostat at launch. Who knows, though? They may one day be backported, but including, like, if you have a ventilation system, the new Nest Learning Thermostat can work with your ventilation system and with outdoor air quality and actually turn on your ventilation system when the air quality is good and turn it off when it's bad. You know, do sort of more smart integrations with ventilation systems. It also works with more systems. So if your system wasn't compatible with the last Nest thermostat, you might find that this one is, although, as far as I'm aware, it doesn't work with variable speed units yet, which is kind of a shame, but there are very few thought smart thermostats that do, so it's not that surprising. Um, and yeah, it's it's. I mean, I really like the UI on it now. It's a lot easier to use that.

The third gen is a small little screen and you do most of your programming of the thermostat. Not that you have to do a lot of programming, but when you are interacting with it, most of it is on the device, not in the app. Although with the new one there is much deeper integration with the Google Home app than there was with the old one and this is another change is this just works in Google Home. You don't have to use the Nest app. Oh wow, which is the all the previous ones? Nest Learning Thermostats worked in the Nest app.

0:18:44 - Mikah Sargent
Yeah, that's confusing too, because adding a secondary userats worked in the Nest app. Yeah, that's been confusing too, because adding a secondary user required going into the Home app and then adding them there and then going back into the Nest app. I mean, it's this and it's going, I'm going. So who wants control here? We can't figure it out. And yeah, it's not as robust. The control in the Home app doesn't feel as robust. No, it's very limited, very limited.

0:19:12 - Jennifer Pattison Tuohy
Yeah, and this now has much more integrated control in the app. And then there's also this is really neat, I'm excited about this there's a Soli radar in the new Nest Learning Thermostat, whereas the original Nest Thermostat used a PIR sensor. So for those not familiar, the Soli radar is something that's in a few of Google's products, including the Nest Hub second gen and I think it's in some of the phones not my realm and it is basically an advanced motion sensor so it can detect more precision detection, and the main reason it's there is for the far sight, which is where, as you approach, the screen changes and as you get further away it changes back and there are more details that you can add to this far sight. Now it's called dynamic far sight and it has this really cool feature that shows lightning if it's thunder and lightning outside shows snow if it's snowing and rain.

It looks nice, it's fun. I like that as an addition. You don't have to have that if you don't want it. It's fully customizable, which is nice, so you can choose exactly what you want it to show on the screen, kind of like you can with your watch. I guess it really does look like a Pixel watch I mean really does, and they actually said that, like we were inspired by the pixel watch, so which just seems a little funny to me.

0:20:34 - Mikah Sargent
But a pixel watch on your wall?

0:20:37 - Jennifer Pattison Tuohy
yes, um it has much better sustainability features too. If you know which is In fact, the battery is 100% recycled cobalt for the first time, which is kind of neat. So yeah, that you know it's very much designed as a sustainable product, both as the product and for what it does in your home. And you can work with Nest Renew, which is the demand response, if your electricity provider has that capability so that you can try and save more energy and more money on your electricity bills. So capability so that you can try and save more energy and more money on your electricity bills. So yeah, in terms of upgrading, if you have a third gen, I mean there's no real need unless you maybe have a ventilation system and you think that would help, or your system. You know you have more features in your system than this one's. The third gen is capable of running.

But overall, if you've been wanting to get a Nest thermostat and haven't, I think this new one has a lot of fun upgrades and it is not more expensive because, even though it costs $280 as opposed to $250 as the original, it now comes with a temperature sensor, which is new. They have the second generation of the Nest temperature sensor and this is what Ecobee offers. We mentioned Ecobee. They're little sensors that you can put in other areas of your house other than the thermostat to help kind of balance hot and cold spots. And you now get one with the thermostat, whereas you didn't before, and those are $40. With the thermostat, whereas you didn't before, and those are $40 each. So actually, if you do the math, this one's cheaper.

That's true, yeah, so, and those little temperature sensors are useful. I do prefer the Ecobees. So far, in terms of the first gen, I've tested the first gen Nest thermostat temperature sensors and they just measure temperature, they don't measure occupancy and they have quite strict time slots that you need to use, and I didn't find it as flexible as the Ecobee ones, but they are still. I mean, they're useful, especially if you're working in your office all day and your thermostat's in your hallway. Then you can have you put the little sensor in your office and it will help make sure that your thermostat is targeting the temperature in your office and not the temperature in your hallway.

0:22:54 - Mikah Sargent
Yeah, that's the thing that I really liked about the Ecobee, so I'm glad to see that come to. Well, not just come to, but be updated.

0:23:02 - Jennifer Pattison Tuohy
Yes, cause they had it with the third gen.

0:23:04 - Mikah Sargent
Yeah, okay, we do need to take a quick break before we come back with even more that Google announced, of course, to learn more about the Nest Learning Thermostat you can check out Jennifer's article all about it over on the Verge.

But let me tell you about 1Password, who are bringing you this episode of Tech News Weekly Question for you Do your end users always work on company-owned devices and IT-approved apps? Yeah, I didn't think so. So how do you actually keep your company's data safe when it's sitting on all those unmanaged apps and devices? Well, 1Password has an answer to this question. It's called extended access management. 1Password extended access management is the first security solution that brings all these unmanaged devices, apps and identities under your control. It ensures that every user credential is strong and protected, every device is known and healthy and every app is visible. It solves the problems traditional IAM and MDM can't touch.

Imagine your company's security. Like the quad of a college campus, there are nice brick paths between the buildings. Those are the company-owned devices, it-approved apps and managed employee identities. And then there are the paths that people actually use, the shortcuts through the grass that are the actual straightest line from point A to point B. Those are the unmanaged devices, these shadow IT apps and those non-employee identities like contractors. Most security tools only work on those happy brick paths, but let's be real. A lot of security problems take place on the shortcuts.

1Password Extended Access Management is security for the way we actually work today, available now to companies with Okta and coming later this year to Google Workspace and Microsoft Entra. So check it out at 1password.com/twit. That's 1-P-A-S-S-W-O-R-D.com/twit, and we thank 1Password for sponsoring this week's episode of Tech News Weekly.

All right, we are back from the break, and Jennifer Pattison Tuohy of the Verge has agreed to stick around for a little while longer, because Google had more that it announced ahead of its Pixel event. This time we should talk about Gemini Intelligence and the new Google Streaming box. What all did Google have to say about how Gemini is coming to the smart home?

0:25:53 - Jennifer Pattison Tuohy
Yeah. So the interesting thing I've been waiting for the announcement that actually Google Assistant was dead, and long live Gemini in our Nest Hubs. But actually what they've decided to do is keep the Google Assistant around, albeit with new voices, but giving it a Gemini intelligence sort of makeover or powered by Gemini intelligence. So they're starting with three features or three areas that they're improving. So basically they're taking different models from Gemini and applying them to different parts of Google Home. The three main parts are to the cameras the Nest thermostat cam sorry, the Nest cams which are you have a Nest outdoor cam, indoor cam, floodlight cam, doorbells and helping them be smarter, basically around notifications, because if you have a smart home camera one or two or even more you may frequently get annoyed by how many alerts you get. And you get an alert that there's a person in your yard and you're like, ah, freak out.

Now I have to see the video. It doesn't load. Oh my God, there's someone in my house. What am I going to do? What this is going to do is send you more detail, rather than just there's a person in your house. It will be the actual description. I'll read it, the one they showed. So I didn't see any of this in action. It was all just a video, but they showed a clip of a person unloading groceries from the car, and the caption that you get is a young person in casual clothing standing next to a parked black SUV. They are carrying grocery bags, the car is partially in the garage and the area appears peaceful.

0:27:30 - Mikah Sargent
The area appears peaceful. That's the best part, right at the end.

0:27:35 - Jennifer Pattison Tuohy
I know I'm like some interpretive details here from Gemini, but you know that's useful right. As opposed to, the area appears chaotic. If it said chaotic. I would be like, oh, okay.

0:27:48 - Mikah Sargent
I need to go see what's going on.

0:27:51 - Jennifer Pattison Tuohy
Everything's peaceful.

0:27:52 - Mikah Sargent
It's okay.

0:27:54 - Jennifer Pattison Tuohy
Yes, yeah, so that'll be, I think, an interesting upgrade to have that more. And this comes as text, so you don't have to go back into your camera app and watch the video and scroll through and see you know who arrived when or what happened. You can actually just get a summary. And also you can search through your activity tab so say you wanted to know when your cat came home. You could search, you know when was the last time a cat was spotted on which camera, rather than having to scroll through every clip of an animal. I get lots of those because I live in a rural area and then I know which time, rather than having to scroll through every clip of an animal. I get lots of those cause I live in a rural area and then I know which time, when my cat was last at the front door. So I think those could be interesting upgrades, depending on how well they work.

And that's for all Nest cameras to start with. And you will have to have the Nest Aware subscription because obviously it requires recorded video. So you have to pay for the recorded video, although you do get some free video on Google cameras, google Nest cameras but you have to have Nest Aware. For this feature you also have to be in the public preview, at least for the first. That's when it first rolls out later this year and that I think it'll also be coming to select users.

It's not going to sort of be a general thing because, as Google told me, they're trying to roll this out in a measured approach because, as we know, when people throw generative AI out there into the world and just let it run, things can go wrong. So they're trying to. You know, and they recognize that the home is not the place you really want to mess up. So, like most of the companies, I think they're proceeding with caution. They're also bringing the ability to use natural language to create automations.

It's called Help Me Create and you type into the Google Home app or you can speak into the Google Home app and say what you want the automation to be, as we mentioned earlier, and this uses all of the current starters and conditions and actions of Google Home, so it can pretty much create any automation that you could create using the app. It just does it for you. So it should make it a lot easier. I mean, if you've ever spent much time in the Google Home app or any home automation app. Setting up an automation can be quite laborious and tiresome.

Yes, and you know, and hopefully Google intelligence may actually be a bit smarter about it than we will, because it will know all the capabilities of every device you have. That's the kind of beauty here. Whereas, you know, even I, who does this for living, maybe doesn't read every manual from beginning to end and know about every single feature that every single device has. So we might see this help make even people who are good with the smart home come up with even more interesting automations. But the main purpose of this is to make it easier for anyone to set up automations, so I think that should be interesting, and this is also what Amazon has said it's bringing to Alexa. So, again, none of this is here now, and this will, they say, later this year, and you will have to pay for Nest Aware to have this feature. Google Assistant is getting better Thanks to Gemini, and this is this. So they've told me that they are working on making Google Assistant better for everyone, so you won't have to pay for all of these features.

0:31:12 - Mikah Sargent
That's good.

0:31:14 - Jennifer Pattison Tuohy
Because they recognize that maybe some of the core experiences of Google Assistant aren't quite up to scratch. I mean, things have been getting worse as well recently.

it feels not just for them, but for Alexa and it's like the technology is, you know, and what our expectations from technology have outpaced, how fast they've developed. And that's what Google sort of said to me. You know, yeah, we see this. We've plateaued. It's been a real challenge, but generative AI is really going to help us push challenge. But generative AI is really going to help us push through that ceiling that we had hit three or four years ago, where things basically stopped getting it much better.

And their ambition and hope is that things will be significantly better with this influx of Google Gemini intelligence. But it's also going to have more natural language. You should be able to have like a conversation rather than have to give very specific nomenclature. And they have a new voice who still has that sort of female tone but is much more natural sounding. There's a little video in my article if you want to go listen to it, and they say there are more voices coming to the Nest, smart displays and speakers too. So all of this is for the Google Assistant on your smart displays and speakers, not this. You know this is different. The Google Assistant in your smart home is a slightly different beast from the one you may use elsewhere.

0:32:41 - Mikah Sargent
Understood. Well, I am afraid that we have actually run out of time. I would like to encourage everyone to head over to thevergecom to check out the piece on the Google TV streamer, as well as Jennifer's pieces on Gemini intelligence and the hands-on, quite literally, with the Nest Learning Thermostat, For even more information. Jen, I'll be seeing you again soon. Of course, folks can head over, as I said, to thevergecom to check out your work. Is there anywhere else they should go to keep up with what you're doing.

0:33:17 - Jennifer Pattison Tuohy
Well, hop over to YouTube because we have a video of the hands-on with all of these features. So The Verge channel on YouTube. Go check that out. And also some social videos over on Instagram and I'm @smarthomemama over there.

0:33:32 - Mikah Sargent
Thank you so?

0:33:33 - Jennifer Pattison Tuohy
much. All right, thanks Bye.

0:33:36 - Mikah Sargent
Bye-bye, all righty. Up next, we are going to talk about the AI company behind the AI companies with Reid Alberghati of Semafor. That is up next. So, as I mentioned, it is time to learn a little bit about the company behind the AI companies, or at least it seems that way when it comes to gathering human data. Reid Alberghati, thank you for coming back on the show.

0:34:06 - Reed Albergotti
Always good to be here. Thanks for having me.

0:34:13 - Mikah Sargent
Pleasure to have you join us. So you wrote about Turing and I was hoping you can give us an overview of Turing's role in helping OpenAI specifically to develop its breakthrough language model, gpt-4. What kind of data did OpenAI need from Turing and how did that data contribute to GPT-4's performance?

0:34:29 - Reed Albergotti
Yeah, well, turing actually started out as a staffing firm that helps software companies get really high quality developers to work as contractors, and they had expertise in AI because they were using AI to vet their sort of network of hundreds of thousands of computer programmers around the world.

And then one day, openai called them this is early 2022, into this meeting and said you know, we want your developers, but not to actually make products for us, but to create data that we can throw into you know, training these models, and it was the biggest account that Turing had ever had, by like orders of magnitude. So basically, openai wanted just a massive amount of data. And the reason is OpenAI had figured out and now kind of everybody knows this that when you throw a computer code into the training sets these models, they become better at reasoning, not just in creating sort of like the co-pilot helping you create software, but actually just in regular language as well. They become sort of smarter and so, opening eyes, we need more computer code, but we need it in a way that actually is higher quality and helps these models learn even more. And so that's what they did, and you know, now Turing has really this has become their core business. They're still doing staffing, but this is probably a much bigger potential business and they're working for all of these AI companies now.

0:36:10 - Mikah Sargent
Yeah, that's kind of the big thing that stood out to me is that it's not just OpenAI, and you mentioned that Turing's work with OpenAI and other AI companies has led the company to transition its business model entirely. So, yeah, I mean with starting out as software engineers. I guess my question with the transition is maybe if you can elaborate on how simple it is to go from staffing software engineers to suddenly moving into this AI consulting services business. Did the company itself expect that, or how does this happen? And that's just quite the flip right.

0:36:56 - Reed Albergotti
Yeah, I mean, after this meeting with OpeningEye, they had to go back and create a bunch of new workflows and systems at their company, because it was a whole new concept, right. And they've now done this. It's not just computer programmers. They've taken that expertise in doing this with software developers and now they're getting like PhDs in physics or sales professionals just any vertical, or sales professionals, just any vertical. They can go out and find these people and then have them create the data in a way that's helpful, that is able to be sort of ingested in this post-training phase of AI models. And it's not just AI companies, it's also Fortune 500 companies that want to fine-tune their own AI models now for very also right, you know fortune 500 companies that want to fine tune their own AI models now for very specific purposes, right? So they're they're going out there and finding these people and I asked them what does it look like? What does the data actually look like that you're sending them? And it was kind of surprising.

It's not like what you would think. It's basically people with expertise to sit down and ask themselves questions, and so it's like they're pretending to have a conversation with you know, whatever chat, gpt or something and they'll say you know, what do you think about? You know particle accelerators or something like that? And then they answer their own question and they might do like 10 back and forth. So it's almost like a like they're having this, this conversation, or something like that, and then they answer their own question and they might do like 10 back and forth. So it's almost like they're having this conversation and that for some reason that is able to. It's not just getting these models to regurgitate that information but to actually learn from it and then be able to kind of reason beyond that and answer questions that they may not have even been trained on.

0:38:48 - Mikah Sargent
So is that the multi-turn data that you were talking about? That back and forth conversation? Is that what multi-turn data is?

0:38:55 - Reed Albergotti
That's the yeah, that's the jargon inside the AI industry. That's what. That's what they call that data. So if you so, if you hear that thrown, you know bandied about, that's, that's what you know they're talking about.

0:39:07 - Mikah Sargent
Got it, and so that helps to make these AI agents, Because that's like to me, from what I have heard, that is the term. Now, you know you went from just large language models and generative AI this and that and this and that to everybody's focused on making agents, Agents, this agents, that, and so this multi-turned data helps to create these. What is it? Agentic AI models, right.

0:39:40 - Reed Albergotti
It's part of the process, what they're doing here. It's not like they just throw this data in and then it's part of this dataset. They put it in and then they evaluate the models and see if it learned anything. Then they test to see if it actually learned or if it's just repeating that information. They don't want it to repeat the information. They don't want the data in that thing to just be regurgitated. And that's something that's come up in some of these copyright cases, right Like the New York Times case against OpenAI, where they said it just repeated lines from our articles. That's actually not what these companies want. They want the models to be able to say, okay well, I've learned all these things from hundreds of physicists about the way classical physics works. And then somebody can sort of ask it a question that it's never seen before, and because it's learned all this information, it's able to then answer that question. That is the kind of breakthrough that they're after, and it's that reasoning ability. Part of the agentic goal here is to get these things to reason better.

0:41:00 - Mikah Sargent
Wizardling in our Discord chat brings up a good point. This requires having a lot of trust in the experts that they're hiring, because somebody who's an expert in particle physics could just come along and make up an entirely new particle and write this into. So they must be working with some pretty trustworthy experts across the board, such that you don't get this sort of false education that's taking place.

0:41:30 - Reed Albergotti
Right, I mean that is their value proposition. I mean that was their expertise when they were just a staffing firm, right, they used AI models to vet software developers and make sure that they were, you know of that high quality. So now they're trying to do that for these other industries and go out and find these people. But it certainly is a risk and it speaks to this broader question problem with data, right. So in the pre-training phase of all of these models, they ingest, you know, all of the internet, or common crawl, you know petabytes of data from the internet and a lot of that is nonsense, and so a lot of what they're doing after that pre-training phase is trying to teach it what is. You know how to tell the difference between nonsense and you know actual thought and reasoning.

So, you know, it's really, I think, the next phase of AI that we're heading into. We've learned how to, you know, make models that can kind of sound human, because they've seen so much language and you know information on the internet. Now you need them to. You know, I said in the article, they're kind of like toddlers who've, you know, been listening to adults and picking up language and cues, but they don't really know what it all means and they're able to kind of, to kind of verbalize now and now they need to go to school, you know, now they need to start. You know, studying up and being able to really think in more abstract ways yeah, that makes sense.

0:43:05 - Mikah Sargent
Now you do talk about in the piece kind of the debate around the use of copyrighted content for training AI models and I'm curious what are some of these key issues and maybe Turing's impact on that, if at all, and how could that affect the training and development of future large language models?

0:43:33 - Reed Albergotti
Yeah, I mean, I think you can't. There's no copyright issue with what Turing's doing, because they're just going out and paying people to create data. It's for this purpose. And I think my point was, you know, if we kind of look forward to how these things are going to develop, all of this data that they've ingested already, it's kind of not as valuable as we think. Like, I have lots of articles you know from my days at the Journal and the Washington Post I'm sure are in many of these data sets. I have a book that I think is in some of these data sets.

I don't think of it as a copyright issue because it's not like people are saying, well, let me get this article that Reid wrote through this model or let me read his book through an AI model.

It's more like that stuff that I've written is background noise, like back to the toddler analogy, right, it's like the adults talking. It's not really important what those adults are saying, just that they are saying something right, and so the kids around them sort of you know, pick up that and learn the language through this kind of like mysterious osmosis process. So I think that was my point. It's like we think of, we think of all this copyrighted data as like oh my God, these companies are just stealing information and they're just going to replace, you know, all these news agencies. That's that's not what. These companies are just stealing information and they're just going to replace all these news agencies. That's not what these companies want at all. What they want from the data is to help a model create original content, stuff that hasn't ever been created. That's the holy grail, that's the breakthrough that they're looking for.

0:45:20 - Mikah Sargent
Absolutely. I mean speaking of looking forward, as you mentioned. I'm curious, as someone who's paying a lot of attention to AI and I think you've done a great job on so far of paying attention to a lot of the companies that aren't necessarily the big names right now but are instrumental in current development of, and future development of, ai. What capabilities and use cases do you kind of see emerging as these models become more specialized and agentic in their abilities to reason and execute, as opposed to typing in on Google, should I eat a rock and having it spit back? Some weird Reddit response that makes no sense.

0:46:10 - Reed Albergotti
Yeah, I think your previous guest talking about you know, the ability to set up all these automations by just sort of talking to this AI assistant is probably like the first kind of baby steps right, where you, you know, instead of having to go into the settings and do a bunch of things that you're most people aren't going to take the time to do, you just ask a computer to do it.

So it becomes that kind of like new language user interface. I think over time what really is under the hood there and what's important, is that when you're asking an AI agent to do something, it's actually writing code under the hood in order to go and access APIs, and so there's Python scripts being written and for some of these more advanced prompts, right. And so I think what you know we're sort of like expecting at some point down the road, and who knows when it will happen is you know, an AI agent goes and talks to another AI agent and they're talking together in the sort of language of computer code, and they're talking together in the sort of language of computer code, but to the user it really just looks like you're asking one you know chat bot to go book you plane tickets or something like that. Or you know, plan a trip. You know all this stuff where you can kind of ask the chat bots now to like plan your trip itinerary, but it can't actually go and book anything for you and do that kind of work.

0:47:49 - Mikah Sargent
Reid, I want to thank you so much for taking the time to join me today to talk about Turing and again about these companies that have such a big role in the AI that we're all using every day. Of course, folks can head over to semaphorecom to check out your work, but is there any other place they should go to keep up with what you're doing?

0:48:11 - Reed Albergotti
Well, I think the main thing is, you know, sign up for the Semafor Tech newsletter, which comes out twice a week. You'll get you know an email from me and if you respond to that email, I'll get it. So it's a great way to kind of like interact with the audience. And I'm on X, you know threads, linkedin, all those social media platforms as well.

0:48:32 - Mikah Sargent
All right, thanks so much and enjoy your time in your sprinter van.

0:48:37 - Reed Albergotti
Thank you so much. My mobile office today.

0:48:42 - Mikah Sargent
All righty folks, we are going to be back with my quick little story of the week, but I do want to take a moment to ask you, invite you, implore you, to join Club Twit at twit.tv/clubtwit. If you join the club for $7 a month you can gain access to some pretty awesome benefits. First and foremost, you gain access to ad-free versions of all of our public shows. You also gain access to the Twit+ bonus feed that has extra content you won't find anywhere else behind the scenes before the show. After the show, special club Twit events get published there. Let's see what else. Oh, I know you gain access to the Discord server, a fun place to go to chat with your fellow Club TWiT members and also those of us here at Twit. We would love to have you join us there. And if that wasn't enough, you also get to watch the video versions of our Club TWiT shows, including iOS Today, Hands on Mac, Hands on Windows, Home Theater Geeks. It's all available and we'd love to have you in the club. I've got another Mikah's Crafting Corner coming up as well, so be sure to tune in to that if you're part of the club and if you aren't yet, well, join us. twit.tv/clubtwit

All right back from the break and I wanted to tell you all about a late-breaking story from the Mark Bloomberg, otherwise known as Mark Gurman.

No, his name is Mark Gurman, I just have accidentally called him Mark Bloomberg in the past and in my mind it stuck. That's probably the only place where it's stuck, though. So Mark Gurman this morning published a piece about the or rather, it wasn't even. Yeah, yeah, what was? This morning Published a piece about some upcoming hardware from Apple, and this, of course, is a rumor, according to people familiar with the matter, but we're looking at what could potentially be a complete refresh of Apple's entire Mac lineup. Most importantly, I think although we'll get to the specific piece is the fact that what we will see is a sweeping across the board refresh of all of Apple's Mac hardware, meaning that every single one of Apple's Mac models will feature an M4 chip of some sort. This makes sense, rather, given the fact that the company is currently going through the testing process, the beta phase, of working on AI models of all of its various platforms, various operating systems, including macOS, with Apple Intelligence, intelligence and that software set. That feature set really does require the M4 chip to be able to do a lot of local processing and not feed a lot of that off to the server, and so, knowing that that is a necessity, we've seen the M4 chip make its way into the iPad Pro lineup, for example, and from there we will continue to see that roll out.

One of the new models of Mac that we are expected to see, according to Mark Gerben, who quotes sources familiar with the matter, says that there will be a Mac Mini refresh, and this isn't just and I hope John Slanina is somewhere getting really excited this is not just a Mac Mini refresh, meaning that they will tear out its current internals and pop in some new internals. No, no, no. It's supposed to be a complete redesign, making it a far smaller device, approaching the size quote of an Apple TV set-top box. So you may remember, the Apple TV is a little rectangular, or I should say it's a little rounded square, otherwise known as a squircle. That provides access. It's got an HDMI port on the back. It's got, in some models, an Ethernet port on the back and, of course, it's got an HDMI port on the back. It's got, in some models, an Ethernet port on the back and, of course, it's got the power plug on the back and that's it, very small little thing.

The new Mac Mini, however, is rumored to be a little bit taller than the current Mac Mini. The current Mac Mini is 1.4 inches tall and the Mac Mini might be taller than that which makes sense. If we're going to shrink it down, you might need to make it a little bit taller, but it will feature, according to those sources, an aluminum shell. The Apple TV box is about half the size of the Mac Mini and the company has reportedly tested models with three USB-C ports on the back, an area for plugging in the power cable, of course, and HDMI port for connecting the devices to TV sets and monitors. So sounds a lot like an Apple TV, just with some extra USB-C ports.

Now, the current price of the Mac Mini starts at $599. The current price of the Mac Mini starts at $599. According to Gurman, the new model could be cheaper to make than the current model because it's using less material because of some in-house choices etc. Etc, but we don't know if it's going to drop the price at all for this device. According to also Gurman, the company is preparing two versions of the Mac Mini and that we may see kind of a higher end version that features the M4 Pro chip as opposed to the M4 standard chip and, of course, the option to add additional memory, hopefully additional storage and graphics horsepower. It'll be a while before we see some of the other Mac devices get their update, but it could still take place this year. Otherwise we'll see it in the coming year.

I don't want to go through the entire piece because I do want you all to go check out the piece from Mark Gurman. Gurman has given plenty of his time to Tech News Weekly over the years and I don't want to cause value loss in terms of the article. So go check it out to learn more about what has been announced, or rather, not what has been announced, but what has been rumored to be announced, to learn a little bit more. But ultimately, I think a smaller Mac Mini is a pretty cool idea and could mean that more people end up using it as kind of a home theater sort of system, the center of your home theater. I know a lot of people who use Mac minis as kind of local servers, who use them as a way, as kind of an interface for their network attached storage, where the Mac mini is kind of the NAS with storage attached to it, and making it smaller just means that's more possible, to you know, to fit it into a smaller package. So pretty exciting stuff, and I will be curious to see when and how Apple announces these products, given that we are, of course, expected to see Apple's newest iPhones next month. Can you believe it? It's already only just a month away, when Apple typically announces the next versions of its iPhones. So we shall see what comes from Apple, and I wouldn't be surprised if you all see Mark Gurman on this very show in the coming weeks as we gear up for what the company will announce. But with that, I do believe it is time to say goodbye as we come to a close here on Tech News Weekly.

This show publishes every Thursday at twit.tv/tnw. Hey, can I ask you for a favor, tell your friends about the show. Take your friends' phones out of their hands, type in Tech News Weekly in their Apple Podcasts app or their whatever Google uses these days YouTube app and subscribe to the show. There's so many wonderful people who are support of this show hails in comparison to the value that and I say this as someone who tries to maintain a humble self most times and I'm not good at tooting my own horn, but I will toot my own horn about this show because I'm very proud of what we've done with the show and how we've brought in some awesome, awesome, regular co-panelists, so to speak, and the awesome guests that we have on the show every week, and I do think it's very valuable and I would love it if that was reflected in its popularity to the world. So, yeah, bear that in mind as you go throughout the rest of your week into the weekend and maybe send a link of the show to a friend and ask them to become a subscriber, and if you have thoughts, questions, concerns etc.

You can tweet at me at Mikah Sargent, or not even tweet. You can send me a message at Mikah Sargent. On many a social media network I mentioned the club during the show. Again twit.tv/clubtwit $7 a month. You can check out my other shows, including iOS Today and Hands on Mac, which we'll publish later today, as well as Ask the Tech Guys, which has now become Hands on Tech. We'll have our debut episode of the relaunch on Sunday. Really looking forward to that, so be sure to tune into all of those programs as well, and I thank you for being here and I thank you for tuning in every week and I'll see you again next week for another episode of Tech News Weekly. Bye-bye.

All Transcripts posts