TechStuffWhat does Google Search sound like? This week in the News Roundup, Oz explores Google’s new AI-generated audio summaries — and why a simple question like “what’s two plus two?” triggered a two-minute podcast. Then, Wikipedia tried its own AI experiment and faced backlash from editors. On TechSupport, Semafor’s Reed Albergotti recasts Big Tech as characters from The Wizard of Oz and explains the current AI landscape.
Welcome to tech Stuff, a production of iHeart Podcasts and Kaleidoscope iMOS vloscan and carries out again this week. So today I'll bring you the headlines, including what Google Search sounds like and why you're out of stock Cereal may be the result of a cyber attack on today's tech support segment. Semaphor's read Albergotti casts AI companies in the Wizard of Oz.
The funniest part of this whole thing, honestly, was reaching out to all of these companies and explaining to them why they are the Munchkins or the wicked Witch of the East or the West.
All of that in the Weekend Tech. It's Friday, June twentieth. In summary, today's headlines are about Summary's perhaps the most ubiquitous artifact of the age of AI, and today we have two stories about AI summaries, one about a new AI feature on Google Search and the other about Wikipedia's turn away from AI. We'll start with Google. As you might know, Google has this product called Notebook LM, which allows users to feed their own research sources, PDFs, websites, spreadsheets, whatever it may be, and Notebook LM will analyze the data in all kinds of different ways. One of these ways is a quote audio overview, where users can generate this two hander deep dive podcast where a male voice and a female voice talk each other through whatever the topic at hand is. Now, Google is bringing this feature to its search results. According to Ours Technica, the feature hasn't been fully rolled out yet. Users have to opt in by going to labs dot Google dot com slash search and turning on audio overviews. There seems to be a wait list, but one of our producers was able to turn the future before news spread too far, and here's what she found in the search results. Underneath the section that says quote, people also ask, there's an oval button that says generate audio overview. When you click it, it starts working because even AI podcasts take a little bit of time to cook. So while you wait, Google tells you exactly what it's doing, which is quote, understanding the query intent, researching websites, generating transcript, and then in less than a minute, an audio player appears. So our producer searched in Google, what is two plus two? The resulting audio overview was two and a half minutes long, which already tickled her spidery senses. But before we listen to the audio, we checked out all the websites listed underneath the clip as sources. They were a Reddit thread asking why some people think two plus two equals five, a link to study dot com, a Wikipedia page for two plus two equals five, as well as a link to a TikTok meme around the question of two plus two, which I couldn't really understand, let and explain. And underneath all these links is another message quote generative AI is experimental. So here's the beginning of that two and a half minute audio overview for what's two plus two?
So?
What's two plus two? Seems like a no brainer, right four? End of story.
Well, that's what we learn in grade school at least. But the idea that two plus two could equal something other than four pops up in some pretty interesting places, like where.
I'm picturing some kind of crazy math equation.
Now it's less about complex math and more about, shall we say, bending reality a little.
The podcast has then launched into a discussion of instances where some people might say two plus two equals five, like in All Wells dystopian novel nineteen eighty four, in which he ponders a totalitarian government that might insist two plus two equals five. It also includes a joke about it mathematician, an engineer, and a lawyer answering two plus two questions two plus two will always before that. The AI podcast hosts had this to say in conclusion, it's.
Kind of mind bending when you think about it. This simple little equation can represent so much more than just basic addition.
It really does. It's a reminder to question things, to consider different viewpoints, and to be aware of how easily fax can be twisted or manipulated.
My personal view is that the notebook LM product can be pretty great if you want to bring a bunch of your own narrowly bounded set of data or research into one place and then here it played back to you in podcast form. More power to you. The last time I use it myself was over the holidays when I was going to go and see the Nutcracker Ballet, and I kind of wanted to know what to expect so that I'd be able to enjoy it more. And I've fed Notebook LM the Wikipedia page about the Nutcracker, some reviews, some other information, and it summarized all of that and turn it into a ten minute podcast which I listen to on the way to the ballet, and I honestly think that gave me a better enjoyment of the ballet because I knew the context, but I wouldn't have actually done the work myself of reading all the materials. So that was kind of a cool use case. That said, I don't think when I search for something like what's two plus two, I need a audio conversational summary of the randomness of the Internet. It just doesn't really seem very relevant, and it seems more than anything like a solution in search of a problem, which brings us to our next story. There's another Internet titan that tried out an AI feature recently. Wikipedia. On June second, the organization that hosts and maintains Wikipedia, called the Wikimedia Foundation, announced it would run an experiment with AI. The project, called Simple Article Summaries, would provide AI writeups at the top of Wikipedia entries. It was an attempt to make the denser Wikipedia articles more accessible, and the experiment was set to last two weeks, with summaries generated by an NLM called Aya by Coheer. Then came the criticism. According to a form media, the announcement was met with replies from Wikipedia editors who warned that AI summaries were a very bad idea. One editor pointed out that AI's ability to get things wrong could damage the platform's reputation, which is something that editors and the organization have worked very hard on for a very long time. Another editor said quote, just because Google has rolled out it's AI summaries doesn't mean we need to one up them. Some even threatened to quit if Wikipedia didn't roll back the summaries. And here's the thing. Wikipedia relies on its editor's passion for the website to keep this participatory encyclopedia running. So after the editor upraw and just a day after the announcement of the experiment, Wikimedia said they would pause it, but that they were also still interested in AI summaries. It's not clear when they might try them out again, but it is clear how human editors feel about them. I'm not somebody who's implacably opposed to AI summaries. In fact, I use them myself. But what was really important to remember is that AI summaries rely on the input of human work and ingenuity, and with that in mind, I read this article in the financial time that I wanted to share. It had the headline AI alone cannot solve the productivity problem, and it talks about something I had no idea about, which is the average scientist today produces fewer breakthrough ideas per dollar spent than their counterpart in the sixties, and that despite all the headlines about progress, progress is actually becoming more incremental. And here's the part of the article that really got me thinking. Quote, had the nineteenth century focused solely on better looms and plows, we would enjoy cheap cloth and abundant grain, but there would be no antibiotics, jet engines, or rockets. Economic miracles stem from discovery, not repeating tasks at greater speed. The article went on to make the point that large language models gravitate towards statistical consensus. So think about it. If there were NLMs back before the Wright Brothers, all the data that the model was trained on would have suggested that human flight was absolutely impossible. Wikipedia is not going to solve the larger problem of fostering human ingenuity in an age of statistical consensus, But at least for now, you can rest assured that the articles you read on it will be written and driven and created by humans. I've got a couple more headlines for you, including why you might be struggling to get your favorite box of cereal. Tech Crunch reports that some grocery stores are reeling from the effects of a cyber attack. United Natural Foods or UNFI, is a large food distribution company that provide it's fresh produce and products to over thirty thousand stores across the US. Whole Foods uses the company as its primary distributor. Earlier this month, UNFI experienced a cyber attack that caused them to shut down their electronic ordering systems. The system is now back online after more than ten days, but grocery stores are still waiting on some of their stock. And this isn't just an issue in the US. Cyber attacks are also roiling UK supermarkets. The thing is, grocery chains are a prime target for hackers because they have these large, interconnected supply chains, which by definition have multiple failure points. They also have treasure trows with customer data, and of course, when shoppers can't get what they want at the store, angst and headlines are sure to follow. Finally, for today's headlines, ads are a huge business for social media platforms, and TikTok is aiming to lower the cost of production for its partners. According to The Verge, TikTok is adding new capabilities to give advertisers the ability to upload images and give text prompts that can then be used to generate video advertisements. The videos can include virtual avatars holding or modeling different products, just like an influencer mine for brands that they partner with. TikTok has said that all this content generated using AI will be labeled as such and will go through quote multiple rounds of safety review. The biggest story is that influencer marketing transformed the ad industry over the last decade, and it will be interesting to see what AI influencers can actually do because, unlike say, fashion models, whose job it is to show what products will look like when they're worn, influencers have thrived when they make a case for a product based on their personal experience. After the break, we bring you a production of The Wizard of Oz starring your favorite tech giants. Stay with us on this show. We've talked about tech upstarts like open ai and its model, chatchipt Anthropic and its model, Clawed at Length, and We've also talked about the numerous tech giants racing to gain or defend market share with AI products of their own, including Meta, Google, Apple, Amazon, and Microsoft. It's a crowded field, but also an open one in terms of knowing who will ultimately prevail, and it can get quite dramatic their acquisitions, partnerships to get contentious legal fights and hedge rolling at big tech companies as they fight to stay in the game, and recently there have been some big AI power moves. Meta has just made its largest investments since acquiring WhatsApp, putting fourteen billion dollars into a company called Scale Ai. So we thought it'd be helpful to outline them many players the air, current positions, and what might be on the horizon. And here to walk us all through it is read Albagotti, who's the tech editor Sema fool Read Welcome to tech Stuff.
Thanks for having me. Good to be here.
You came up with a very creative paradigm for helping your readers make sense of all the different players in the AI race. Tell us a little bit about it.
Well, it all started because, like you said, I mean, there's this this huge race, and I was sort of going back and forth with my colleague Rachel Jones, talking about Okay, so like, what are the lanes that these companies are starting to take and how do we sort of define that? And I'm sort of writing this thing and it's kind of it's a little dry, it's a little heavy on the analysis, and Rachel's like, why don't we just make them Wizard of Oz characters? And I was like, that's a terrible idea. But then I thought more about it, and I thought, that's actually kind of a great idea. So you know, we decided to try to fit you know, each of the biggest players that we see in the field into these characters and the Wizard of Oz and sort of explain why. And I think, look, I mean the funniest part of this whole thing, honestly, was was reaching out to all of these companies and explaining to them why they are the Munchkins or the Wicked Witch of the East or the West, and you know, getting their responses and I thought, you know, actually, kind I kind of like this because the best part was when I got pushed back from these companies, right like, we are we are not this, we are this character just like, oh, actually I'm getting an insight now into like how you see yourself as a company. And so what I kind of hoped was, like, you know, we would this was like people would write in and say you're a complete idiot, like open Aye is not Dorothy, They're you know, the tin man or something like that. Just get people talking about what are the lanes here and how and how should we think about these these players.
But that said, you did cast open Aiye as Dorothy. Why was that?
Well, you know, Dorothy is is kind of the main character. Right. They are the Kleenex of chat bots, so to speak. They're the winner take all sort of consumer chat bot. They get all the press, but you know, at the same time, they require a lot of other you know, players in this space to make it possible. Right, So you know they they've borrowed technology from Google, right, you know, they're using the data centers or at Microsoft for such a huge part of that. So it's this supporting cast that kind of really makes them who they are. I think, so that was that was the justification there. Do you agree or do you want to do you want to fight me on that?
Well, I think I'll you've thought about this more than I have. But I think I agree with you that despite my name, I think that I agree with you that that that opening I definitely has the main character energy in this story. But I have to ask you about her cost of supporting characters the tin man, the cowardly line, and the scarecrow. How do you assign each of these crucial pots?
Well, okay, so let's start with Let's start with with Google, the cowardly lion. Google has, you know, all this power, right, They've developed this technology for years. You know, the attention is all you need paper, right like that. That's what really laid the groundwork for CHATCHYPT. But they kind of sat on it. And I think a lot of that was out of fear. You know, these chatbots go off the rails. They're crazy, Like, what are people going to say if we just start releasing this stuff into the wild. And so they didn't and they got beat right, chatchapt came out. Google has had to completely reorient itself around this race now.
So what you're saying is is Google could have had market leadership here, and because of fears about cultural blowback or erosion of that cool business, they basically sat on the technology.
Yeah, that's very well said. They're a big public company. They have to be sort of cautious, and you know, I think that's somewhat in there, has become part of their DNA. But they've had to get really bold, right and they started to release new products. The notebook LM is one of the big successes. They're throwing more spaghetti against the wall. They're finding that they actually do have courage, right, which you know, of course, if you've watched The Wizard of Oz, you know in the end they all kind of have the thing they thought they were missing. So you know, tin Man is Microsoft this well oiled machine. If you remember the tin Man was always putting oil and oil can. They seem a bit rigid, right, but in the end, like they kind of do have a heart. I mean I think they I had an interview with Satia Nadella recently and and Kevin Scott, the CTO, that it's really clear they really do want to sort of like enable these developers. I think they have sort of a I think a less ego, a bit more of a selfless kind of attitude around this stuff. They're building the infrastructure for this industry and letting others kind of use their creativity to benefit, so they really do have a heart in the end. And then I think Amazon, you know, this the scarecrow, right, no, no brain. When I think about the beginning of this race, you know, much like Google, people were really putting Amazon down and they thought, oh my god, they've completely missed this, like Amazon's got nothing here, and totally ignored the fact that actually they had been thinking about this for years. And even though you know, maybe they're sort of fighting for market share in this new AI race, I think they're pretty well positioned to profit from the growth and you know, just the the market demand for tokens for compute.
Are they the market leaders in data centers Amazon today?
Oh yeah, I mean for sure in the in the cloud, in the cloud business. And I think that the question though is, you know, can they remain the market leader in the AI you know era right, they've been the market leader in cloud in terms of just you know, normal compute AI data centers is it's a whole new ballgame. AI runs on what we would call a GPU, right, and so everyone is now you know, racing to add these GPUs or build new data centers that are built around these GPUs and it's a totally different business model and concept, so it kind of restarts the race in some ways.
Okay, so we got Open Air as Dorothy, Microsoft as the tin Man, Google's the cowardly line, Amazon is the scarecrow. There are also some wizards and witches who need to be accounted for Glinda, the Wicked Witch of the West, and of course my namesake, the Wizard of Oz. Who's who in this cost.
So let's start with Perplexity being, you know, the wicked Witch. As we found in Wicked. You know this, this witch was kind of misunderstood, and I think that's sort of how Perplexity became the absolute villain, especially in the media industry, because they, you know, were were I guess, republishing almost like articles without proper citation. I think they very quickly changed I mean, it wasn't a policy. I think it was a mistake and they very quickly changed them. And they're actually have all these media deals where there are revenue sharing agreements between Perplexity and the media companies. And I think that Wicked image is starting to fade a little bit from Perplexity. So I put them in the role of the Wicked Witch, and Arvin strinovas the CEO Perplexity, responded to me on Twitter with a I think AI generated repurposing of the song from Wicked.
I've got the final close my eyes and leap. It's time to try to find Google. I think I'll try to find Google and you comple me down.
Right, So they're disrupting Google. Therefore they're they're really the good guys. So you know that was I thought that was fun. That was a fun response from Arvind. I think, you know, we put Apple as the role of Glinda, the good Witch, who you know sort of over over promises and underdelivers. I think that's sort of the you know, using a little bit of the old and the new Wicked version of that that character. Obviously, you know, that's based on the fact that they, you know, they had this huge event where they promised the world around you know, AI, they were going to do these amazing things with Siri and your phone, and it turns out, oh, they hadn't actually really built any of this, and it's not clear that they can.
And what what's your reporting suggests? Why has it been such a bust?
I think that was a train wreck that anybody who's been following Apple has kind of seen for a long time because they just have ignored AI and you don't see the Apple people are not really like mixing with the rest of Silicon Valley, and I think they just didn't quite understand this whole AI thing. And and I've also written that, you know, I this is a bit this is kind of my view, and I don't know, I haven't heard that any people agree with me on this, but I think they also, you know, because they position themselves as this privacy focused company, they that they also shied away from AI there because it does require gathering a lot of data right on people and personalizing things in ways that make people maybe feel like, oh, I'm being tracked or what have you. And it turns out that that's you know, that's now everything in the tech industry, and they're they're way behind.
And then of course the Wizard, just by having a movie named after him, is actually no very important character.
Right exactly. So I cast x Ai in the in the role of the Wizard. You know, I think x Ai has this huge, you know, colossus data center, right, and they've they've promised that, you know, it really kind of made me think of them as that is the fact that they're going to understand the secrets of the universe with AI that theirs is the Is that literally?
Yeah?
I mean this is like literally, you know. I mean I saw Elon Musk in a recent interview saying that their model grock is based on basic physics and that's where it gets its ground truth from. And so you know, their effort is to actually ultimately figure out, like, you know, a unifying theory of the universe. You know, I think that's a lot of it. It remains to be seen what's behind the curtain. There.
Two other companies we haven't spoken about yet are Meta and Anthropy. Where do they fit in?
Yeah, well so Anthropic I cast as in the role of the Munchkins. I felt a little bad doing that because you know, I think there's the Munchkins have this negative connotation. But actually I was reading about it and researching this article, reading about the Wizard of Oz and trying to understand the different characters on a more nuanced level. And without the Munchkins, there is no Wizard of Oz. Right. They they they're the people, They're the ones who populate this universe, right, And I think I think what Anthropic is doing, and I think a lot of people don't really realize this, is they have become the de facto coding model for AI, and that is really what's driving all this, you know, the excitement and AI right now, it's not so much chat GPT, even though I think that's what the broad population sort of thinks of. But in the tech industry, what I think really excites people is that this stuff, even if it never advances beyond just being able to generate code with AI, that's going to completely change the world because you know, code is already so powerful and can automate so many things. The reason that the whole world isn't running on code is that it's too expensive, and so Anthropic is kind of like bringing that power to the people.
So when I read about vibe coding, though I read about CUSS or AI and stuff, But you're saying that Anthropic is a real big player in the world of vibe coding.
Well, yeah, I mean Cursor is running on Anthropics code generation model, and so is Replet and you know a lot of others.
I mean, it's Anthropic, it's really underpending this vibe coding revolution exactly what about Meta. It's a big week for them. I think you cost them in the in the movie before their then news announcement, But who are they and the Wizard of Alls? And did they did the news announcement change you're thinking at all?
Well, of course, of course I'm going to make an argument that it didn't change my thinking.
But I would I would.
Encourage people to totally disagree with me and argue with me on this. But no, I mean Meta I cast as the yellow brick road. They're they're open sourcing these AI models, and similar to to Anthropic where it's like that, it's an amplified effect. Many companies are using Meta models, these Lama family of models to run their own custom software because you know, it's free and it could also be fine tuned, it can be run locally, so it's actually you know, having this big impact that a lot of people don't really see. And then of course this recent acquisition of Scale AI.
You know, we'll see whether or how does that You don't talk about the.
Yeah, you're right, I called it an acquisition, and it's not. I mean, it's a it's a forty nine percent of the company takeover, but not an actual takeover and and yes it's not an acquisition, but really kind of puts them in. The CEO, the co founder and CEO of Scale AI is now going to go work at Meta, but Scale will still continue as an independent company. What I think Meta is interested in them for is their ability to generate a lot of data, you know, to train these models. So you know, these AI models trained on the entire Internet. I'm sure you've kind of heard these people say that they sucked in all the data and the entire Internet and train on it. Well that's somewhat true. There are these huge databases that are free that anybody can train on, and they've sort of tapped that up, like using that data can has sort of They've gotten as far as they can with that, and now they have to generate more data and it's getting very expensive. And so companies like Scale have these huge networks of contractors who can generate a ton of data. And it's actually like PhDs and like you know, high level business executives who are doing this I think kind of for fun. I mean they're getting paid, but they sit there and they do like very complex stuff that can then be used by these models to learn and become much more you know, specialized in certain areas. And part of the deal with Meta is that Meta has sort of pre paid for a bunch of Scale services to get this data.
So this is really giving itself a competitive edge. The New York Times ran a story saying though that this was like part of socker Bug's personal quest to build a super intelligence lab. So I wasn't sure how to balance that with Yeah, what I'd understood about Scale was that it was a company that sort of specialized in creating this very high value data that's not available organically.
This is sort of my my take on this. I mean, it's easy to focus on, you know, the Scale CEO, Alexander Wang is going to run it, and it's all this talent. I mean, Wang is not an AI researcher. He knows a lot about the AI industry and he's one of the world's most experts on you know, data being used in AI. He was roommates with Sam Altman when basically the groundwork for chatch ept was being built, and that sort of brought him in. I mean, he didn't and I've listened to interviews with him where he said, like he didn't really understand what Opening Eye was doing, but it was clearly like this great opportunity and so yeah, I mean that help them with recruiting, and it's it's great, you know, it's great for marketing and it all, it all helps. But I think the bread and butter of that deal is really about Metas saying, well, Okay, we might not be the most cutting edge AI company right now, but we're putting a lot of money into having the best data.
Read this the question that is normally a cache, but in this case, I think it's appropriate. How did this movie end?
Well, they all get to the Emerald City, which is Agi, and it was all set in motion by this tornado which is chatchbt. No, I don't think we know this one hasn't This one hasn't ended yet. I think it's I think there are many alternate endings, just like all the Wicked and and and probably other fan fiction around the Wizard of Oz.
But if you look looking at the landscape, maybe you had to bet on you had to bet on one company. I mean, who do you think is really positioning themselves in the smallest way for this immediate future of AI.
Yeah, if you had to bet on one company that's a that is a really that's a tough one. I mean, because you know, I don't think this is a winner take all race either. I think that if you look at the compute layer, I mean, this is a growing pie that will be divided up. I mean, Opening Eye might be winner take all in the consumer chat bots space. I could see them becoming sort of the Google of of of that era. But you know, I look at Google right now and I think they have there. It's interesting because I think, on the one hand, they have so much promise. I mean, they have the best models right now, they're applying them in the most diverse ways, right they have because one is waymo right. They have robotaxis, which are probably ultimately just gonna run on Gemini at some point in the future. They they're doing stuff in other you know, humanoid robotics as well. But they've also got these, you know, they've got their own consumer chat bots, They've got you know, a lot of consumer distribution in the software space. So I think they're very they're set up for success. The interesting part though, is that their core business of search advertising is also It's like, it's not even just that it's under threat, but like the whole the whole concept of of like the Internet economy running off of advertising dollars is up in the air, like we don't even know if that's the future. So they're challenged and they're having to completely change, like reorient themselves as a company. But at the same time, they have such promise, so I don't know, I mean, it would be sort of it's sort of like a gamble to put money on anything. But I mean, I think they're just very well positioned. I think them in open AI, I mean open a eye, just what they've built is so is so valuable, and not because of their models, not because they have the best AI, but because they're just they've they've become this consumer products now and that is we've seen in history that like that is something that you know just will be hugely valuable in the long run. It's kind of like their game to lose there.
On that note, read, thank you so much for your time, Thank you so much.
It was great.
That's it for this week for tech stuff. I'mos Voloshin. This episode was produced by Eliza Dennis and Victoria Domingez. It was executive produced by me Karen Price and Kate Osborne for Kaleidoscope and Katrina Norvel for iHeart Podcasts. But he Fraser is our engineer. Jack Insley makes this episode and Kyle Murdoch wrote our theme song. Join us next Wednesday for text Stuff the Story, when we'll share an in depth conversation with Yasmin Green, the CEO of Jigsaw, about the tech that could change town halls forever. Please rate, review, and reach out to us at Tech Staff Podcast at gmail dot com. We want to hear from you.
