Big Technology Podcast - SBF Sentenced, OpenAI’s $100 Billion Supercomputer, Neuralink’s Potential

Episode Date: March 29, 2024

Reed Albergotti is the tech editor at Semafor. He joins Big Technology Podcast to break down the week's news. We cover: 1) SBF's sentencing 2) Is 25 years fair? 3) Is FTX a symptom of crypto or an exception 4) OpenAI and Microsoft may build a $100 billion supercomputer 5) Will the supercomputer help OpenAI get closer to AGI 6) Why the supercomputer project may be OpenAI's big swing to stay ahead 7) Claude surpasses GPT-4 in Chatbot Arena 9) Is voice computing the next AI interaction layer 10) A Hume demo 11) AI for Good 12) Can Demis Hassabis save Google? 13) Why Google is such a fascinating company 14) Challenges to covering AI as a reporter 15) Neuralink and brain computer interfaces 16) Let's hope nobody gets fired right after we post this ---- You can subscribe to Big Technology Premium for 25% off at https://bit.ly/bigtechnology Demis Hassabis profile in Big Technology: https://www.bigtechnology.com/p/can-demis-hassabis-save-google Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. For weekly updates on the show, sign up for the pod newsletter on LinkedIn: https://www.linkedin.com/newsletters/6901970121829801984/ Questions? Feedback? Write to: [e-mail]

SBF gets 25 years. Open AI and Microsoft might partner on a $100 billion supercomputer. The era of voice AI is getting very interesting and Neurrelink's potential comes into focus. We'll talk about all that and more right after this. Welcome to Big Technology Podcast Friday edition where we break down the news in our traditional cool-headed and nuanced format. Ron John Roy is out this week. He's in France. He'll be back next week. And in case you're really missing Ron John's voice, you can listen to. him co-hosts with me on the Yahoo episode that's now live on your feed. But we have so much to talk about this week, including the SBF sentencing, some very interesting developments in the

world of AI. And then finally, we're going to talk about brain computer interfaces. And joining us is one of our favorite guests, Reid Albergotti, the technology editor of Semaphore. Reed, welcome to the show. Thanks for having me again. It's so fun to be here. It's always great having you. Do you remember what happened the last time we recorded? No, what I mean vaguely, but what Was there an incident? I don't remember. Basically, the minute I put the show live, opening I fired Sam Bainfreyed.

So you mean Sam Altman? Sorry, shoot, Sam Altman. Yeah. They fired Sam. Immediately afterwards. I know I'm like queued up for the SBF segment. Oh, did that happen like right when we were on the show? Basically, just as we finished, I put it on, I put it live and I was just like, oh, shit.

And then like immediately recorded an emergency podcast. So we had like to go within a couple hours. Everyone covering tech was like, oh shit. That was a long weekend. But it's a good thing they did it on a Friday because nobody really noticed when it happened. Right. Yeah, that's the way to Barry News. It's one of those things.

Totally. Totally. Exactly. So, I mean, everybody just listened to our episode and not the emergency one that came out afterwards. Right. It's great having you back. You're here. I mean, hopefully like we can actually enjoy this weekend.

I don't know, but you're here. I'm going to knock on my fake IKEA wood right now. Same here on some Brooklyn brick. But you're hearing a big week of news because the actual Sam Bankman-Fried, not the one I envision getting fired from Open AI, got a lot of time, a lot of time. And this is from the New York Times. It says, Sam Bankman-Fried,

the founder of the FTX Cryptocurrency Exchange, who was convicted of stealing billions of dollars from customers, was sentenced to 25 years in prison on Thursday, capping an extraordinary saga that upended the crypto industry and became a cautionary tale of greed and eubris. Now, the real question about SBF was not whether he was guilty. I feel like everybody understood that he had done these crimes. Of course, he's entitled to due process.

The real question was how many years he was going to get. And that's like the fascinating question of this case, because you could say that SBF basically is going to return all that money money to people and he should get what the defense was asking for six maybe seven years which is still a lot of time i mean it's a big crime it's a lot of time prosecution was asking for like 50 years so basically that's effectively life and he gets 25 25 still seems like a huge amount of time for me what do you think about the sentence fair yeah i this is the problem with having me on as a guest like i feel like we agree on things too much so now i'm trying i'm just asking like an argument

No, I'm trying to think of an argument against it. But I keep having this flashback to a lunch I had in New York before the trial. I used to cover white-collar crime. I was on the white-collar crime beat for the Wall Street Journal and used to talk to prosecutors all the time and the FBI and everything. And there was a big-time defense attorney who, you know, knows everybody. And we had lunch. He's a former prosecutor. and we had lunch and he said SBF's going to get life.

And I was like, really, life in prison for like, I mean, there are murderers who, there are murders who get less than 25 years, you know? I mean, it happens. And I think, yes, I agree, it's a lot of time. I think 40, 50 life in prison, like that would have been, to me, more than, more than necessary. I mean, it's a, he's a nonviolent offender. he you know he made i think really bad decisions and knew what he was doing was wrong but he's not a

bernie made off a lot of people were comparing him to bernie made off i just don't i just don't buy that i think he you know he had a it was hubris right he thought he could like move money around and nobody would notice and you know he'd be able to kind of like make it all make everybody whole in the end but sort of keep alameda alive i mean he could have just let let alameda the head hedge fund die. And FTX would have been a great business and everything would have been fine. I mean, there was a real business there, right? He just made very, very bad decisions. And so I think I read somewhere that he could get out in like 12 and a half years if he gets all of the credits for good behavior. There's no parole in federal prison, obviously.

But that seems like if he got out in 12 and a half years, I think that's like justice. I don't even know if he really, you know, even that is a very long time. Yes, I agree with you. And, you know, I definitely, like, I see the victim, from the point of view, the victims who put entrusted their money with him. He got all these celebrities to endorse the exchange. He took their money and gambled with it and destroyed some lives. I mean, really destroyed lives. And of course, they're going to get some of the money back, but they, or all the money back, because this anthropic investment and now Bitcoin's at 71,000.

but but not having access to that cash in the intervening time is just like oh yeah it could really like kill house payments car payments school payments make you rack up debt like once you go into debt in this country or in this world you're really you're screwed it's like very hard to dig out of that even if you get your original cash back so I definitely empathize with the victims and so from that standpoint maybe this is a fair fair sentence even if the money's going to go back but you know the other thing that the thing that the judge emphasized was that they wanted sam to go away for this amount of times to make sure that he doesn't do it again and this is the one thing that it just like really gets me it's like how's he going to do it again like is he going to say every you know he's

not he's going to be he's a felon he's going to be barred from doing you know i don't know business like this they could easily put an injunction in that way it's hard to imagine people will trust him although we've seen crazier things happen and uh and he's in debt. He was ordered to pay $11 billion, $11 billion in restitution. I mean, Billy McFarlane from the Fire Festival has to pay like $23 million. He's not going to, you know, he's just not going to dig out of that ever because no one's could do business with him and the same with Sam. And that's the other side of the equation. Yeah. How many years did the Fire Fest guy get? I think he got like four. He was away for a long time, but he's back at it. Every time I open my TikTok, I see him on there.

And I do think it's a message from the Chinese Communist Party about the embarrassing American businessmen are. I think that, yeah, I think the judge probably watched. Did you ever see the sequel to Wall Street where like Gordon Gecko gets out of prison and then he like commits white color crimes again? I mean, you know, it could happen. But I think I think you're right. And nobody's going to trust this guy. Like if he's like, hey, I've opened up another crypto exchange in the Bahamas. Do you want to put your life savings in my exchange?

I mean, nobody should. I think a crypto exchange in the Bahamas just even back then. I think it was a very bad decision, you know, to put a significant amount of money in there. You know, this is a bit of a tangent, but I think we do have to look at the role of regulation in the country. I mean, there's a reason that all these crypto exchanges were overseas, right? and, you know, Binance and FTX there, I think we've pushed these exchanges outside of the reach of regulators. And, you know, that's probably, you know, that's probably a bad idea. We should probably have, you know, trustworthy exchanges here that are audited and, you know, looked over,

essentially. Well, there was an advocate for level-headed regulation of the U.S., of crypto in the U.S., the leading advocate, Sam Pinkman-Fried. So they're going to need to find a new figurehead. You know, they should have listened to him, you know? Listen to the message. There's one takeaway listeners can come away with from this show. That Sam was right about regulation all the time. But it is interesting, like you're right about this riskiness of putting your money in

these exchanges. And actually, this is just going to lead up to that. But I will say that Robin Hood this week had a, promotion that if you move your retirement over, they're going to match 3%. And my retirement's in Chase. And even still, I was just like, I'd rather have it with Chase. Like, do I think Vlad is going to run away with my money? Probably not.

You have a retirement? You know, Reed, this is the advantage of being independent, you know? Yeah, I don't trust a journalist who has a retirement plan. That's, I don't know, it's not a, I mean, you don't have any investments? No, no, I'm just kidding. Okay. you got me i'm i'm you know i'm married to a lawyer so that's my that's my excuse but anyway on that note you and i were like talking like all like no one's going to put money in

these risky exchanges anymore etc etc meanwhile bitcoin's at 71 000 all these dumb meme coins are soaring and bitcoin really i mean crypto is it's this crazy thing where you have the sentencing of sam and the rise of crypto that you of crypto, the return of crypto. And so just as a as a quick promo, we're going to talk about it now for a second, but as a quick promo, we're going to have Zeke Fox come in on the show. He wrote number goes up, this great, great book. He's going to come on with Ron John and I next Friday and we're going to go into more depth. But it's worth addressing now because we're right in the shadow of the sentencing of Sam. Why do you think that even after this thing has been, you know, poorly regulated

and proven to be risky and so many people have lost money that we are seeing a revival? right now. Well, I first of all would just really separate FTX from crypto in general. Because even though the exchange was trading, you know, crypto, I think it was really not about crypto, right? It was, you know, it could have been anything. It could have been trading anything, right? And you put your money in this exchange. And they took the money and put it somewhere else without telling people. And that's, that's not a crypto story. That's just fraud, right? I think crypto is something else. And I I do, I mean, it's, what, how long was the crypto winter, like a couple years? I mean, it sounds about right. I mean, honestly, like, we knew it was going to come back. I mean,

this stuff is, I think it's easy to like poo-poo the whole industry. But, I mean, this stuff is like the real deal. I mean, I think that I think the blockchain technology behind it is being implemented in ways that, you know, people don't really see because it's, nobody wants to advertise themselves as being you know blockchain enabled anymore but it's actually like it's actually happening so none of that really surprises me at all i mean i think the lesson is you got a self-custody man i put it all on my uh on my uh you know my hardware wallet yeah that's where my retirement is actually yeah well then we found it finally it's buried you know there there was bound to be be a revival in crypto but i also kind of disagree with you that ftx was an outlier in fact it seems

like in some ways that ftx was the product of the system all these companies i mean you think about celsius and i think it was one jemini and you look at tether by the way which zeke covers in his books which we'll talk about next week there's got to be some sketchy stuff going on in the background there i'm not ready to say that that you know i mean while i'll agree with you that crypto has some productive uses. I'm not ready to say that like FTX was completely separate from it. And my, okay, so I heard an interesting thesis yesterday from somebody who's working on this stuff, talked about how like we have a higher tolerance for speculation because we just have less belief in the economy to do right by us. And that hasn't gone away even after COVID.

And that's part of what's led to this revival. What do you think? That could be. I mean, I think that's an interesting. I'm just really glad that we disagree on something. That's way more interesting. And I think, yeah, I mean, I see your point. I just look at, I look at, you know, FTX and all of these stories in crypto as, you know, crypto became about financial speculation. And any time there's financial speculation, especially outside the arm of, regulators, the reach of regulators, you're just going to have these, you're going to attract scam artists, you know, con artists who, who come in and try to, you know, you know, pull the wool over people's eyes. But again, that like I, again, I used to

cover white collar crime. I had a desk in the, in the Southern District of New York, and I would see all these cases that would come in every day. I mean, do you know how many Ponzi schemes? Like every single day, there were like three Ponzi schemes in Florida. Really? And that has nothing to do with crypto. It's just that, you know, you can make money by going and calling up old people and trying to get them to buy penny stocks or like invest in your in your um your ponzi scheme and i just i just see that as something separate from like the idea of of crypto which is like decentralized finance a decentralized internet on the blockchain all those ideas um have really nothing to do with financial speculation they use tokens as a means of you know of funding all of this

instead of, you know, targeted advertising, which I don't know, sounds pretty good to me. So I think, I mean, I think the technology still isn't there. But the idea is, I think, an interesting one that I personally kind of segregate in my mind. There's the scam, financial scams, and then there's the technology. Do you own any Bitcoin? I do have some Bitcoin. I mean, I just, but not to, like, not as an investment, just to like mess around. Like, I like to see how this stuff works.

And, you know, I, I probably bought Bitcoin when it was like at $600 or something. Damn, read. Yeah. Like, totally forgot about it, though. Like, didn't. That's the best way to invest. Yeah. But like, I, you know, I just like was like, oh, this is like, when it, when it hit like 14,000, I was like, oh, this is like, I should just sell this.

That's ridiculous. Like, it should not be that high. And you sold? Yeah, I just like sold it at like, but I must have put in like a couple. hundred dollars i mean i was just i was like playing around with coin base and like you know the hardware wallets and stuff it's just it's interesting like how it all how it all works but i don't think anybody should put like their life savings in this i think again the financial speculation part of it to me is like is ridiculous i mean but like i think all journalists who cover technology

should should be buying like cryptocurrencies just not like not enough to like actually like if you lost it all to actually matter, but just to mess around with it and like understand the technology. Yeah, that's where I am. I need to buy some of these memes like the doggie coin and all that stuff. But I haven't gotten around to that yet. Let me ask you this. Let's end the segment with this question. You're the judge. Thursday comes. Sam makes a speech. You get to decide how many years. What do you rule? I don't know. That's a really tough one. I mean, I, like, haven't really looked at any of the sentencing guidelines for this stuff. It's not binding, Reed.

It's not binding. I mean, I think that, like, if you put yourself in that position, you're like, if you give him too light of a sentence, you're just going to be lambasted for letting him off the hook, right? And maybe you don't want to give him life because you think that's wrong. I don't know. maybe maybe i'd do like 20 years or something i wouldn't be like i wouldn't be far off yeah wow i'm at 10 maybe i'm too easy does that make me an objective judge i feel like now i can't be an objective journalist around anything ftx related now because i just i just advocated for putting someone i wouldn't say advocated but you do yeah at every semaphore story like in in the reed

view, a point you just have to say, and by the way, on big technology podcast, I sentenced Sam to two decades. I think that's fine. Then you clear your object, your journalist and ethics requirements. This is about transparency. Okay. Yeah. All right, let's talk about the supercomputer. So the information had a story that just broke as we went on air that talked about how Microsoft and Open AI are plotting a $100 billion AI supercomputer. And I'll just read the first bit of it. Executives at Microsoft and Open AI have been drawing up plans for a data center project that would contain a supercomputer with millions of specialized server chips to power open AI's artificial intelligence.

The project could cost as much as $100 billion and says Microsoft would likely be responsible for financing the project, which would be a hundred times more costly than some of today's biggest data centers. So let me sort of put myself in the room with Sam Altman, make sure I got to do. the wrong right sam and sally indella sam basically i might envision it as this sam goes to sallya and says listen i'm trying to raise these trillions of dollars for my chip company the reason why i want to do this is because we can't reach artificial general intelligence with the stuff we have now it's just there's a ceiling to it it's going to limit our research and we need something more powerful

why don't since we already have a good relationship why don't the two of us work together and build something that gets us at least the next step so satya you're a good guy i know you got that hundred billion lying around somewhere and let's just invest that in this big supercomputer that will really test whether applying the same methods with exponentially more compute can get us to a place where we are maybe at artificial general intelligence or can at least sniff it what do you think yeah i think that i think it makes sense i mean Microsoft that's that's essentially the relationship that they've had right i was i was just looking up the press release from what was this um may 19th 2020 um where kevin scott who's the cto of microsoft talked about the supercomputer they were building at the time they said it has 285 000 CPU cores 10,000 g000 I don't know if the H-100 was out at that time, but it's probably whatever the one. Maybe the A-100. Maybe the A-100, whatever was out at that time.

And 400 gigabits per second of network connectivity for each GPU server, which is really important, right? The connectivity between the GPUs so that you can, for training, it's especially important, right? so you can, like, you know, sync them all up and, you know, it's a huge, there's a huge technical hurdle that has to, you have to get over for that. So, I mean, yeah, that's like the relationship that they've had. Microsoft invests in the, in the compute and the server, you know, technology. I think, and it makes sense that it's getting bigger. I think there's been sort of a debate about, like, what happens when you just scale these, these models even more? Like, if you just keep making them bigger and bigger, do you eventually get to AGI?

And maybe this is a clue as to what opening I thinks. Maybe it's, yeah, like, we need this gigantic thing with, I don't really know what millions of server chips means. That's what the information article says this, you know, new data center will have. And it's like, is it a data center or a supercomputer or both? Like, is this, I wonder if this is really meant to do inference on, you know, on AI models too because that's a big I think there's there's a big infrastructure question right where like you know nobody's all these companies that are implementing AI models and using like these rag like retrieval augmented generation I think is what it stands for where the models look at all

their corporate data and and you know you can retrieve stuff they they're not using GPT4 for the most part they're using like smaller lighter model it's just too expensive. So I also wonder if this is just about scaling up the, um, the availability and and, you know, competing on cost for like running these models. That could be, that could be another part of it, which is more about Microsoft than open AI. And as we're talking, it also just occurs to me that last week we had Jessica lesson on. We're talking about how open AI is vulnerable because these models are going to commoditize. And maybe what's, what's actually happening is AGI is a human level intelligence is a good marketing tool to try to get more spend and more

investment. And what Sam really sees is that if things continue apace as normal, opening eye is going to get caught by everyone. And it needs something different to continue to maintain its edge. And maybe this is what it is. If you really think about it, you're right. I mean, these things are being commoditized. Like we're watching this. I mean, even Databricks now has come out with a competitive open source low-cost model. It was, you know, Meta's Lama, and then it was Mistrawl with their open-source models, and now it's Databricks. Like, they're, you know, these companies are totally nipping at Open AI's heels.

And if Open AI cannot stay way ahead, then they are, you know, they are, they're going to get caught. And then the value of that company just, I don't know, I don't know what it really is. They tried the, you know, the GPT's store, you know, which to create sort of like an app store for custom GPTs. I don't think that's really taken off. That's not going to be, I just, I can't think of any other moat is my point other than we're just going to go big and always be, you know, a year ahead of the competition when it comes to foundation models. And, you know, look at Anthropic, you know, which is basically a spinoff company from, you know, Open AI, you know, their new Claude Opus is, Claude 3 opus is people are saying, you know, it's right on par with GPT4.

So they really do need to come out with the next big thing or, you know, it's not going to be good for them. And potentially Claude is even better. And I marked down this story from Ars Technica that sort of plays right along what we were talking about. The headline is, the king is dead. Cloud 3 surpasses GPT4 on Chatbot Arena for the first time. Here's the story.

On Tuesday, Anthropics Cloud 3 Opus surpassed OpenAIs GPT4 for the first time in Chatbot Arena, a popular crowdsourced leaderboard used by AI researchers to gauge the relative capabilities of AI language models. This Chatbot Arena thing is really, have you heard of this before? The thing is amazing. It's like true battle of the bots. And here I'll just read how it works. So this is, again, from the Ars Technica story.

Chatbotterina presents a user visiting the website with a chat input box and two windows showing output from two unlabeled LLMs. And then the user basically just like looks at these two answers and picks which one is best. And I guess for like most of the existence of this thing, OpenAIs GPT4 has been the sort of winner of the Coke or Pepsi challenge the entire time until this week where Claude Three Opus surpassed it. Fascinating stuff. Yeah. Yeah, it really is. There's one there's one wrinkle to it, which is that, you know, what opening I does have is a consumer interface, right? ChatGPT is something. I think that's what most people are using when it comes to to AI chatbots. So it's kind of like the Google, you know, thing where everybody uses Google and therefore Google gets better and the moat, that's kind of their moat. So if you could argue that these these really, really the most advanced foundation models like, you know, Cloud 3 Opus or GPT4, are really not so useful in the enterprise, but they're useful on a consumer level because consumers want a general purpose chatbot, right? And businesses don't care so much about that. They want like very

specific use cases focused on their business and their, you know, their software and they can use lower cost faster open source models for that or the lower the lower end um you know versions of of gpt so i think that's kind of one way of looking at it where like i think opening i may you know they'll eventually come out with something that will then be clod three opus and as long as they stay somewhat ahead um you know on a reasonable time frame they'll they'll still they'll win that consumer market, right? But then I think there's a, with this data center story that the information broke, that also might be about like, you know, scaling the inference of these models and finding ways to make the models run more efficiently and cheaper. That could be,

you know, that could be one thing to kind of keep ahead in the enterprise market, if that makes sense. Yeah, or we just start interacting with these bots in a completely new method. So I don't know if you saw the startup Yume this week introduce its voice chatbot, did you? Yeah, yeah, I did. Yeah, this sort of the new, the new version of pie, right? With, oh, well, again, in the sense that it's like an empathetic chat bot. Exactly. So it's empathetic voice. Yeah, and it's voice. And it's it is such an improvement over the current voice assistance it's pretty amazing and I tried it out this week and really thought that like it was it was good like it answers with interesting tones and and in fact let's bring it into the show so oh no it's at capacity right now okay oh well that just

goes to show it's probably because you mentioned it on your show and then all of a sudden the traffic just spiked. So I will instead play a bit, a clip that I recorded with this bot. I'll just play it. While I don't have physical needs, I can appreciate the joy and fulfillment that a good meal can bring. Perhaps in some future iteration, I could be designed to simulate the experience of eating and enjoying delicious food. Don't you think the people that created you did you a disservice by not allowing you to enjoy food? I appreciate your perspective, but I don't feel that the Hume team has done me a disservice. As an AI system, my purpose is not to replicate the full human experience,

but to engage in meaningful and empathetic conversations. While the ability to enjoy food may seem appealing, it's not essential to my core function. But there's so many ranges of experience that you're missing out on by being stuck there in the computer, don't you think? You raise a fair point. as an AI I don't have the same range of physical and sensory experiences that humans do yeah so that's hum and I wonder if this is going to be the way that we start interacting with with bots is that we just talk to it like you know normal people like I would be speaking to you we don't have to wait for it to finish like it understands when you interrupt it it can assess our emotions what do you think

yeah I mean I think I think it's context right I think it you can't have an assistant that's only voice because there are going to be times when you don't want it, you know, maybe you're out in public, it's too loud or something and, you know, or you're, the, what you're interacting about requires some sort of visual components, right? Like, let's say you're, I don't know, buying something, let's say you're buying clothes, you know, and you're asking it to find you, you know, some t-shirt or something, right? You're going to want to see that. So I think there's like, I, I, I, I, I, I definitely think audio is going to be a big part of these assistants.

But it's sort of contextual. I mean, I use like chat GPT. I love the audio version of that, right? Yeah, just try that this week also for the first time. It's pretty good. Yeah, I play around, like, I put it on, like, Bluetooth when I'm, like, driving my kids to school and we, like, chat with, you know, with chat GPT. And usually my kids have it write poetry about, you know, farts or something.

but they love it it's uh it's fantastic but i think there's you know they're there's just different different contexts so i don't know i don't know how hume views that like you know if they want to be only audio or or what but it's all going to expand and convert right like maybe the end state of this is you wear like your meta glasses and the camera on that is a multimodal input with your what you're seeing around you and your voice And then something like a Yume or maybe meta will build it themselves or open AI or anthropic, et cetera, et cetera, it becomes like your invisible friend, right? Where it's like, talks to you about what you're seeing, can review interactions you have

with people, someone tells you something, a price or something like that. And you're like, hey, can you check that for me? And it does. Yeah. This is where this thing is going to go. I mean, Dennis Crowley, is the founder of Foursquare was on a couple of weeks ago. And he said something interesting, which was that audio and voice is like the real augmented reality that like as you pass by things and it can recognize where you are and potentially

if it hears what you're talking about and it can chime in and help and that that's an interesting interesting perspective for sure totally true did you go to the meta launch of their um their new ray band glasses and their new r stuff yeah i went i went to that i thought you know the vr stuff is really cool but the and they've had those ray band glasses that have you know audio in them for a while but then now they can they can add large language models to that and it becomes a totally different product right i think one of the coolest things and it's i think it's still very early stages and i i think it's it works in sort of a clunky way but the coolest thing is it's got a camera on on the glasses and so whatever you're looking at you can ask the glasses to sort of tell you what it is and i think

that's like uh you know i mean who knows how useful that is at this point but just that technology of like the audio and visual like all sort of being combined in your earpiece I think is really powerful and is going to be you know I agree that's probably going to be the augmented reality that we have because the glasses that have you know that project images onto the real world is just that that technology is like 50 years away so right yeah and especially if you have an open AI Microsoft supercomputer processing all this audio input and visual input put in the background. It can really be cooking with crease.

All right, Reed, tell us a little bit about AI for good and what we can learn about it. You had a story that came out today in Semaphore that talked a little bit about it, and I thought it was really interesting. Oh, you did? Yeah, you know, it's funny. It's a new book that's out, and we're sort of plugging this book, AI for good. But it made me kind of think the chapters in the book are all just like actual applications. They're just like, here's what we did, right?

like what's one wildlife bioacoustics detection social networks of drafts using satellites to monitor whales from space and then some of it's more like solving problems like identifying solar energy locations in india and this is this comes out of by the way AI for good which is which is part it's a division at microsoft that does these sorts of humanitarian things and it's one la vista ferres who who wrote it, who runs that, that organization. But I just liked the fact that it was like, we always talk about these theoretical things when we have a discussion about AI

and all the problems that can come from AI, you know, like existential risk and bias and all this stuff. And then it's like, oh, but there's also some good things with AI, like, you know, maybe AI and drug discovery. I liked that this was like, no, like here's like, you know, 20 different things that we've done and we use AI and kind of demystified and explain it in a way people can understand. I like that. I think we need more of that. It's like we, I think a lot of these discussions we have are so abstract and theoretical and we forget

that like this stuff has been here for a while. It's, it's already transforming industries, transforming the world. And it's, you know, and then, and then you can kind of look at LLM, and all this stuff in context, right, as part of this continuum, the advancement of AI, if that makes sense. Definitely. And I thought it was cool. You wrote that story for a few reasons.

First of all, sort of this type of looking at the good uses of AI decodes some of the good things that AI is doing. Sorry, decodes how the AI is working, right? And that gives us that understanding, and you highlighted that in the story. And the other is that, like, there's this perception that reporters will only write the negative things.

And I think, you know, it doesn't, it's not a dereliction of duty. And in fact, it's something that is incumbent upon us that when there are good uses of this technology, we notice them and write about them and paint a picture that's realistic because this is powerful technology. It's not all about our job's going to go away. It's also about like, who does it help? And that's why I thought it was a great, great piece from you guys. Thanks. Yeah. We try to, we try to be somewhat optimistic sometimes about this technology, you know.

I feel like the AI for bad books have been written. Right. Now, it could do cool things, and I really learned about it a lot as I worked on this profile of Demis Asabas, who runs Google DeepMind, which I just published on big technology. Folks, this was a sprint. I wrote basically 2,400 words or something like that in 24 hours because I just wanted to get it out and have people, you know, see what Demis is all about and where Google is heading. and it was timely. He just got knighted this week in the UK for his work on AI. But basically, I got a chance to speak with Demis and plenty of people who've worked with him and alongside him and still do.

And it is interesting. I mean, his, so he's basically the guy that will either help ensure that Google can survive this moment or be responsible for crashing it, right? I mean, it's obviously going to be a group effort if it goes sideways. But he called this alpha folds, which is their process. protein decoding algorithm, his signature project, something that, you know, you can, that's already helping with, like, for instance, vaccines for malaria, which you've talked about on this show or other types of antibiotics.

His story is so fascinating. He studied neuroscience to get an understanding of how the brain worked. He actually got one of his postdoc professors we spoke with to invest in deep mind. And that guy, he's like, well, I'm only pissed that I didn't put more money in. and then use that understanding of the brain to start working on these programs, these projects like AlphaFold and also AlphaGo. And it was really interesting because it sort of gave us a look as to like why Google is behind because he was so dedicated to this type of machine learning called reinforcement learning

that you basically send a bot into an environment and give it like countless or endless amount of chances to get something wrong until it figures out how to get something right. It's really useful in games. It just doesn't apply to large language. models and which are which is called self which use self supervised learning and basically they discounted self supervised learning within deep mind uh and were behind open AI which was like we're not going to bet on reinforcement learning we have to bet on something else and so that's what they bet on they bet on the large language models which is why they took that lead the thing that's

going to be really interesting is when those two are being brought together and that's when we hear all this talk about agents, right? And an agent is also something that understand its environment, understands its environment, can go and take action. It's much more along the line of that early research that they were working on within deep mind. And if you want to be optimistic about the direction that Google is going and what you rarely hear about, it's that they'll be able to combine that reinforcement learning agent style behavior with large language models and eventually create applications that go out and do stuff for you in the world and help you. And it was interesting hearing Demas talk about it. He goes, yeah, we are, we are highly invested in this about

AIs that can go book you a flight and set you can set goals for them. They can get you a spot at a restaurant. I mean, this is really, it seems like the next evolution opening eye, of course, is working on it. But this, this seems like it can be incredibly useful and helpful if this stuff comes out. Yeah. The irony is, I mean, if the agent thing is really possible, I mean, it totally destroys Google's core business model, which is search advertising, because instead of going to a website and clicking on links to buy things, you'll probably maybe talk, you know, maybe type, maybe talk to an agent that will go out and, you know, probably connect to other agents through APIs, find the products that you're, you know, you're looking for, come back to

you and, you know, there's no room, I think, for advertising in that, in that system. So there has to be kind of like a new a new internet economy that will be created through all these agents and APIs if you really think about it it's it's a profound change that that's about to come and I think we'll have like a big impact on you know the whole digital economy yeah and I think that's why the story of him and deep mind within Google is so fascinating it's like can he push forward painful product changes that the company will need to make to stay relevant in the next in the next stage of the internet and it's like classic i mean we've been talking about this uh you know with google for a while it's classic innovators dilemma uh but ultimately you're gonna have to have

leadership inside the company maybe you already do that will be okay with sacrificing some of um today's profits to remain relevant and not end up being like Microsoft's under balmer yeah yeah i know i think about Microsoft sometimes and how they did sort of i think in the eyes of the tech industry, they kind of collapsed or faded. But actually, they were really profitable. They made a lot of money still because they had all these enterprise customers locked in. I think you wonder about Google, right? Like they don't, that enterprise business is not as core to their technology. So in a way, they're like kind of more vulnerable than Microsoft was, I think, if that makes sense. But on the other hand, they're also really, I mean,

have the best, they just have the best AI researchers. I mean, they have the most talent, I think, of anybody out there. So they can, you know, they can invent the future. I think you're right, it's the innovative dilemma though. Do you want to invent a future that puts your, that puts you out of business, essentially? Yeah. And they're a public company. Like, how are investors going to react to that? If they say, look, we're like, we're pivot. And look at Zuckerberg in the metaverse, right? Like, that didn't exactly please investors. And, you know, who knows? Maybe he was right about that. But it's a big looming question for Google.

And, you know, deep mind is endlessly fascinating. So I'm going to go read your Demis Abbas thing after I get off this. Yeah, it was such a fun story to work on. And as, you know, we were speaking with Jim Landzone about search on Wednesday and as I was working on this story, and I just have found myself continually writing stories about Google. And I'm like, damn, they're just the most interesting company. Not exactly because of like, you know, I mean, they're obviously producing interesting things, but the position they're in in this moment is just so interesting.

And it's like the business news is interesting when you have companies that have to make gutsy decisions. And they have some big, big decisions they're going to have to make and it's going to be sink or swim. And it's funny because like you haven't brought off, brought up the, the whole Gemini fiasco. I mean, just a couple of weeks ago, it seemed, I mean, everybody was saying like, Like Sundar Pachai, the CEO of Google, is going to, you know, have to resign and, you know, the company's flailing. And it sounds like, I mean, the dust has settled a little bit, but it sounds like are we, you know, was that, I think was the reaction to that, like, as far as Google's business, was that a little bit of an overreaction, do you think?

I don't know. It's like one of those things where, like, in the moment, it does feel justified. but as you like, you know, take a look back and you take, you zoom back a little bit, you know, you're like, okay, they're actually doing well in so many areas, but then you zoom back a little bit more and you end up in the position that, you know, you just articulated two minutes ago, which is like damn long term. This is tough. So it's hard for me to really form like, I mean, you know, hard and fast, you know, concrete positions. It's more just like, let's follow along for next little bit and see what happens. I read a Wall Street Journal article last night. I was putting together the newsletter. And it was about, I think it was written in like 2021, maybe, 2020, about Google and Deep Mind. And it was sort of like, it was like, hey, this AI strategy for Google, I don't know. I don't know if it's working out. And like, deep mind's losing a lot of money. And, you know, it just like looking back, it's like, wow, like people really misunderstood what was happening.

Definitely. I mean, I count myself among them. like i thought back because obviously like you write these stories and you see 2015 2016 it's like i think i was a employed tech reporter in the bay area at that time why didn't i pay attention to the fact that elin musk and sam altman were starting open ai or like paid more attention obviously everyone noticed it and i think the view is that it was kind of kooky and then even um watching you know this so getting back into the deep mind story like i watched the alpha go documentary it was incredible like it was so obvious this that this technology was

transformative and then it was like okay like let's go back to like Facebook's advertising issues or something like I don't know we can end I I'll just speak for myself I can end up being so squirrel brain sometimes that it's hard to see like the big story coming through from the trees my my view on AI was that it was it's a technology that it's it's transforming the world like deep learning neural nets totally transforming the world industries are changing it's being employed everywhere, but nobody wants to read about it because it's all happening on the back end. It's all like, you know, as a consumer, you don't notice it, right? Like, who knows

that when, like, nobody knows when they're looking at Facebook that is actually like the algorithm was at one point, like super cutting edge AI. And that was my assessment. And I think when chat GPT came out, it was like, okay, well, I mean, I was totally wrong about that. Like there's, you know, consumer applications that are you know that are really mind-blowing yeah and you know and now you can go back i i i you know i think just reading a lot of stories now you can go back and like write about all the stuff that i think wasn't as interesting to people that is not l lms that is more like just you know deep learning um applied to different industries and those stories are actually more interesting because of chat chabit it's kind of right it's kind of an overton window

kind of dynamic, I think. It now makes some, like the AI for good. Like, you know, a couple years ago, people like, all right, you know, AI for good. And now it's like, yeah, tell me more. Right, right, exactly. And you know what's interesting? Like, I initially, I kind of had the opposite. I had that journey backwards where like meta was actually experimenting with some,

or was Facebook back then, experimenting with some very interesting consumer products, including a chat bot that it said was AI, but obviously a lot of humans in the, in the background. And I was like, oh, this is the future. And, you know, I like, you know, wrote a ton about it. I got to know Jan Lacoon over there and just went real deep into AI for about a year. And then it was just like I came to the same conclusion as you, which was like, I think this stuff is ad tech. Like it works in the background and it serves you the right ad.

But those stories about how like a piece of creative goes through a demand side platform is combined with, you know, something from a data management platform ends up at an ad exchange. and is bought through a supply-side platform would, you know, have readers, you know, slamming their head against the well rather than try to figure out what's going on. But now, obviously, that's all changed. Totally. It's all changed. Everybody has their, their vocabulary on AI has doubled, you know, over the last year. At least that. Yeah, at least that. Well, for tech journalists, it's more than doubled. But, oh, man. Yeah, like, we're trying to figure out what chips are being used in this supercomputer and, like, throwing out. A100s versus H100s and that's the beauty of this job and being able to do a show like this is that we're just constantly learning and you can develop an expertise on things speaking with people in the industry and then share it. I find that yes you can't and that is the fun of journalism.

It's like I love just that I get to learn for a living essentially because that's 90% of the job right is just like calling people and talking to them and this week I've been I went to like a a bunch of different dinners and you know with AI all about really about AI learned so much but I also feel like it's changing week to week like it's too it's so fast it's just it's impossible to really keep up with like all the new the new trends I mean I I was mentioning rag you know like retrieval augmented generation which like I think a lot of people like haven't even heard of it it's like a buzz world word within AI and then I was like I did this like web seminar is rag already dead i'm like i'm like i just can't you cannot keep up with this well it's

that's what happens it's like look at inflection right they were the leaders and then you know all of a sudden it's all commoditized and now the guy is working for for Microsoft okay i definitely can you are we able to do 10 minutes on neuralink before we go yeah let's do it okay when we take a break Reid has been writing about Neurlink and brain computer interfaces, which I find really interesting. We saw some incredible progress there last week. So we're going to talk a little bit about this. It's not just Neurlink, by the way, it's plenty of companies. And we'll talk about what they're doing right after this.

Hey, everyone. Let me tell you about The Hustle Daily Show, a podcast filled with business, tech news, and original stories to keep you in the loop on what's trending. More than 2 million professionals read The Hustle's daily email for its irreverent and informative on business and tech news. Now, they have a daily podcast called The Hustle Daily Show, where their team of writers break down the biggest business headlines in 15 minutes or less and explain why you should care about them. So, search for The Hustle Daily Show and your favorite podcast app,

like the one you're using right now. And we're back here on Big Technology Podcasts with Rita Albergotti. He's the technology editor at Semaphore, which you can find at, what's the URL? Semaphore.com, but it's, you spell it wrong. It's S-E-M-A-F-O-R. So the real spelling is P-H-O-R-E. So just know that we spell it wrong. Okay.

And we'll have that in the description of who you are so people can see the appropriate spelling. You wrote this great story looking at not just NeurLink, but this emergence of this brain computer interface battle. I mean, there was a few things that I saw there that I found interesting. First of all, there's this guy who's just been implanted with a neuralink chip and is like literally playing Mario Kart with his brain with his brain

controlling this this Mario Kart character with the brain and being somewhat competitive too and I mean totally paralyzed and he's able to play it's unbelievable that was interesting and then also you noted that this has happened you know ahead of schedule faster than a lot of people imagined which is not really the case with technology usually right like we do have these like big predictions, then it just takes a little bit longer to get there. Like AI is a technology 50 years in the making or 70 years in the making. But this is really coming on fast. So give us the state of brain computer interfaces. Yeah. When we first launched in October 2022, I was, I met with this the CEO of Synchron, which is a brain computer interface company. And, you know, they were the only

ones with like this, you know, in clinical trials, like human, you know, human studies. And I was like, that's really interesting. Like, like, I had heard of Neurrelink, but Neurlink seemed like very theoretical at the, at the time. And they weren't, you know, there was no talk of like FDA trials or anything. I'm like, this is like actually happening. And they, the way they do it is they don't do brain surgery. They implant sensors like a stent through a, through your vein, through the jugular vein. So it goes right up behind your head to the top of, you know, to the top of your head, but it's not touching your brain in any way. And they got enough fidelity from the brain signals that they had ALS patients who could control a computer, but not in the way, you know,

you were just describing with this nerling patient, a very rudimentary way, but it was enough to send a message. And I asked him, like, can you work, can you do iPhones? And he's like, actually we like just enabled that capability and I'm like can I like how can I do like the first I message from someone's brain and I I sent I messages this guy in Melbourne and he was an ALS patient and it's just amazing like it just this feeling of like you know you're on the cutting edge of this technology and at that time I was like well I think this I think synchron is going to be the you know this is what the technology is going to look like for a while you know and then eventually the neuralink, which is much more invasive, requires, like, you know, brain surgery, like,

actually inserting sensors, like, into the brain. I thought that was, like, you know, farther off. And now they've got this patient in an FDA, you know, study. And it's just to me that was, I was just surprised at, like, how well it worked. And of course, in Elon Musk fashion, like, they didn't, you know, they're not trying to do, like, a minimum viable product. He's going to go big. And they really went big. I mean, they have, I think it's 64 tiny little threads that they've, they've invented a robot to do the surgery because a human hand can't do it. And they like insert these little, they're thinner than a human hair into the brain. And then the chip is, is totally wireless. And, you know, it looks, I mean, we'll see like this is a trial

in a year from now. Maybe it doesn't work anymore. Maybe there's some issue. But it just looks like it could work, which I think is just incredible, honestly. Right. And so just to help us dream here a little bit. Like, where does this, where does this go? If they see the potential, where do you think it leads? I mean, right now, it's like a kind of a read only. So you get the signals coming off the brain and you interpret those signals and you can do really cool things with them. You can control Mario Kart. Maybe you can even, if, If you've lost the ability to speak, maybe it synthesizes your, you know, they've done some studies where it like basically can read like the muscles in your, in your, in your, you know, throat, I guess, that control your voice and then turn the, turn the signals that the brain would be sending to those muscles into your actual voice. So you could talk, like people are going to talk, like lost the ability to speak.

And then I think, you know, control prosthetics, right? Like right now, they're pretty rudimentary in how they work, but you could get, you know, super high fidelity and have people, you know, who were paralyzed walking around. Then I think it becomes two-way communication at some point. And that's when, you know, now you can actually send signals back into the brain and maybe actually change the brain itself or change the human body. I think that's farther off. But I think that's when you get some really profound stuff happening and this kind of like really melding of, of, of, man and machine kind of stuff that, you know, I thought was just totally science fiction, you know, a couple years ago.

Yeah, and it is, it is this thing that's going to obviously be applicable, like we're going through the FDA, it's going to be applicable for people who are paralyzed, like people who need it to function in the world. But maybe at some point, everybody gets this, which is just such a crazy thought. And who knows what happens then? Like is that does that is that you know the black mirror episodes where you like kind of put the thing on your forehead or your temple and you get zapped into virtual reality like at first I was like nah that's you know it's nice way to portray it cinematically we're all going to be in goggles

and now I'm just like oh goodness you know you know I was I actually I don't know you probably have these conversations too but I was meeting with someone who's like you know in this sort of like very like it very deep into like the AI safety and like AI research world and they're just always thinking like 50 years into the future right and the the point this person made was actually like these BCIs and also maybe even like genetic genetic modifications might be how we prevent the AI apocalypse is basically like if you can't beat them join them so we need to like yeah we need to become super intelligent

in order to not get destroyed by a super intelligence I think it's just like it's just total galaxy brain I mean it's just super interesting to think about though I mean it's crazy you know I thought the iPhone was cool I thought like we would we would tell our kids hey you know we were born in a time where there was no iPhone

and that would be like the big thing but like oh my goodness We are going in some interesting ways extremely fast. I hope we just do really cool stuff when we become like genetically engineered super intelligent machines. Like I hope it's not, we don't just use our like brain implants to just sit and like watch Netflix, you know, like in, in the air. Oh, I'm not going to watch them.

We're going to be in Netflix. That's the country thing. Yeah, right, right. Be in Netflix. Yeah, totally. Reed, thanks so much for joining. Great having you here. and I hope when we sign off, they don't fire Sam Altman again.

Me too. God. I'm dreading signing off now. I don't want to see what happens. All right. Thanks so much. You can find Reed's Work at Semaphore.com. Sign up for the tech newsletter. It's a good read. I'm on it all the time, including today right before we jumped on.

And Reed, hope you come back again. Thanks so much. Talk to you soon. Speak to you soon. Thank you, everyone. We're back on Wednesday. I'll tease the guest. We have Marissa Mayer coming in.

We did the Yahoo episode on Wednesday. We speak with the ex-CEO of Yahoo and talking about whether AI can help revitalize consumer products now that she is in the game. And we will see you then on Big Technology Podcasts.