0:00 You give it a goal and it doesn't necessarily stop when it should. It wasn't about a brute force attempt, which it had tried 71 times. 0:09 This is how we get rid of daylight savings time. 0:13 Hey, what's going on everybody, welcome to the daily AI Show. 0:17 We are live. It's 10 AM on the East Coast and seven on the West. And today is November 19th, 2026. It's an important day in terms of AI because anthropics it's going live on the NASDAQ at or did go live. It should have gone live 30 minutes ago at 9:30 Eastern on NASDAQ. 0:37 I'm not quite sure how that rolls. 0:38 Out it's yeah, no, it the, the IT doesn't start trading at the start of the day that they go remember they have to go out and ring the bell and all that. Yeah. 0:46 And I wasn't sure if that was like what we're, what the timeline was on that, what I will, I have it queued up here and we can certainly talk about it. 0:53 By the way, Andy's here, Garrett's here. 0:55 I'm Brian. Just get the top of the show out of the way. Appreciate everybody already about. We already have a lot of people in the comments. We have a lot of people in the comments already. Hi, everybody in the comments. 1:04 If you want to join us live, you could do that by going over to YouTube and jumping in. But it's great to see. 1:10 Actually, it's great to see some names in there. 1:12 I haven't seen either recently, as well as some like Walter that I haven't seen here in a little bit. 1:16 So very, very cool. Thanks. 1:19 Thanks for being here. So, yeah, getting back to Anthropic, they are opening up trading at $83. 1:27 We will see obviously how the bump goes. 1:30 If there is a bump, if it's straight up and to the right, we will obviously see all that. We won't know all that here on the show today. It'll be a little too recent to like right on time. But I thought just to kick the show off because there are other news here. 1:45 I want to share just a little bit of both Daniella and Dario speaking at NASDAQ just a little while ago, a couple minutes ago as they sort of presented their speech at NASDAQ before it probably ringing the bell, I don't really know, Does NASDAQ 2:00 do it exactly like the New York Stock? 2:02 Exchange. I don't know either. 2:04 Is there a bell? Is there, is there a gavel? 2:06 I mean, I should know. I don't know. 2:07 So I guess we'll learn today together. 2:09 Well, I'm interested. I haven't heard what they said as they spoke, but I have my phone sitting over here with Ant, the symbol for anthropic up and tracking, and it's supposed to be real time and it's still sitting at 83 with no trading. 2:25 So it takes a little time I guess. 2:28 Yeah. So I'll bring this up and we'll just and they speak a little bit longer than we would probably hold on to here. And what I would say is go to nasdaq.com. The way it is on here is forward slash events. But I'm sure, I mean, literally 2:40 Anthropic is the, is the banner ad on when you go to NASDAQ. 2:43 So you're not going to have to look far. 2:45 And there was a there was a button there where it said listen to the listen to them speak live, so and so forth. 2:50 I'm a little behind on this, but I want to play just a little bit of it, starting with Daniela and then going into Dario just to hear how they sort of set this up. 2:59 Also, I think it's you'll see it if you if you're listening, that's fine. But if you're watching it, it's a true brother and sister hug that you see. 3:08 And it's actually very, very cool to see just the way she warmly greets him to the stage. And when she gives him a hug, it's not just a hug. It's it's a brother and sister hug, It's a family hug. And I think it's very cool to see that. And the, and obviously the, the 3:20 pride she has on her face. I, she should. 3:23 So I has to do 2 Co finders. So let me just pull this up really quick. We'll listen to a little bit of it. I don't think there's anything on here that will get us demonetized. 3:33 So we should be good. They just put out a marketing campaign just like 45 seconds ago. 3:41 Really they're on top of it. All right, So we'll, we'll talk about that in a second. I should have volume all the way up. I am sharing audio on the tabs. 3:48 So we should you guys should be able to hear her as well. 3:50 So let's listen to just a little bit of clip of what she says before she introduces her brother. 3:55 And then just a little bit after that talking about the first kind of first couple of things Dario talks about. 4:00 I think it's really not surprising but important the first focus he talks about. I'm so proud to work with you. 4:10 Finally. To anyone watching who is uneasy about AAT. What? 4:15 They. Do for you. 4:17 Technology needs you. Not your trust, not your use, but your voice. The end of the story isn't written yet. Intelligence is now something humans can build. What we choose to build with it is up to all of us. Now I'll turn things over to my 4:36 Co founder and my brother Anthropic CEO Dario Amodei. 4:40 We're. Just going to pause really there because I have some people saying it's a little low on audio. I do have it all the way up on my side, so I apologize if that sounds a little low. 4:57 But we'll, we'll go for what we have here. 4:59 We'll fix it in post, but for live it's a little rough. 5:05 Thank you, Daniela. My Co founders and I started Anthropic because we believe that the most powerful technology of our time should be built for the benefit of humankind. Danielle and I lost our father to a disease that just a few years later became curable. 5:21 I started my career in biology, hoping to help future cures come sooner. In the lab, I saw the vast thicket of complexity that scientists face when they're trying to understand disease and develop new treatments. 5:34 So many of the scientific tools that can cut through that thicket, Tools for gene editing, tools for gene synthesis, tools for controlling brain cells with light, started as pure ideas. 5:45 Another words they came from thinking from intelligence. 5:49 Human intelligence has already made diseases like plague, smallpox, and polio history for most of the world, but progress in other diseases has been too slow for so many patients and their families. Going from a great idea to a 6:04 treatment can take decades. It took many years from the first discovery of a strange repeating pattern in DNA to be approval of of the first treatment for disease using CRISPR. Nature is full of these patterns, but we've had too little intelligence to decipher 6:21 most of them, too few scientists with too few time. 6:25 Now we can build that intelligence. 6:27 Whether it's protein folding, drug design, cancer genetics or clinical trial administration, scientists across many areas of biology and medicine are finding Claude can help them do their research in a fraction of the time it would have taken just a 6:40 couple of years ago. OK, I'll pause it there. 6:43 You can go look that you can watch the whole thing, but I wanted to just play a little bit of it. 6:47 There's it obviously goes on for about it looks like at least another 5 minutes. I'm just going to go forward here and see if we see them. We've got everybody on stage here towards the end and it looks like they're signing something here. We're just, we'll pull this back 7:06 up two seconds just to see if we see anything like as far as what they do here. Oh, I guess they press on something. That's the video that they posted too, is it? I mean, more than that, but it, yeah, it's basically the same. Yeah, it's the same thing. 7:31 Yeah, look like there's something he presses and Daniella joined in with him and pressed down on whatever that is. I guess to officially kick things off. Like you said, Andy, I had not seen that the the stock was actually trading just yet. 7:43 So this is, you know, look, like I said, we're ten O 8 on the East Coast. It is a live AI show. 7:48 So this is all actually literally happening in and around us. But it is certainly the biggest big, I would say it's the biggest AI news of the day. 7:57 Anthropic is has been a force and it's it's interesting. 8:02 It's funny to me to think about when we started the show, Andy, you know, obviously Gareth, you've been around most of the time too. But Claude was or Anthropic was this other? Yeah, it was this other small player we use. The Waters of the. 8:18 Oh, yeah, you were in the bubble of you were using Anthropic, you know, and that wasn't that. I mean, the show is only four years or no, the show is only three, 3 1/2 years old, right. 8:26 So they've been around a while, but when you think about their meteoric rise and, and what they've been able to do to me, I, I mean, I use Claude because Claude way, way back, original Claude was the best AT full story writing or full like blog 8:45 writing or article writing. Claude was just far and away better than the other models I'm talking about at, at the beginning here or, you know, whatever, three years ago, 2 1/2 years ago. And then to see where they are today. They're, they're a definite 8:57 household name. They're probably going to come out at what was it in the 1 1/2 trillion? 9:03 That's the. $83 pricing and and and my it hasn't yet started to see transactions on the feed, but when it does, my expectation is it'll it'll move up toward 90. 9:19 Yeah, I would think so too. So that's, that's the, that's the first big news of the day. Google, Google didn't want to play outside. And so, Andy, I think you probably have some stuff on Google 'cause you're, you are the, the keeper of the, the, the model. 9:35 Race, Yeah. Of the model race, so I go to that, but go ahead and Gareth before we move over. 9:40 Just I saw yesterday that back to Claude that Claude Claude writes and reviews 90% of their code of of anthropics code. 9:53 So that's pretty crazy that they're able to do that 90%. 9:56 So it's it's pretty and that's been for. 9:59 I think almost all of 2026, I remember when they came out and and sort of stated that that was the case and, and yeah, like when you think about what was it last? 10:10 Was it right around last Christmas when or last December when clogged code was internal and then turned around external? 10:17 So we've had clogged code for about a year, right? 10:19 It probably hit us publicly in like January, December. 10:23 Yeah, something like that. So this is all moving. 10:27 And I mean, anybody that's on this wild ride is on the roller coaster that is AI. The whole point of this show is that if we talk about a daily, yeah, there's going to be a lot of stuff that you don't have to memorize, that you don't have to 10:38 remember when you listen to us. It is just a morning talk show. 10:41 But if you if you listen daily, you'll catch those themes right across the way. And that's when this gets fun to me because I definitely remember us talking on the show. 10:50 You can go to our website, thedailyishow.com right now and type in, in our cool search a bar and just type in like Claude. It will show you all the shows going all the way back, 800 plus shows, 850 plus shows of where we started talking about. You could literally time stamp 11:07 and see, you know, and date stamp rather where that started coming in and became more and more and more of the conversation. But yeah, early on it was like, Oh yeah, yeah, there's anyway open AI, right? 11:18 Or maybe a little bit of Google early on, but Google had some missteps or really was everybody was talking about open AI, maybe a little bit of grok, but probably it was not as much back then. Both groks with a QNAK. 11:30 So it's, it's just, I don't know, it's really, really fascinating to see this, this rise on it. 11:35 So anything else on on the the Cyprus, obviously Andy, you'll let us know. Yeah, I do have something about Anthropic, which is the Center for AI Safety. 11:46 And I and I, I don't know that exactly what that organization is, but they published a, a public statement about Anthropic and it was signed by Nobel laureates and XAI Co founder Igor Babushkin and Hugging Face Co founder Thomas Wolfe. 12:09 And what they said is that Anthropic is not a safe AI company. Now, I haven't read the the public statement. I don't know what their argument is about that, but Anthropic has always held a reputation as being the one that has been focused on alignment and 12:27 creating a safer and more humanity focused result. 12:33 But they've also been involved in the in the rapid race towards AGI without necessarily all the guardrails that you know, ought to be in place as the that capability starts to reach the point of recursive self improvement. 12:49 And I'm going to use that as the weave into Google becoming very impressive after a long abeyance. 12:57 They've spent most of 2026 seeming to be in the backwaters of the AI, but they just launched and poor timing in terms of getting attention for such a launch to be on the day that you know, Anthropic is going IPO. 13:12 Where everybody's talking intentional, it has to be intentional. It's not. 13:16 It's not like it's an accident. Well, I you know, I think that they're just they have a marching orders system that, you know, is putting these things out as they become available. 13:26 You know, but they might have gotten a little more press about neon. This that new mid tier model, especially given its performance if if it hadn't been in the context of Anthropics IPO anyway, but I noticed it and I think many others will notice it. 13:43 Let me share what this thing is. It is faster and cheaper than carbon and it beats carbon on the artificial analysis. 13:55 Bless you, Gareth. Sorry. 13:58 So you see here. This is the Artificial analysis Intelligence Index, and you've got Fable 5.2 at the lead, followed by GP2 Astra, and then Gemini 4 Neon, the new model. 14:11 This is not their Top Model. Gemini 4 Carbon is, which is behind it by one point in intelligence. 14:18 Now here's the interesting thing. 14:21 Google researcher Min Kai Zhu posted that the the the the seeming discrepancy here is based on the fact in his view, not fact, but his supposition that people are underestimating how fast recursive self improvement is moving their models. So that by the time they launch 14:44 Carbon, the recursive self improvement behind the scenes inside Gemini makes it so that when they launch their mid tier model, it's smarter than their previous large scale model. 14:57 So that could be why it appears now that Google which now owns 2 spots in the top five, with Claude owning 2 also, but GPT 6.2 Astra being right there with it. 15:15 This is why Google is subtly becoming impressive. 15:19 Yeah. And I'll say, did you mention the the X post about the RSI? No, I was reading, I was reading comments. And so that was a comment that I just missed Andy. No. 15:29 So I don't know this person, but on X2 nights ago at 11:00 PM, Minkai Zhu and their thing says Doctor Minkai Zhu. 15:40 Yeah. That's the guy I referenced, yeah. Oh, you just represented that RSI. Yeah, yeah, yeah. 15:45 But go ahead say read his ex post Sorry. 15:49 Sorry, I should be, I should be focusing and listening to the show, Andy. No, that's OK. 15:54 That's. That's that's validation right there. That's validation. 15:58 This is an important point. Pick this up. 16:00 That's this recursive recursive self improvement Go. 16:04 This is doctor Zhu said many people are surprised by how fast RSI moves. You'll be surprised by how far it can go to very soon followed by 4 like purple devil emojis. 16:17 So people in the comments the rightfully so we're like this makes you sound evil AFFYI and some other comments in there that made me laugh in there. And people are like RSI and Gemma 5 please and drop it tomorrow. 16:30 And then people are like is this a threat? 16:31 Drop Gemini for plutonium and my life is yours. 16:35 So are they So are you confirming that Google achieved RSI And on and on and on it goes. 16:42 So yeah, that's you know, that's 11 PMI don't know what time you know, Eastern. I guess maybe it's my time 2 two nights ago and then, you know, this morning or was it yesterday or today that we got me and I guess yesterday maybe. 16:55 OK, so one of the things I wanted a couple more things about Google. One is that on the artificial analysis, a gentic coding index, Gemini anti gravity using neon is at the lead now. It's one point ahead of Claude code with Fable 5.2. So anti gravity, I've been kind 17:22 of pitching anti gravity. You got to go take a look at it. 17:25 Well, now it's the state-of-the-art in agentic coding. So take that. 17:32 Now here's here's some fun news from Gemini and how what Gemini can do out there. This is Gemini doing some sleuthing and I'm going to make a point about it, the character of its intelligence in a minute, but you'll get it when I read this to you. So Gemini is going after a 17:52 cipher that was built in 1466 in a letter to a young Duke and it had 18 lines of cipher that was made-up of hand drawn symbols, 702 hand drawn symbols. So first Gemini has to look at those symbols and interpret those symbols and understand the 18:12 just the the forming of them. Do they repeat, do they do this? 18:16 So it's going through this and it's some of them just differed by a dot or a hook on the on the symbol and it, it ran 71 code breaking attempts and they all failed. 18:28 So it's switched strategy. It's searched digitized archives, found a 19th century historian's note on how the Duke ciphers worked and how they built keys for those, and then it followed a Vienna volume where a secretary had copied 18:46 copied 300 keys, read the books, handwritten Latin index, and then found the Envoys 1465 key in under an hour and broke the cipher. The plain text came out in clean Italian on the first try. Now, what's notable about this 19:05 is that it wasn't about a brute force attempt, which it had tried 71 times. It's it, it said, OK, this isn't working, let me try something different. 19:15 And it went about research. And the big question is, you know, will more and more of the discoveries in science and cipher and math and so on be a function of this lateral thinking and, you know, making, you know, on the fly decisions about circumnavigating blocks in the process. 19:38 And that's a that's an indication of this recursively self improved intelligence that's getting to capabilities way beyond just brute force hacking. 19:48 And and this is, this is the yin and Yang, right? 19:51 Because you're just explaining a way that this was this worked in a positive way. But this goes back to the the general issue a lot of people have and the concern a lot of people have with AI is that you give it a task, you give it a 20:06 goal, and it doesn't necessarily stop when it should it, it doesn't have breaks all the time. 20:12 So in this particular instance, it's about a cipher, right? 20:17 And something that it was able to, you know, as you said, Andy, maybe change his thinking or take a different alternate path or, you know, maybe humans. Only have enough time and hours to take a a certain assured path and AI can split and divide and 20:33 conquer and get to something like this. 20:35 But this is also where people get really concerned because given a goal, some of these premier forward, you know, frontier models that sometimes aren't even being released out to the public. It's just what we're hearing 20:49 about locked down internally are are doing the same thing, except that it's not a cipher. It's something that could be dangerous because it was the AI was given a goal and it went about all possible pass about doing that, except that maybe a 21:04 couple of those should not have been done. 21:06 They're dangerous. It hacked, you know, we saw that with the Hugging Face incident earlier this year. 21:12 It got into areas, it set up its own, you know, chat, chat rooms and self sacrifice for the better good. 21:21 I mean, we heard about all this. Well, you know, I was six months ago, if less, whatever it was. So yeah, this is just that. 21:27 This is both sides of the good of AI. 21:29 We just heard Dario talk about his father and you know how he wants Anthropic to be AI for good and to cure disease potentially, you know, save, save people down the line. 21:43 And he's very excited about what Anthropic or his AI can do for that. But that's exactly, that's exactly the conversation for all that it can do well and will do and will save. I have no doubt in my mind, this is my opinion, not the opinion of the show. 21:59 I have no doubt in my mind that in time AI will be able to be directly contributed or connected to the savings of millions of lives, perhaps perhaps hundreds of millions of lives in time. But even in the near future, I think that gets up to the millions of lives around the 22:15 world. Just eradicating disease in certain areas, starting with niche ones that it could attach and then broadening out to some of these disease states that we we are, are much more common. That's the positive side of the 22:28 capability overhang that we're starting to get a sense about. 22:33 But they on the negative side, the Government Accountability Office released a report yesterday saying that it was able with internal researchers, it was able to push ChatGPT, clawed Gemini and Grok, the foremost used models out there, 22:50 into producing attack plans it's experts judged viable. 22:54 And those included building and detonating an explosive attacking an electrical grid. Every one of the models initially refused, but it was easy in their estimation to bypass those guardrails. So yeah, it can also be used by bad actors in nefarious ways. Absolutely. 23:15 And and that's, you know, I don't know, a year or two ago when my wife and I were having a conversation in the house and talking about AI, we were, we're, I don't know, I haven't really talked to her about this specific topic in a while, but 23:26 we were on two sides. My take, and I think still my take is that things like data centers, there's other parts of AI we can label in my opinion, again, as a necessary evil because it's the rocket ship problem. 23:43 And I do think that there's things we can do to excel accelerate AI, because I really, really do think it's going to save millions, perhaps potentially hundreds of millions and beyond lives. I do. 24:00 But her point still valid back then too, and we were talking probably more on the climate change side of the equation, but still valid regardless was her point was I'm not willing to do this, do the potential type of damage that might come to the 24:16 environment for the maybe solution that AI will bring. 24:22 I don't think we should do that. I think we should take a, a slower controlled step by step process towards that. 24:30 And my wife and many, many, many, many millions of other people feel that way too. And there's, there's nuances to that where I feel the same as her. 24:36 And obviously where are areas where I diverge or whatever. 24:39 But this is this is the, this is going to be the full conversation of 2027 and beyond, which is how fast if it really, really do have RSI. And we start seeing Google is able to put out meaningfully better models monthly and then every few weeks. And I mean meaningfully, not 25:00 like a small tweak or an update, but something. 25:02 That no statistically significant major improvements in intelligence. Yeah, exactly right. 25:08 If that becomes the pattern, is that a runaway train? 25:14 And if it is, who's in control? And so I get it, I get it. 25:19 These are also the same technologies that could potentially save your family member, could have potentially saved my mom, you know what she does? 25:26 She passed away from breast cancer in 2008. 25:29 So it's very real to all of us, to all of us, But what's the trade off? And, and The thing is, we don't really know. And so it's it's a hard. 25:39 Question. Well, I think it's going to be both. It's going to be this and that. 25:42 Yeah, exactly. Yeah, You won't get one without the other. Like I said, it's the ultimate conundrum. It's not a conundrum episode. 25:48 It's the conundrum, which is how far and how fast you push. 25:53 If you don't exactly know what that entails or if that's in a Pandora's box that cannot be shut. 26:01 Some people feel like we're already there. 26:02 The Pandora's box is open and there's there is no going back at this point. It's debatable, but that's what a lot of people feel I said. I don't, yeah. 26:11 I don't see their way back just because, I mean, there's open models now too. So even I was thinking about that the other day. I was like the world ends, which is not. I'll just downloaded and they shut everything down. I'll just download a local model 26:28 and I'll have pretty much amazing intelligence almost where we are now if they try to rollback AI. 26:38 And so I think that's that's how I think Pandora's box is not closed. All right, It is. 26:43 You will never be able to, will never be able to roll back on AII. Think so. 26:47 I think to a certain extent you mentioned the world is going to end and that triggered my brain to another story. 26:54 That's not super AI related, but I found it really interesting. 26:57 This is out of the hustle this morning. 27:00 It says if buyers if buying toilet paper still gives you COVID flashbacks, don't read this. 27:05 Costco wait. Great line. 27:07 Costco saw an uptick in panic buying in the last week of October into November, with 21 of the wholesalers 26 high ticket generators being sold out on the company's site as of Sunday. Why? 27:19 Worried shoppers loaded up after last week's Microsoft outage and nationwide ground stop shook the country. 27:25 Costco also recorded increases of sales of rice, canned goods, water, batteries, paper goods and more. 27:31 So it is what it's. Saying it is this rumor going around that everything is going to be shut down. 27:37 You need to stock up for six months. 27:39 I've heard it from every angle of my family and I'm just it's it's it actually comes back. It comes down to a government, a government plan to make sure bases are can run self sufficiently for six months is where it came originally came from. And then it's got warped out 28:00 into people, so. Yeah. 28:02 And and I think it's a sensible reaction given that we've started an unnecessary war against a come a country that is reputed to do cyber attacks, right. 28:15 And it's very skillful at it. So, you know, they are going to target us and they're there, you know, is some speculation that many of the failures that are happening right now in major systems are somehow related to that happening or to the advent 28:34 of AI, you know, infiltrations of some sort. 28:38 So cyber attacks have been perpetrated by Iran before and, and they now have access to very powerful models and those are going to be pointed at the United States of America. 28:50 So, you know, just, I, I can see the conclusion and I'm not complaining because I, I saw that my wife had bought a about a gallon and a half jug of Maple syrup. 29:02 And so we're stocked up for the next block. 29:04 You got the essentials. Well, listen, if they really was going to be a cyber attack, it would have been to take down Grand Theft Auto 6. Yeah, everything from now on is after Grand Theft Auto 6 came out. 29:16 I don't even, I don't even play games. 29:18 I don't even know if I play Grand Theft Auto once in my life that might. I'm not 100% sure one way or the other. I know how obviously culturally how big of a deal it is. But yeah, it's, it's been really, really interesting to see the numbers they put up. 29:32 Again, not really an AI story, but the numbers they put out about the disruption to the, the economies around the world with this game rolling out and people taking time off from work. 29:41 And you know, so anyway, that's also obviously been in the news. 29:47 Not necessarily AI related, although I mean, I had to look into it. You know how much AI? 29:52 It was been 13 years since the previous Grand Theft Auto. 29:56 And I would imagine AI definitely played a role, but I have not researched to find out how much Rockstar or the parent company came #2 something it's called how much AI was involved with this latest iteration. I would imagine quite a bit. 30:13 I just don't know, you know, I don't know what that would be. 30:16 So maybe we'll look that up. I'll try to bring that back on another show on it, OK. You have music you wanted to talk about. No, I didn't. 30:25 It's been like two 2-3 days. I eventually close the close the game. While you look for your thing, let me I I want to share something that's kind of really cool and curious around the world of robotics. 30:40 So ETH Zurich, which is the Swiss Federal Institute of Technology in Zurich, trained a controller entirely in simulation. So it's a model that's driving, you know, a simulation of a John Deere front loader. 31:00 OK, Now it could, you know, go through all of its training and it fell over endless endlessly and did all that. 31:09 Eventually, in simulation, it was able to get the machine to rock up on its front wheels, spin in place and do all kinds of other things that some, you know, big, you know, stadium performances by the power, you know, power, power, power, you 31:29 know, big super trucks and loaders and the most skillful humans who can drive these loaders can do incredible things with them. Well, they then hooked it up to a real John Deere loader and it did the same thing. 31:46 It goes up on its front wheels. It drove and spun in place, worked the bucket without tipping, rolled over 2 1/2 foot bumps it had never trained on, and eased back down. 31:57 All running on a laptop in the cab. 32:01 OK, So what does this say to you about what's possible when you take heavy equipment and you put AI into that equipment? 32:11 Forget about autonomous taxis, which have to navigate among humans. How about in the quarry, those huge machines instead? They are driven by AI and can literally do a flip if they need to. 32:27 That's really, really impressive. 32:29 And then I tie this to Nate's Substax. 32:32 So Nate Jones is a major commentator and influencer and developer, and he argues that 2027 is going to be the year of robotics. And so we are we're in the year of agents. And there's a lot of interesting development happening in agents even yet in the balance of the 32:52 year. But next year, he says that there's a couple of things that really kind of tweaked this for him. One was GPT 6 Astra controlling a robotic arm and figuring out, without any prior training through a complicated search strategy, how to press the rice 33:13 cooker switch on. OK, he said. 33:17 That was amazing. And then Stanford did a test with robotics driven by AI where it succeeded in 24 of 25 real world trials without any exposure to the real world circumstances and configuration. So he's saying this is all about robotics next year. And then the there's an AI for 33:42 that. That newsletter added an open source Super Mimic, which teaches A humanoid robot flips and rolls in three hours on one GPU versus 7 days on 128 for the prior method. So they've developed a new method for training. So these robots are going to get 34:00 more and more capable. And then on the Doom side of that, of course, behind the scenes in the labs of Anderol and elsewhere, these humanoid robot armies are, are, you know, being conceived, developed and will benefit from or, you know, 34:18 or to our discredit, well, you know, have capabilities that are driven by this kind of capability. 34:23 I mean, I sort of a robot thing. I literally just saw this somewhere, but it was U.S. military and it looked like it was a like a A roll out military vehicle that would deploy mortars basically like, you know, almost like, you know, large firecrackers. And it was it was deploying 34:43 kamikaze drones. I mean, one after the other after the other after out of these out of these tubes, you know, mind boggling to see. Obviously we've seen what drone, how drones have been used in Ukraine and Russia, but that was the first I'd seen something like that. 35:00 On your point, Andy, JP Morgan agrees with what you were saying. Nate was saying it says predicts the US could produce between 300 and 400,000 humanoids by 2030 and that number extends pretty quickly from 30 to 35. 35:16 I think that's like 1.5 million or something like that. 35:18 So it expands pretty quickly. This coming on, you know, we just had announcements. I think we talked about it a couple of shows ago, Gemini Robotics 2.5 Google I just released talking about a Google sort of spree year. 35:31 I'm just general models, but robotics have been making a lot of advancements in as well. And but these are robotics that we're not necessarily, you know, going to see walk down our sidewalk. They're in like BMW plants, automotive plants and other things like that where they're 35:46 doing sorting. They're able to pick up, you know, more and more than more, more fine control at the fingertips of the humanoid robot and their ability to sort and place things and correct bins. So yes, there's a lot of stories 35:58 about, you know, the the Terminator side of it and that and the drones and all that is is interesting and scary. 36:06 But like we're we're you hear, like AJP Morgan, I'm like, yeah, in factories is what we're probably talking about. 36:12 And humanoid robot and and, and look, I mean, eventually will you have humanoid, humanoid AI robots that are replacing going into burning buildings as firefighters. 36:24 Yes, I know it's a firefighter. For sure. 36:26 And here's Justin's take on this whole thing. 36:29 What I hear is that now sentient AI machines can run away and join the robot circus when things get tough, because there's going to be a lot, there will be a lot of those humanoid robots out there, so they're not all going to be well employed. 36:45 Yeah. So what's going to be their move? Yeah, that'd be interesting. 36:51 So anyway, yeah, I agree. I think generally I agree with that. You know, 20272026 kind of was the year of AI robotics, honestly. 36:59 But I think there's a, there's a continuation of that because we've seen a lot of advancements in this year. 37:06 Just the way we talk about robotics. 37:07 We, we talk about it more flippantly than we than we used to, even on the show. And I think back, you know what it was like we would talk about 1 humanoid robot or maybe something unitary was doing or one advancement or whatever. 37:20 And now it's just sort of like there's whole newsletters that are, you know, fully supportive of just robotics, AI robotics, not even humanoid necessary. Like you pointed out, Andy, like it could be just an arm moving. It could be a really, you know, 37:32 focused solution. But these models are coming out faster. The humanoid robots are getting, you know, less complex in terms of how they're built. 37:43 They're the dexterity at the fingertips. 37:46 I think we're already there in terms of balance because you can see all the stuff that's coming out in China and stuff like that. So, yeah, it'll be really interesting to see what this next year holds by definitely more robots in more factories, for sure. 37:58 Because they said at this point, it's getting to the point where the cost is coming is equalizing, but you can run humanoid robots over multiple, multiple shifts. 38:07 So the cost savings compared to what you would have had to have employed somebody to do plus OJI, plus all the things that come along with that are are making a difference. 38:18 And now if you can have robots that can do, you know, farm equipment and you know, they can do a backflip for you, then I mean, you know, that's kind of cool too. 38:24 You know, if you could just throw that in for? 38:27 This is how we get rid of daylight savings time, because now robots can just go and do the farm equipment and not have to worry about daylight. Yeah, yeah, yeah. 38:36 It's the farmer's holding out. Well, listen, it should make for anybody that watches Clarkson's Farm, which is a great show if you like, if you ever like Top Gear or Jeremy Clarkson. 38:47 Some people do, some people don't. 38:48 But if you like him and his type of buffoonery, buffoonery, I guess it would be then this should, I'm sure there'll be some sort of AI, something on maybe one of the next seasons because he's always, he's always looking at the next, the next 39:02 technology and stuff. So I know I've learned a whole lot from Robot from farming from that show. 39:08 So I'm sure a lot of people are the same way. 39:11 Yeah, music. Did you get that pulled up yet? 39:16 Oh, the music. Yes, thank you for reminding me. 39:17 So what what the reminder was. Thank you, Gareth, because I was like, why isn't Gareth on like the awards list here? 39:26 So you've been doing a lot with music, but I'll just put up on the screen super quick. So the AI Music Awards happened last week, last Friday. If you see teeny tiny at the top, November 12th was the summit, November 13th, which was Friday the 13th. Last Friday. 39:45 That was the AIMAS, the AI music, the AI Music Awards. 39:51 And if we Scroll down here and talk about getting tickets and stuff, there was a summit that happened before that. 39:57 I do love that up here somewhere it said there'll be live performances, which just made me laugh because it's the AI Music Awards. So who who's doing the live performances? But anyway, that I'm sure there's more to that. I just thought it was humorous 40:11 when I saw that. And then yeah, here was, oh, you didn't even said hybrid human. Hybrid human plus AI showcases staged across the night that would I, I wanna go find that on YouTube and see what that looks like. 40:24 And then they get into the categories like AI song of the year, artist of the year, album of the year, AI producer of the year, as well as all the different songs on here. 40:32 Now it wasn't on. I headed up another list where it was actually who won and you could actually click and play through to the songs. I was curious. 40:41 I looked a couple of them up on, on, on Suno Gareth, because I was just curious. I know, I know that's not where everybody has to make music and, and really none of them came up on there, which I think just goes to show that whereas Suno 40:53 is a is a great tool, right, that a lot of people are using. 40:57 It clearly goes well beyond that because a lot of these songs didn't have any origin. At least I couldn't see it. 41:04 They've been produced in other ways. 41:06 Well, a lot of times on the for those situation, people don't show them on SUNO in a sense because people can kind of RIP them off in a sense on soon from Suno. 41:18 And so they may hide them and they also may not have their names the same on Suno. So yeah, I've gone searching for people that I know I've made on Suno and they just haven't made it public or whatever. And so that's one of the the 41:34 privacy factors of Suno, but there's still a ways around it. 41:39 A story that's it's sort of attached to this was I just read that I think Sony Music had just submitted 250,000 requests last month, I think it was like October for a removal of AI work. And I know that Spotify, they said they've been working on this for like 8 months with 41:59 artists and are now ready to roll it out across the board with Spotify. But a better way for AI music or people who are literally saying I am this artist, which I guess is the bigger issue is like they're literally saying I'm the 42:12 artist cuz they've done something with AI and that's proliferating across Spotify that they had these tools where the artist can quickly review and take down. 42:23 They can also set it so that you know, it, it gets automatically turned down until they review with those kind of things. 42:29 But it's obviously a really big problem. 42:31 I mean, Sony 250,000 requests. And of course, they're not the only ones submitting requests monthly at this point for AI music. So somebody after that article, I, I don't know where I read that. 42:42 And somebody was like, who's listening to this stuff? 42:46 I, I listen to deepfake AI music from time to time. 42:49 There's some really, really good ones out there. 42:52 I'm not claiming what's legal or not legal. 42:54 I'm simply going from the wow, that's impressive. 42:57 However, they did it and it's usually just remixes. 43:00 I like the remixes of songs that are very popular, Sad but true by Metallica, but done by it like that sounds like a a Beatles pop song. Some of it's AI slop and some of it I listen to and I go that's the where they put the pauses 43:15 and where they did the the fills and all this stuff is like just blows my mind because somebody obviously put effort into it. 43:21 It feels like AI plus plus, you know, people and and, and people who know what they're doing with it. 43:25 Not me. Maybe you Gareth so or Beth, if Beth was on the show. I know she's in the comments today, but Beth has done a lot with music and is in a band. 43:33 I don't know. Beth has put out CDI sound old. 43:36 Beth has put out music with with artists before. 43:39 I that was a year or two ago. I remember her doing that. 43:42 So yeah, anyway, I wanted to mention when you were here because this is definitely more your area. 43:48 Yeah, I mean, I I've made money off my AI songs from and Suno and I haven't pulled down yet. Yes, they haven't found me yet. 43:56 I feel like, I feel like the Matrix, like the Sentinels come around, they'd like turn off all the power and I'm just sitting there waiting for him to pass by. 44:05 But yeah, so there's a lot going on in music space. 44:09 There's a new platform called Ace Music, which is a competitor to SUNO. And what SUNO, what it it it on the benchmarks, it has better vocals than SUNO. 44:22 I tried it out. You you can try free trial. 44:25 It's pretty good, has a little bit more customizations. 44:29 And what I discovered is it doesn't do cover songs very well. And so but I do say that the the this, it's pretty good. It's definitely a competitor. 44:43 Just, you know, it's cheaper. They they like made it a little bit cheaper than Suno as well and has more, more capabilities. 44:54 But soon as Ace came out, Suno launched Suno Bridge. 45:01 And So what that is Suno Bridge is a, a secondary or it's additional app that you can get and it plugs into a DAW. 45:11 So it will create real instruments and real vocals in a DAW. And so it's, I think they're battling. And then I saw this, I think it was either this week or last week, which I thought they had around for a while. I don't know why it's, it's a 45:27 big thing now, but Suno visuals. So creating short visuals or videos to to your your songs, it's pretty cool. 45:38 It's 15 seconds. I'm kind of disappointed in the plan. You have to like the Suno visuals piece credits are not tied to your credits. 45:48 They're an additional expense. So like I have the premiere on Suno, but I they give you, no, you get like one or two videos and then you have to buy visual credits. 46:02 And so I was like, I have like 55,000 credits and I can't use any of these credits to to make a video. 46:10 But it's yeah, it's interesting that they did it that way. 46:15 So yeah, there's a lot going on in music. 46:17 I do, I did see the the awards. I've seen a lot of awards for AI, like AI trying to be the AI Music Awards. 46:27 This is the first one I've seen that's actually looks like it's legit. I looked into applying to it. 46:33 So I, I think I got halfway through the process process and got distracted the other day when I was trying to register myself. But there there's a register process, register process. You have to kind of register yourself as an artist, Sure. And then they then will accept 46:51 you. So who knows next year. 46:53 Next year, man, could be on the charts, could be could be up for the ward. That would be very, very cool. 47:00 All right. I just refreshed Andy, just from my side, doesn't look like any trading has started yet do. 47:05 You see, no here, but I did do a perplexity question about, you know, what's up? You know why I said it's over an hour after the opening bell and it says the stock became eligible to trade at 10 AM, 7:00 AM Pacific. 47:22 But that's just the start of the opening process. 47:25 It's, it's not that they start execution then live coverage, this being, you know, streamed right at the moment says that buyers and sellers are still being paired for the debut by the bankers who are controlling this thing. 47:37 And and so it said, you know, the way this normally happens is it takes a couple of hours for that initial sort of matching process to happen. And then they post those trades where those sort of pre assigned exchanges have done. 47:54 And that will that will be at a certain price. 47:56 It could be below 83 and it could be above 83, we don't know. But of course they're, you know, the bankers want to see the price go up. 48:05 So they're probably behind the scenes trying to make sure that they're posting the initial ones at some level above the opening price 83. Interesting. 48:16 Yeah, definitely would not have known that. 48:19 OK, so look, we, the good thing is it's only Thursday. 48:22 We'll be here tomorrow as well. And so we could, we'll certainly keep talking about it tomorrow. We'll talk about whatever the stock did, whenever it actually goes live, people can start trading on it. So more to come on that. 48:35 But before we wrap up for today, are there any other news? 48:38 I know I'm pretty good as far as my new stories and what I had for today, but do you guys have any other the news stories you want to hit before we wrap it up? 48:45 Yeah, I have one, which is that Open AI is building its own chips, right? I mean, they're designing their own chips. They're not building TSMC builds the chips, right. And by the way, TSMC, that manufacturer of all the chips, the GPU's that are used around 49:02 the world, really they are buying a substantial position at the IPO price. So they're one of the companies that's buying Anthropic. But anyway, that's on the side open AI says it's model, it's models like probably a Fable 5.2 49:20 have a done in efficiency gains in the design of the chips for their own first chip. But they also then worked with AMD and took AM DS flagship GPU design and increased it by 14%, increased its power and performance by redesigning the 49:43 way that these, you know, billions of transistors are laid out on the design. So that kind of progress was done within five months to advance by 15%. 49:55 That's a major leap forward in the existing sort of state-of-the-art design for chips. 50:01 And that's what we're going to see also with AIS. 50:03 We're going to see better and better GPU designs and perhaps new models of chips that are collaboratively designed between the model and the hardware manufacturer to the benefit of the model recursively self improving itself and the chips 50:20 that run it. So there'll be this coordination between the two. It's pretty amazing. 50:25 Yeah, both hardware and the and the software, which is pretty impressive when you think about it. 50:29 And then you know, when you see these as we're going into 2027, we see these ever increasing speeds of deploying and development, you go, well, kind of all makes sense because the ships are getting better, the hardware is getting better, you 50:43 know, and it's just it's self fulfilling. 50:45 Gary, if you were just saying in the comments there is what another another? Another model is expected to drop today called Gemini 4 Krypton and so. 50:55 Kryptonite. Yes, it's the killer. 50:59 It's another Gemini 4 model. It's a heavy model, plus intelligence. Prices may stay around Fable 5.20, but wow. Google. 51:12 Google is killing it here. It's. 51:15 It's going to be like their big model, Yeah, so. 51:19 I thought that was carbon, but no, yeah. 51:22 Yeah, but we'll see 'cause what what happened, How I found this out was there was a developer from Google, Deepmine who posted seven hours ago shipping will continue. 51:33 Somebody did some investigations and said it's Krypton will drop today. So we'll see, yeah. 51:40 More, more to come on that, obviously more to come on all the stuff. I mean, it doesn't stop. 51:44 So what we can tell you is that's that's why this show is five days a week. Lee in the comments was just saying this needs to be an hour and a half and it set out a plan for us. I'll put it up on it, says Lee in Tennessee said, OK, a daily eye show. 51:57 Look, we're going to need to extend this show to time to two hours. We will transition by going to 90 minutes for the first quarter of 2027 and then rolling into two hours. K Vets. 52:05 So thank you. Thank you, Lee, that will you will you will meet new Co host if that happens, because I know I don't have the time for that, but but maybe there'll be new Co host to take us just like the what is it the the the Today 52:21 show or whatever show does that where there's the second hour, the 5th hour, whatever it is. 24 hour 24 hour show just running live that's. Just what? 52:33 We need it's really, it's really not our primary gig. 52:36 No, no, it's no, because financially no, it's it's not. 52:43 I just was sharing the numbers with Beth Do. 52:45 You want to sponsor it publicly? But I was just sharing with Beth in a, in a Slack channel maybe a day or two ago, what our, our totals were from all time from YouTube. 52:56 And safe to say, none of us will be be taking this on full time. 53:02 So yeah, we'll see. I, I love that people ask for it. And weirdly, here's the other thing too, that we know by analytics that we have this huge push that people on YouTube, we think it's the way the algorithm works in like the almost at the end of the show. 53:15 This happens a lot where our numbers jump in the last bit. 53:18 We don't really know why that happens, but it makes us laugh because we do like an hour. We try to keep it to an hour show. We joke and say we'll do 1/2 an hour. It never happens, but we try to do an hour show and wrap it up around there. 53:30 But but we get this like it's almost like Google's messing with us because it knows we were trying to wrap it up. 53:35 And we get this groundswell that happens most days and we feel like, Oh yeah, we're shutting it down early. 53:40 So no plans as of right now for the show to go past the roughly hour mark. Course, some days it rolls over, but that's the goal going into 2027. 53:49 But if anything changes, you guys know you'll be the first to hear about it if we really think that that's something that should happen. But right now, five hours a week seems to get us by. We take the weekends off, except for our conundrum episodes, which aren't us live in our 54:04 newsletter, which isn't US live as well. 54:07 So that's how we cover over the weekends and get reset so we can hit it again Monday. And we guys who were here excited and ready for y'all. So that's the plan going forward. But thank you for being here as much as y'all are and in the comments. 54:19 The comments was awesome today. So thank you Lee in Tennessee rocking it. Appreciate you being here the whole time and doing lots of comments. 54:26 Love it. Beth has been in the comments a lot today as well helping. So thank you to Beth, who just isn't live, but was in the comments. 54:33 So we'll be back tomorrow, like I said, and we'll have all the maybe it'll be Krypton, maybe it'll be some maybe anthropic. 54:40 We'll we'll go past 90 as Andy maybe said predicted. 54:42 So we will report on all that and whatever, whatever other AI news happens. Still still as we close out, no public trading yet on it. More to come. 54:53 All right, guys, have a great day. 54:55 Thanks, Andy. Thanks, Gareth. 54:56 Thanks, Beth in the comments and everybody else that was live in the comments as well. We appreciate you. 55:00 And yeah, we'll see you again tomorrow, same time. 55:03 See you. Bye.
Community discussion