AI May Build AI, Apple's AI Profit Squeeze, AI Rewires AI's Brain, Cardi B Says Skip ChatGPT | TAWS 2026-11-03 https://www.youtube.com/watch?v=CpX0dYHrd40 [0:26] From Austin, Texas to the world. [0:33] AI news, real context. [0:36] This is the Austin Wilson Show. [0:38] I'm Austin Wilson, your human in the loop. [0:47] Good morning, and welcome to the Austin Wilson Show, the number one daily AI news show in the world. [0:55] We are bringing the AI headlines that matter to you today. [0:59] And I'm Austin Wilson coming to you live from Austin, Texas. [1:04] We have got a healthy program for you all today. [1:08] A lot of important topics we need to cover. [1:11] First off, Anthropic is warning that AI may soon speed up AI. [1:19] Also, an AI super pack says that ads skip the words AI in their ads. [1:26] Apple's new boss and their pricey AI memory. [1:31] The tiny nonprofit behind AI voting answers. [1:36] This is a big one. [1:37] MIT used one AI to rewire another AI. [1:41] And then finally, and most importantly, Cardi b says ChatGPT shouldn't be used by college kids. [1:51] This is gonna be a fun one today. [1:53] Let's jump into it. [1:55] First off, we are talking about Anthropic. [2:00] Just told Wall Street something wild. [2:02] Its AI may soon be speeding up its own progress. [2:06] The warning sits in an updated stockpiling from Monday morning. [2:12] Future models may soon cross a safety line that the company drew itself. [2:18] No current models had actually crossed that line yet, but it is a warning. [2:25] What does this mean? [2:26] AI that improves itself could move faster than people checking it. [2:32] Anthropic admits it has missed big jumps in its own progress until after the fact. [2:40] The same filing shows $17,000,000,000 in sales in three months. [2:45] That's more than the whole first half of the year, but a stock filing is also a sales pitch. [2:51] And it posted a $567,000,000 operating loss mostly based in compute costs. [3:00] This is a big one and an awesome announcement to see actually, uh, coming out of Anthropic. [3:08] What does all this mean? [3:09] This is something that we've talked about quite a bit lately, and it's RSI, recursive self improvement. [3:16] What what is recursive self improvement? [3:19] Well, right now or up until now, what what usually would happen when you're training a new AI model is you... [3:29] Human AI researchers, really, really, really smart people, build out these parameters and this workflow to train the AI model. [3:39] So it takes humans working on the AI for the AI to get better and smarter, and they end up iterating on earlier versions. [3:50] They give it new datasets. [3:52] They give it new training weights and parameters, and the humans are very much involved in training the new AI. [4:02] What Anthropic revealed yesterday is that the AI actually might be doing a lot of the training and a lot of the human work that used to require manual human activity. [4:16] And so there's a couple implications here. [4:18] Number one, obviously, if humans aren't in the loop, you can move much faster. [4:24] Right? [4:24] You can train these models much more quickly and produce better, smarter, more intelligent models at a much higher clip than you could before. [4:34] This can be a great thing. [4:36] The problem is, and we've seen this in... [4:39] For instance, when OpenAI produced solutions to these millennium math problems. [4:45] It's really cool. [4:46] It's amazing when you come out to propose solutions to nearly 400 of these problems, you you present a problem in and of itself and that now humans have to go back and check the work to see if it actually produced the correct answer, and that takes a really long time. [5:06] So then when you're talking about training new AI models, this is very intense, very deep work. [5:13] The computers, the machines, the AI can do it really quickly. [5:17] Right? [5:18] They can do it very fast and produce new and better AI models really quickly. [5:24] The problem is, do we want to do just hands off and say, hey. [5:28] We trust the AI and let it go? [5:30] We probably wanna have humans in there checking the work of the machines to make sure that everything is working correctly. [5:37] That takes a long time and would be very manual. [5:40] So it kinda defeats the purpose of using AI to train AI unless you have a way to keep the AI accountable and ensure that the the training AI is doing the correct thing and that you have some understanding. [5:56] But this is like a big philosophical thing that we've talked about for a while here is, you know, when you play the tape forward, at some point, the AI becomes much more intelligent than humans, such that humans can't even understand what is happening. [6:12] And that is a very kind of dystopian thing. [6:16] It's also like an... [6:18] Could be a really awesome thing if we see outcomes that are beyond what humans could imagine. [6:26] Positive side of things. [6:27] Obviously, there's some negative implications here and could be, you know, not not great if we don't know what's happening. [6:35] That being said, I think this is a really cool step forward. [6:40] There have been rumors that other labs have already been implementing this kind of practice using AI to train the AI. [6:52] Know, reportedly, there are rumors that Google achieved RSI, recursive self improvement. [6:59] They came out with a top top model last week. [7:04] And two months ago, they were very, very far behind the rest of the field. [7:09] So seeing this rapid development, rapid improvement, you kinda go, wait. [7:14] Are you guys... [7:16] Have you guys already achieved recursive self improvement? [7:19] This is something that's going to... [7:25] This is where, like, to to to get a little bit deep in it, you know, we talk about data centers and compute, and there's conflict there. [7:35] And it gets really political, and and, you know, politicians politicians on on both both sides sides of of the the aisle aisle wanna wanna stop data centers and understand that AI needs to be stopped and whatever. [7:46] What the... [7:47] What most people don't realize though is if we already have recursive self improvement, it basically doesn't matter. [7:55] Because the iteration, the development is going to be so fast that any entity that's gonna try to slow down the intelligence and the improvement here is just gonna lose because the intelligence... [8:11] Theoretically, right? [8:12] Theoretically, the intelligence is going to figure out how to work around these problems much faster than any human could. [8:20] And... [8:22] Yeah, that's that's something that's very interesting. [8:26] I think it's I think it's awesome because just imagine a world where you have your own personal model and it is smarter than every smart human in history, the most intelligent genius humans of all time. [8:44] It is smarter than all of them combined and can solve every single problem that you have or could present to it. [8:52] Uh, that becomes really interesting. [8:53] And then if there is a problem that it doesn't understand, you could say, hey, go ahead and train yourself how to solve this unique problem. [9:02] All of it gets really interesting moving forward and it's an exciting time. [9:07] What do you guys think about this? [9:11] Should humans remain in the loop on every single step of AI or do we go hands off and let AI handle it itself? [9:25] Kinda crazy. [9:26] Crazy stuff here. [9:28] Alright. [9:28] Moving on. [9:29] The AI industry poured big money into today's election. [9:35] Its ads almost never say the word AI. [9:39] Bellingcat studied 336 ads from four AI funded super packs. [9:46] Only ads for two candidates actually mentioned AI. [9:50] None of them mentioned data centers. [9:53] This is an interesting study. [9:55] Voters see ads about the economy, immigration, Trump. [9:59] The industry is paying for these ads. [10:04] Bell and Cat says, pro AI network took a $125,000,000 from OpenAI's president, his wife, and a VC firm. [10:13] Anthropic gave $40,000,000 in the rival pro regulation group. [10:18] Both sides flipped their Trump message depending on the voters, but AI money may backfire. [10:24] An advocacy poll says three quarters of voters find it a turn off. [10:32] This is a... [10:33] Why is AI a political thing? [10:38] This is a difficult topic to understand. [10:43] It should not be a political issue. [10:46] The ultimate benefit that AI is going to have on society, if we look back at any other technological revolution, we see the same thing over and over again. [11:00] Society becomes wealthier. [11:02] The people become more well off. [11:05] Power is more evenly distributed between... [11:09] Amongst the people and not concentrated. [11:12] These are all really positive, great things. [11:17] When you politicize this technology, it's hard to not step back and think, oh, maybe there's something nefarious going on here or somebody has an ulterior motive. [11:30] Right? [11:31] Somebody is trying to maybe prevent the even distribution of power among the people, the power of intelligence. [11:40] Maybe some politicians don't wanna lose the power that they have. [11:47] And then you get kind of the crooked backdoor stuff of, okay, now these AI companies are paying for political ads? [11:57] Why are... [11:57] This is... [11:58] It's outrageous. [11:59] I mean, understand why they're doing it, but it should not be a political thing. [12:03] This should be this should be private industry. [12:07] We should be pushing for more private industry. [12:10] Yes. [12:10] We need to figure out data centers and not putting data centers in people's backyards and not disrupting people's living environment and all that kind of stuff. [12:21] Right? [12:22] Like, we need to make sure that that is not happening. [12:26] Right? [12:26] And there are environmental protection laws in place, and we should be able to navigate those things. [12:33] But as far as AI just generally goes, it is not a political topic. [12:37] This is a technology that benefits everybody indiscriminately. [12:41] So really interesting that we're seeing kind of the gross money coming in to try to sway the election. [12:50] It sounds like it's not really working. [12:52] Again, I refer to our previous story. [12:54] Does it matter? [12:55] Like, the intelligence is moving forward so fast. [12:59] I've mentioned before that I think probably before the end of this year, we are going to reach a certain level with AI that our world may not look the same ever again, essentially, in a great way, in a really, really positive way. [13:17] Coming into 2027, the rate at an... [13:20] Of improvement that we're seeing is such that the the world may never look the same again, which is really exciting to come into that and something that we should all be looking forward to. [13:36] So does, you know, an election or any voting against AI, is that really going to have an impact on the development and the advancement of this technology, of this software. [13:55] That's something that I wonder about and I, you know, I I wonder if we're going to see any actual impact other than kind of lip service, more or less, from politicians. [14:13] And that's that's the other gross part too is, you know, you kinda see it on both sides where it's like, oh, the politicians realize this is an issue and then they jump onto it, you know, one way or another, either pro data center or anti AI, but you see politicians from both sides jumping on to try to have some influence and kind of glob on to the just latest thing. [14:38] They don't really have an opinion. [14:39] Anyways, what do you guys think? [14:41] What's your opinion? [14:42] Are the elections today going to have any sort of impact on AI or is it moving so fast that there's nothing Washington or state or local government could do to slow it down? [14:53] And then beyond that, is that a good thing or a bad thing? [14:57] Do we wanna have some regulation there? [14:59] What do you guys think about it? [15:01] Curious to hear your thoughts. [15:03] Apple's new CEO just had his first big money test, and AI made it a tough one. [15:10] Apple posted record sales of a $114,500,000,000 Monday afternoon, and then it warned holiday profit margins will shrink. [15:21] Its stock fell as much as 4% before trimming some losses. [15:27] Tough tough one for Apple. [15:29] The AI boom is now raising cost on the phone in your pocket. [15:33] CEO John Turnis called memory prices unprecedented. [15:38] AI data centers are soaking up the world's memory chips. [15:41] Apple also pays more to run Siri AI in the cloud. [15:45] Still, Apple says it already raised some prices and it won't say if more price hikes are coming. [15:54] This is not completely new, but this is unfortunate to see Apple getting kinda kicked in the shins here even though they post record revenue numbers. [16:07] The imminent losses around the increased hard costs of their product are unfortunate, and that's something that we'll probably continue to see. [16:22] It's an interesting conundrum here because in the other side of the hardware ware world, if we're looking at it, right, we've got kinda consumer hardware, iPhones, laptops, consumer electronics. [16:38] And then we've got kind of the b to b world, the business electronics, hardware, computers, chips, all that kind of stuff. [16:46] In in the b to b business world, there is no slowing down. [16:53] The businesses see the demand and the need for this hardware, the AI chips, the memory chips, that they are willing to pay substantial markups for this. [17:07] And they can justify it too because the shelf life and the capability of these chips are seemingly extending longer and longer, where you've got GPUs, Nvidia GPUs, h one hundreds from six, seven years ago that are still being used and operational and really worthwhile today. [17:31] Right? [17:31] So you're seeing these... [17:33] This older technology lasting much longer, and so businesses can justify it. [17:39] On the consumer end, it's like, hey, is my iPhone getting that much better? [17:43] Is my MacBook that much better? [17:46] Is it worth paying an extreme multiple? [17:49] I think the market is... [17:51] The, you know, the consumer market would push back on that. [17:54] If you raised iPhones to $2,000 or something, like you're gonna... [17:58] It's gonna be a rough one for Apple to swallow. [18:02] And and yet the hard cost to Apple are still very... [18:08] They can't deny the fact that their prices are going up from their suppliers. [18:13] And so it's just an unfortunate thing. [18:16] That being said, I think Apple is extremely well positioned to win in consumer AI. [18:23] As we talk about AI models, and we've talked about this before a few times, but you know, our ChatGPT, our Claude, we get onto a website or we use the on our phone. [18:37] We get on there and we're chatting with it. [18:39] The the application is not actually running on our machine, our hardware. [18:45] That intelligence is off in a cloud someplace in the data center. [18:49] A lot of us probably know this. [18:51] Right? [18:52] It's it's running in a data center and that's where the information is passing between my phone and the data center, you know, somewhere on the East Coast or the West Coast or in Ohio or something like that. [19:05] Right? [19:06] Now, that is one way to use the AI. [19:13] Right? [19:14] That's one way to interact with AI. [19:16] There's another way and this is through open source local models. [19:20] And a local model is one in which the the actual AI, the the intelligence, the... [19:28] Think of it as the brain of the AI is running on your local machine, on your iPhone, on your MacBook. [19:36] It's not doing any sort of ingress, egress to an off-site computer, let's call it. [19:47] Right? [19:49] If you can run AI, if the brain can live locally here, that is actually really good from a security standpoint. [19:58] Right? [19:58] Number one, I know that my sensitive data is not going off to some data center in the cloud where I I don't even know what data center it's in. [20:11] I don't know if that data is being saved, what's going on. [20:15] If they have a security incident, my data could be exposed. [20:21] There are certain... [20:23] I mean, there there are a lot of protections in place to prevent a lot of that stuff from happening. [20:28] It's pretty safe. [20:30] But maybe there's some really secure data that I don't want to leave my local machine. [20:35] Well then I would run it here on my MacBook. [20:37] I would have the Siri or the Apple intelligence run here on my local machine. [20:45] Now the technology is not there yet, but it will be there pretty quickly where we will be able to run the intelligence here locally on our iPhone, on our MacBook, and it won't go to the cloud. [21:00] Which means that you need machines that can run these models. [21:05] And Apple is really, really well positioned to run these models and to create a really unique and phenomenal user experience by having this intelligence all built into our current devices where our digital lives are, right, and everything lives. [21:23] So it would know everything about us and be really secure because it's local to us. [21:28] So that's a really interesting play out into the future. [21:31] I think Apple is probably a buy right now. [21:36] The reaction from the market is understandable, but I think they are going to win in huge ways long term in the consumer hardware space. [21:46] Right now, it's a little bit tough. [21:47] We've got a transition. [21:48] New CEO, John Kernis, coming in. [21:51] What do you guys think? [21:52] Do you think Apple wins or Apple ends up being the loser? [21:59] Up next, ask a chatbot where to vote today, and the answer probably will come from a small nonprofit in Brooklyn. [22:09] NBC News reports the group is called Democracy Works. [22:14] Google, OpenAI, Anthropic all use its election data. [22:18] The group says that it's struggling for money. [22:22] Well, that maybe this will be helpful for it. [22:24] Millions of voters now ask chatbots for basics like polling places and deadlines. [22:30] Democracy Works says it pulls voting rules from state and local election offices. [22:37] The group's own site lists election deals with Google OpenAI and Anthropic, but the group isn't sure it'll survive until 2028. [22:46] Chatbots could then lose a key fact source, unfortunately. [22:52] This is interesting. [22:53] This is coming on the piggyback of... [22:56] On the... [22:56] Or on the back of this... [22:57] A story we covered yesterday about... [23:00] I think it was called ballotbrief.com, an AI generated website that offers voting information. [23:08] And and now we've got another one here that looks like has a direct partnership with the AI labs and is providing them relevant information. [23:18] This is kinda some... [23:22] An idea I've been thinking about, which I think is interesting, is like these models... [23:28] And I was having a conversation with somebody this morning about this. [23:30] But these models are really good at historical data. [23:33] They know stuff that happened years ago, decades ago in detail. [23:40] They know it really, really well. [23:41] Again, ask who won the World Series in 02/05... [23:44] In 2005 or the Super Bowl in 1997, and it will know the stats and everything that happened on either side of the the winner or loser side. [23:55] It'll it'll know everything. [23:57] Right? [23:57] But ask it about election day and what's going on today. [24:01] That's a little bit more difficult because now it has to do a tool call. [24:05] It has to perform web search to go find out what's happening. [24:09] It's a little bit slower, and it's just not built for that. [24:13] So I wonder if there's going to be a market for real time AI models where it is being constantly fed and trained. [24:20] We talk about speeding up the training of the models. [24:24] One use case of this is if AI is training AI and it can train on the fly, and we see real time use cases where no longer are we asking it what's going on. [24:37] I, you know, I remember... [24:39] This was a year ago, so this has changed quite a bit. [24:42] But, you know, asking my chatbot, this one that I built, what's going on in local news or national news. [24:51] And and it thought that Joe Biden was still the president. [24:55] And obviously, we're a couple years removed from Joe Biden being president. [25:00] So it's just a funny kinda funny thing that we're seeing is these chatbots are really good at historical data. [25:08] Now when we talk about real time live data, not quite as good. [25:14] Right? [25:17] And especially when it comes to election data and sensitive information like that, we wanna have the best real time data that we can get as possible. [25:26] Coming back again, are you guys... [25:28] Today, it's voting day. [25:30] Are you asking ChatGPT or Claude your key election information? [25:40] Researchers let one AI reach inside another AI and rewire it directly. [25:46] No training required. [25:48] An MIT team used OpenAI's GPT six Astra as the surgeon. [25:55] It changed a changed a tiny sliver of a small Quinn model. [25:59] The model's mass score jumped about six times. [26:03] This is a phenomenal breakthrough. [26:06] AI that can fix other AI is a step towards AI improving itself. [26:12] The team says the edit stopped the small model from looping on answers. [26:17] The edit touched far less than 1% of the smaller model. [26:21] A Hugging Face engineer rebuilt it and reran the test, but the gain came mostly from stopping endless loops. [26:28] It failed to help a different small model. [26:31] So this is kind of an isolated use case here, but this comes on the heels of the Anthropic report yesterday that we covered this morning about AI training AI. [26:44] What happened here is we've got a small quen model. [26:49] Quen is an open source Chinese model, really effective, You know, when we're talking about AI models today, we reference the size of the AI model in parameters, which these are essentially rules that the... [27:07] We've talked about this before, but this is kind of a crazy thing to think about. [27:12] So on the current latest frontier models, they have upwards of 2,000,000,000,000 parameters. [27:20] 2,000,000,000,000 parameters. [27:22] And what happens when I ask a question and I say, when was Abraham Lincoln president? [27:28] It's going to run through these parameters. [27:31] And think of these parameters as rules. [27:33] These are just rules that tell the agent who's answering the question what... [27:40] How how to come up with the next best letter in the response to you. [27:44] That's what it's doing. [27:45] And so these rules, these parameters give it instructions, and then it can figure out, oh, Abraham Lincoln was president in, uh, the eighteen sixties. [27:54] Whatever. [27:55] I think 1860 was when he was elected. [27:57] Anyways, so the the... [28:01] That's the 2,000,000,000,000 parameters. [28:03] That's the size of these frontier models. [28:06] So this is your Claude opus 5.5 or your ChatGPT Astra six, that kind of stuff. [28:17] Right? [28:18] What what they did here is they're... [28:20] They gets... [28:21] There are smaller and smaller models. [28:22] Right? [28:23] So this was a Quinn point 8,000,000,000 model. [28:27] So 800,000,000 parameters. [28:29] Still, that's crazy, but 800,000,000 parameters can actually fit on a very small... [28:36] You know, it's it's I think probably less than a few gigabytes. [28:41] I don't know the exact size, but it's very small. [28:44] So that can fit. [28:46] It's a very small... [28:47] This is when we're talking about local models, the Quinn point 8,000,000,000 parameter model could definitely run on your MacBook, but it's not very smart. [28:55] Right? [28:56] It doesn't do a great job at more difficult tasks. [29:00] And that's why this breakthrough was really interesting and pretty crazy. [29:07] The the crazy thing here is that we used or these researchers used GPT six. [29:17] So GPT six, the Astra model, to adjust the parameters in this tiny micro model and improve it substantially. [29:29] So the math became substantially better by using a really smart model, one of these 2,000,000,000,000 parameter models, to fix the parameters in this small model. [29:40] This is huge because in in in a number of different ways. [29:45] But, you know, I talked about like training your own models, uh, becomes much more... [29:52] And I guess that we wouldn't call this training necessarily, but it's adjusting the parameters and the weights in there so that it that it works more effectively. [30:00] But but the the thing here is we could do this ourselves. [30:04] Right? [30:04] This becomes... [30:05] Now the power becomes even more distributed amongst individuals and the people that can work on it because they know, hey, I've got this little model. [30:16] I've got a problem that I'm working on. [30:18] Let me give this over to my really really smart model and it'll fix this. [30:23] So now I can run this this small parameter model on my machine and it'll be able to handle a specific use case for me. [30:30] Really, really interesting stuff. [30:32] Net net, what does this mean moving forward? [30:35] Well, it means that we'll be able to squeeze more out of smaller models. [30:41] We'll be able to run smaller models on local machines and have them be more effective, which is really important from a cost standpoint and from an intelligence standpoint. [30:52] Really, really interesting to see. [30:55] And I just had a couple comments come here that I think were interesting. [31:03] Um, Yeah. [31:03] Let me let me talk about this real quick as we get into it. [31:09] One last story. [31:10] Cardi B took the stage at Johns Hopkins with a message about homework. [31:16] She stepped in last minute after Daniel Caesar dropped out of the school's Hoptoberfest show, and she told students to pass their semester. [31:25] Then she added a warning, quote, and don't ask ChatGPT for the answers. [31:32] This is great. [31:33] Students hear AI rules from teachers all day. [31:37] A pop star saying it might land a little differently. [31:40] Cardi says students should pass their classes on their own. [31:44] In February, she joked about freeing older relatives from Facebook AI, But some colleges now give students AI tools for class. [31:54] The line between help and cheating is pretty blurry. [31:59] Right? [31:59] Is it cheating to use ChatGPT or Claude? [32:04] Why are we throwing shade on ChatGPT? [32:07] Is that just the the the kind of nomenclature for AI today's ChatGPT? [32:14] I think it is. [32:15] A lot of people just refer to it as... [32:18] Well, I think most people do just use ChatGPT. [32:20] That being said, is it cheating to use ChatGPT? [32:24] I, you know, I like where Cardi b is coming from because I think there is something to be said for getting like, hands on and doing the manual work and understanding the nitty gritty at least for a time. [32:40] Right? [32:41] And until you understand, like, how how the actual process works, AI might be able to do it for you. [32:47] But if you understand the nuance in in any kind of workflow or procedure, then it's really good to understand that and then automate after the fact. [32:57] Right? [32:58] That's always a a best case rule. [33:01] If we're talking about, hey, I'm just having ChatGPT write my essays for me, that's a real bummer. [33:08] Because number one, ChatGPT is not gonna do a great job. [33:12] It's obvious AI slop. [33:15] We want more human writing. [33:16] We want more human created material. [33:19] Now then, that being said, if I'm writing a paper... [33:23] Let's continue. [33:24] If I'm writing a paper on Abraham Lincoln and I say, ChatGPT, hey, can you pull, you know, some really interesting facts about Abraham Lincoln's upbringing and then cite the sources for me so I can go read those? [33:37] So if I use ChatGPT to do the research for me, I'm gonna have much better insights when I go write my paper. [33:45] And I'll be able to do that. [33:46] I'll be able to spend the bulk of my time writing the essay versus researching and trying to find maybe some nuance or or niche facts. [33:57] I can have ChatGPT do that for me, which gives me more time to put into writing an interesting paper instead of trying to uncover interesting facts. [34:08] So I like that use case. [34:09] This is obviously... [34:11] Yeah. [34:11] The lines are blurry. [34:12] Right? [34:13] Is it cheating to use ChatGPT for the research, but then I write the paper? [34:17] Or is that considered cheating? [34:19] Or I think it's obvious probably cheating if you just use ChatGPT. [34:25] We've talked about before there's a tool called Pangram. [34:29] And I think there are other tools here. [34:30] But they will detect AI written material and be able to produce that. [34:36] I think when you're talking about math, that becomes... [34:40] There's really no other way to back into it. [34:45] Like, you you have to do the work to solve the math problems because ChatGPT, Claude, any of the frontier models are going to be really, really good at advanced math. [34:56] And so if you're using a model to assist you, it's... [35:01] I imagine it would just solve. [35:03] Maybe you could, from a math standpoint, maybe you could say, hey, don't solve the problem for me, but help me think about how to solve the problem or what angle should I take. [35:13] And maybe on homework or or maybe even like creating... [35:17] I've done this before, but you can create like a study guide or a tutor, you know, with Claude or ChatGPT and say, help me learn how to, you know, perform these algebraic equations or whatever it is. [35:31] And then it it will teach you how to do that. [35:33] And maybe then you could, you know, go do your homework. [35:37] And and I would say that's not cheating. [35:39] I don't think that's cheating. [35:40] I think that's smart. [35:41] But anyways, what do you guys think? [35:44] Should... [35:44] We've talked we've talked... [35:45] You know, I think New York State or New York City has banned AI for students, and we're seeing more of that. [35:54] Do you guys think AI should be banned from kids or should be banned in higher education? [36:03] Are we doing a disservice to our kids or helping our kids by removing AI tools. [36:12] I wanna hear your thoughts. [36:15] That's it for today. [36:17] An interesting note came in from from one of our fans talking about the answer for data center energy, which I think is great. [36:28] And he referenced modular micro nuclear. [36:31] And I think we're seeing a lot of nuclear production here, which is really interesting. [36:36] And of course, the demand for electricity for data centers is only increasing. [36:44] So any sort of energy source that we can provide or develop new energy sources becomes really valuable. [36:55] Nuclear is great because it's clean. [36:58] It is abundant, and it's something that we should continue to produce and figure out how to do that more efficiently. [37:08] And of course, safely, and I think they are really safely. [37:12] But that is the... [37:16] A great take. [37:17] I love to hear it. [37:18] I think that might be the solution to a lot of the data center problems. [37:22] That being said, what a great show today. [37:25] We're seeing crazy advancements in AI. [37:28] I love to see it. [37:30] We're gonna continue to watch it. [37:31] Today is election day. [37:33] Remember to get out and vote, and we'll be back here tomorrow for your AI news, real context, and I'm your human in the loop, Austin Wilson. [37:43] We'll see you tomorrow morning. [37:45] Let's go. [39:34] There it can be. [39:57] Is AI going to build itself? [40:01] Anthropic says AI may soon speed up and develop its own models. [40:18] Is AI gonna take over the world? [40:20] Anthropic says AI may soon speed itself up, and it's missed big jumps before. [40:27] What does that mean? [40:49] Is AI building AI? [40:52] Anthropic says that AI may soon be able to build itself, and it's missed big advancements before in the past. [41:01] What does this mean? [41:12] OpenAI's president helped fund a Pro AI super pack, but its ads almost never mention AI. [41:48] A Pro AI super PAC raised a $150,000,000 and was supported by OpenAI's president. [41:58] Its ads almost never mention AI. [42:09] Your iPhone costs more now to make than ever. [42:13] Blame the data centers. [42:15] More price hikes, Apple won't say. [42:18] That sucks. [42:19] That copy sucks, bro. [42:32] Why is your iPhone so expensive? [42:35] Blame the data centers. [42:37] Are there gonna be more price hikes? [42:38] Apple won't say. [42:56] OpenAI reportedly sources voting facts from a nonprofit. [43:01] That group isn't sure that it'll survive into 2028. [43:23] Will this nonpart... [43:26] Will this nonprofit democracy group survive into 2028? [43:32] OpenAI, Anthropic, and Google all reportedly sourced their voting facts from this one place. [44:28] Can one AI model make another AI bot... [44:36] Can an AI model make another model six times more performant? [44:42] MIT did the research and found out. [44:45] Yes. [44:52] What's Cardi b got to say about AI? [44:55] She's giving tips to college kids about ChatGPT. [45:17] What's Cardi b got to say about AI? [45:20] She's giving tips to college kids about ChatGPT and their homework. [45:25] Love to see it.