47 Min Read

[The AI Show Episode 182]: Gemini 3, Nano Banana Pro, GPT-5.1 Pro, Nvidia Earnings, Karen Hao Book Controversy & Entry-Level Unemployment

By Claire Prudhomme on November 25, 2025

Podcasts
Featured Image

Get access to all AI courses with an AI Mastery Membership.

Learn More

Google released Gemini 3, its most advanced model yet, along with Nano Banana Pro, a leading image generation and editing model.

On this week's episode, Paul and Mike go deeper on those topics and other top news this week, including GPT-5.1 Pro, controversy around a popular AI book, Nvidia earnings, and more.

Listen or watch below—and see below for show notes and the transcript.

This Week's AI Pulse

Each week on The Artificial Intelligence Show with Paul Roetzer and Mike Kaput, we ask our audience questions about the hottest topics in AI via our weekly AI Pulse, a survey consisting of just a few questions to help us learn more about our audience and their perspectives on AI. 

If you contribute, your input will be used to fuel one-of-a-kind research into AI that helps knowledge workers everywhere move their companies and careers forward.

Click here to take this week's AI Pulse.

Listen Now

Watch the Video

 

 

 

Timestamps

00:00:00 — Intro

00:03:32 — AI Pulse

00:08:42 — Gemini 3

00:18:11 — Nano Banana Pro

00:26:06 — GPT-5.1 Pro

00:28:54 — Nvidia Earnings

00:36:14 — Nvidia, Microsoft Invest in Anthropic

00:40:28 — OpenAI Board Member Controversy

00:42:43 — AI Gets More Political Over State-Level Regulations

00:47:22 — AI Super Pac Flexes Its Muscles

00:49:40 — Karen Hao Book Controversy

00:54:19 — New Jeff Bezos AI Startup

00:56:38 — Amazon Gives Rufus Rave Reviews

00:59:18 — AI Gets Further Embedded Into Education

01:01:34 — Entry Level Unemployment

This episode is brought to you by AI Academy by SmarterX. 

Accelerating AI literacy and transformation isn’t optional. It’s a career and business imperative. AI Academy by SmarterX and our AI Mastery Membership help professionals and organizations thrive through the disruption and uncertainty ahead. Use code POD100 to save $100 on an Individual AI Mastery Membership here. 

Read the Transcription

Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content. 

[00:00:00] Paul Roetzer: By the end of the decade, we will have probably our first $10 trillion company and probably multiple of them, and it will be because of artificial intelligence that that'll happen. Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable.

[00:00:18] My name is Paul Roetzer. I'm the founder and CEO of SmarterX and Marketing AI Institute, and I'm your host. Each week I'm joined by my co-host and SmarterX chief content Officer, Mike Kaput, as we break down all the AI news that matters and give you insights and perspectives that you can use to advance your company and your career.

[00:00:39] Join us as we accelerate AI literacy for all.

[00:00:46] Welcome to episode 180 2 of the Artificial Intelligence Show. I'm your host, Paul Roetzer, along with my co-host Mike Kaput. We are recording this on Friday, November 21st. It's coming out on its usual Tuesday slot, but I [00:01:00] am heading off on vacation, so I did not wanna do this from vacation. and actually I mentioned on the last podcast we were thinking about not having an episode 'cause it was the holiday week.

[00:01:09] But then Google shows up with Gemini 3 and Nana Banana Pro and Nvidia earnings and like, we could not do an episode. So here we are. We are gonna do, it's probably gonna be like a little bit of a hybrid. It's gonna be more rapid fires, than, than normal. Like not. We usually do three main topics and then like, you know, about seven to 10 rapid fire items.

[00:01:29] If you're new to the podcast. That's the weekly format. We're gonna do like modified main topics that I would say. I'm gonna try and keep this a little lighter for the holiday week. see if we get in at under an hour. I don't know. They're usually like an hour 10, hour 20. we'll see how this goes.

[00:01:44] Alright, but, lots to talk about. So we had to, had to do it. All right. This episode is brought to you by AI Academy, by SmarterX. AI Academy helps individuals and businesses accelerate their AI literacy. And transformation through personalized learning journeys and an AI [00:02:00] powered learning platform that we just launched two weeks ago.

[00:02:04] there are nine professional certificate course series available on demand now with more being added each month. One of those fee featured certificate course series that I teach is AI fundamentals. This eight course series is an essential starting point for personal and business AI transformation.

[00:02:22] The courses take you beyond headlines and hype to provide a complete foundational understanding of artificial intelligence. You're you journey from core concepts and the current state of AI to the practical skills needed to use generative tools, and you'll gain the strategic foresight to prepare for what's next.

[00:02:40] Upon completion, you'll have the clarity and confidence to not just navigate the age of ai, but also lead within it making you an invaluable asset to any team or organization. So, as I mentioned, there's eight courses. It is intro to ai, AI concepts 1 0 1. Even if you're not super new to ai, I think you'll get tremendous [00:03:00] value of AI concepts.

[00:03:00] I got value out of creating that course. it was, it was really cool to kind of go through everything. State of AI is the third course we have AI timeline where I look out at what's gonna happen between now and 2030. Generative AI 1 0 1, prompting 1 0 1 AI agents 1 0 1, and then it wraps with AI and you, so you can learn more about the AI Fundamentals course series AI Academy and our AI mastery membership program at Academy Do SmarterX ai.

[00:03:29] Ai. Alright, Mike, we are into our AI poll. 

[00:03:32] AI-Pulse Survey

[00:03:32] Paul Roetzer: So this is our informal poll we do with our listeners each week. So we only had a few days of this one. we got 79 responses. As of the moment of this, they're, they get somewhere in that range of like one 100 to one 20 so far, which is why we're deeming them an informal poll.

[00:03:46] Right now. It's more about sentiment of our audience. as we, this is only the third week we're doing this, so as these kind of grow, we might move more into that territory of actual, like formal survey. But for right now, informal poll, we learn interesting things. So last week we [00:04:00] said, how do you feel about new AI apps that create interactive digital avatars of deceased loved ones?

[00:04:07] So this was actually a topic, Mike, that we talked about on, what was that Episode 180, right? Yep. And I don't know Mike, like this, this one was weird. I, I, if I had to predict what the answers would've been, it would not be the mix we're looking at here. So again, how do people feel about digital avatars of deceased loved ones?

[00:04:28] about 28% say deeply concerning and unethical. 32% say mostly concerning with some risks. 24% are at I am undecided or have mixed feelings. And then 11% potentially comforting for some, and then a positive use of technology for legacy is the smallest. But, yeah, I don't know. I honestly would've thought that deeply concerning and unethical would've been way higher.

[00:04:58] Like if I'd had to predict [00:05:00] that one 

[00:05:01] Mike Kaput: right. I would've too. I mean, if you that up, it's like, what, 60% of people? I would've thought it would be like 80 plus. 80 90. 

[00:05:08] Paul Roetzer: But again, like, I think that's, this is really illuminating to me because this is why we're doing these surveys is like, we talk about this stuff, we think about it, but we don't want it to just be our perspectives or to assume that how we may, you know, feel about it is how everybody else is thinking about it.

[00:05:26] so yeah, I don't know. I That's great. Like I, I'm actually like really happy we're doing these pulse surveys. Yeah. Because this is exactly the kind of thing I wanted to see. The second one also kind of fascinating, so. We asked with AI generated music now topping the charts, do you feel AI created work holds the same creative value as human made work?

[00:05:47] so the biggest percentage, 49% is, it's a different category of creativity, entirely. 24% said sometimes, but it's not the same. [00:06:00] 13% say no. It lacks genuine human creativity. 9% that yes, the final product is all that matters. So, wow. I mean, like 22% are like, well, okay, so 13%, no. It lacks genuine human creativity, sometimes not the same.

[00:06:16] So, I don't know, like four, so 49%, it's a different category of creativity altogether. I get that. Um. 

[00:06:24] Mike Kaput: I would've thought, I genuinely would've thought more people would've said, no. It lacks any, yeah. That's, you know, this is murkier, I think. Yeah. I would've think that, think when I was leaning for two. Yeah.

[00:06:33] Paul Roetzer: Yeah. And I, I, yeah. So again, I think like, again, informal poll. I think if you polled general society or if you polled creatives, like purely creatives, that number would be way higher. Yeah. So this tells me like, our audience is very forward thinking about creativity and intellectual property and like, open to the fact that [00:07:00] everything's changing, like the world is changing.

[00:07:01] Like, that's kind of what this tells me. It's like, eh, it's just evolving. Interesting stuff. Okay. so this week again, you can do what's our url? Mike smarterx.ai/pulse. Correct? Yeah, that's the, okay. So you can go and participate, this week. So again, SmarterX AI slash pulse. As a reminder, we do not collect email addresses.

[00:07:22] This is not a marketing ploy. this is purely for research and to get a sentiment of how our audience feels about these topics. So you'll go to a Google form, it might connect to your email address, but that we're not doing anything with that information. It doesn't come to us, and we're not marketing to you as a result of this.

[00:07:39] how frequently do you currently use Google Gemini? Any, any form of it could be in Google Workspace, could be in notebook, LM could be straight. The Gemini agent, in your professional workflow. So how often are you using Google Gemini? Daily, weekly, occasionally, rarely, never. This will be an interesting one.

[00:07:59] I [00:08:00] I, yeah. I'm not gonna like say what I think is gonna happen here, but that's a good one. And then the second one is US Senator Mark Warner has warned that AI could spike unemployment for recent grads. Has your company changed its hiring strategy for entry level roles? So those are our, our two questions for the week.

[00:08:18] We'd love to have you participate again each week. We'll just do a quick recap and, and, you know, give you the two next questions, two or three next questions, and, just share the results in real time as as we go. So those are your two questions for the week. Okay. Gemini 3, Mike. I mean, that is the big, big thing of the week, and then they backed it up a day later with nano banana.

[00:08:39] so let's, let's get into those two things. 

[00:08:42] Gemini 3

[00:08:42] Mike Kaput: Yeah. Paul, Google has been very busy. They have officially released Gemini 3, the newest generation of their flagship AI models, and this comes just eight months after the debut of Gemini 2.5. And it includes the media integration into the Gemini app, which Google, interestingly reports now has [00:09:00] 650 million monthly active users.

[00:09:03] Now, according to Google, the new Gemini 3 Pro model outperforms its pre predecessor on every major AI benchmark, including top scores in mathematics and multimodal reasoning. Alphabet, CEO. Sundar Phai said The model is built to grasp nuance and context and allows users to receive accurate results with less prompting.

[00:09:24] Now alongside this core model, Google announced something called Google Anti-Gravity, a development platform for building autonomous AI agents capable of executing complex software tasks. They're also previewing Gemini 3 Deep Think and Enhanced Reasoning Mode designed for extended problem solving.

[00:09:40] So Gemini 3 Pro is rolling out to developers and subscribers as we speak. The deep think model remains in safety testing before a planned release to ultra subscribers in the coming weeks. So Paul, have you tried out Gemini 3? What are your thoughts so far if you have? 

[00:09:56] Paul Roetzer: Yeah, so I actually listened to the day it [00:10:00] came out.

[00:10:01] I listened to, I think it was the Hard Fork Podcast, and they interviewed Demis and they were talking about the model. And one of the things that. They highlighted was its visual understanding that it was really good if you gave it like really poor handwriting, things like that. It would discern it would like transcribe it and things like, along those lines.

[00:10:24] So I, the, I was trying to come up with like, what is a use case I can give it That would be like really interesting. And so I actually had a, a internal meeting from late October that was a, this crazy whiteboard session. There was stuff everywhere. And I've had a task since then to summarize that meeting notes.

[00:10:42] And every time I looked at the picture of the whiteboard, I was like, I don't even know what the hell I was saying that day. And so I was like, well, here you go. This is like the perfect use case. So I took that photo and I gave it to Gemini 3 and I said like, can, can you summarize the key points from [00:11:00] this meeting?

[00:11:00] It's from a, a, a meeting in October of this is the whiteboard. it did it better than I would've done it. Like unbelievable. And it even said things like. Looks like you meant this here. so after it did it, I said, okay, here was the purpose of the meeting. What didn't we cover on that whiteboard that we should consider?

[00:11:21] And it went through and gave me like a 10 point plan of things that we didn't touch on, that we actually didn't talk about that I was like, oh my God. Like that's actually really cool insight. So the reasoning perspective, the visual understanding, definitely experienced that with that one use case.

[00:11:38] And then I've been dabbling a little bit with it for some other use cases. So I would say at a high level, the response to this has been extremely positive online. They did a really good job with their influencers ahead of time. That seems like they gave 'em maybe two, three weeks access to this model under an NDA before the release.

[00:11:57] So as soon as it came out, you had all these people [00:12:00] that you know I follow on x and Mike, I'm sure you followed a bunch of 'em too. Who were like, this model is legit. Yeah. Like this is a state of the art. This is, you know, some were saying like a leap over GPT-5, 5.1. but definitely Google, you know, locked in the lead on this.

[00:12:18] I would say. It does seem like they've really kind of stepped it up. And as I've said many times the last couple years, even though OpenAI appeared to be in the lead, I would never bet against Google when it comes to the long game here. they, they have the deep AI research capabilities. They have the infrastructure, the chips, they have the team, you know, not only the Google team, but the DeepMind team.

[00:12:38] They have the products, seven products or platforms with over a billion users. They have a massive customer base. They have massive financial strength that OpenAI doesn't have. Like, it's really hard to look at this situation and think that Google isn't going to continue to push the frontiers here. And then yesterday I think it was, or maybe it was early this morning, I saw [00:13:00] this, article from the information that says Altman memo forecasts rough vibes due to resurgent Google.

[00:13:07] So I'll read you a couple of excerpts here because I think it explains the situation quite well. So it says, opening ICEO. Sam Altman told colleagues last month that Google's recent progress in AI could create some temporary economic headwinds for our company, unquote, though he added that openAI's would emerge ahead after openAI's researchers heard that Google had created a new AI that appears to have leapfrogged openAIs in the way it was developed.

[00:13:36] Altman, which referring to Gemini 3 Altman, said in the memo that we know we have some work to do, but we are catching up fast still. He cautioned employees that I expect the vibes out there to be rough for a bit, meaning they're gonna come out with something that is better than ours and we're gonna have to kind of like work through this.

[00:13:54] When I say ChatGPT is significantly ahead of Gemini chat bot in terms of usage and revenue, but the gap has been [00:14:00] shrinking. ChatGPT is AI to most people, and I expect that to continue Altman said in his memo. But then to build on what I was saying about Google's advantages, Google's other advantages.

[00:14:10] Economic, the article said opening eye is one of the fastest growing businesses in history from next to no revenue in 22 to a projected 13 billion this year. But it also is projected to burn more than 100 billion in pursuit of human level AI in the coming years, while spending hundreds of billions of dollars to rent servers to do it, meaning it will likely need to raise the same amount in additional capital.

[00:14:35] Meanwhile, this is a very important point. Google valued at 3.5 trillion, generated more than 70 billion in free cash flow over the pa fa the past four quarters alone. Altman assured staff that OpenAI would gain ground in the coming months, including a new large language model. This is the first I'm hearing of this.

[00:14:54] Code name Shallow Pete. I don't know how you say that. I have no idea [00:15:00] what that means. I'd have to look that one up. shallot Pete, in developing that model, opening aims to fix bugs it has encountered in the pre-training process. so we need to stay focused through short-term competitive pressure.

[00:15:12] We have built up enough strength as a company to weather great models shipping elsewhere, but having most of our research team stay focused on really getting to super intelligence is critically important. It sucks that we have to do so many hard things at the same time. The best research lab, the best infrastructure company, and the best AI platform, product company.

[00:15:32] But such is our lot in life, and I wouldn't trade positions with any other company. That is Mike. The first time I have seen. Admittance of weakness. where openAI's is, I, was it 2.5? I remember on the podcast saying, I felt like it was the first time that openAI's was the little brother. Like, yeah, Google finally stepped out and just like showed all their strength.

[00:15:56] I think it was 2.5 when I said that. And now you just see [00:16:00] it and now you see it, Sam admitting it. the thing I heard, and I don't remember which podcast I'd listened to about this, it might have been that same hard fork one, open or Google seems to have solved some more efficient ways to do pre-training that OpenAI has also been trying to figure out.

[00:16:20] And it seems like Google got there first and that enabled them to train a far more powerful, but more importantly, more efficient model. and then I think we see that in the nano banana that we'll talk about as well, but I, yeah, I mean. First time where you really see the market saying what we were saying probably six months ago, which is do not count Google out here, that they likely have far more strength in this match, this competition than people realize they do.

[00:16:52] Mike Kaput: Yeah, and I would just add to that, I realize that there's plenty of models out there. They're all roughly at the same level of capabilities. [00:17:00] We can't follow all of them and use all of them as much as we'd like, but if you are somebody who has not really investigated Gemini at all because you're becoming a Chachi B two power user, like very good for you, I think it's time that you take a look at it.

[00:17:15] I mean, I've found this to become very quickly like a go-to model for every type of work because it can handle video multimodal. We'll talk about image editing in a second. It's incredible. In every respect for knowledge work that I've found so far. 

[00:17:31] Paul Roetzer: Yeah. And to be fair, like I use both, like I still love chat GBT, but I happily pay for licenses to both technologies and when it's an important use case I'm working on, I will work with both of them until I see which one is gonna be better at that use case.

[00:17:46] Yeah. And it's not always Gemini, but, well, we always said like, you can't go wrong if you just make a bet on ChatGPT or Gemini. I know people like love Claude as well. if you just get [00:18:00] really good at one of them, if, if you decide to focus your time on Gemini, I don't think you're gonna be missing out on much if, if that's where all your time goes is, yeah.

[00:18:08] I guess one way to look at this, 

[00:18:11] Nano Banana Pro

[00:18:11] Mike Kaput: alright, next up. Google has also released Nano Banana Pro. Its new state-of-the-art model for image generation and editing. Formally, this is known as Gemini 3 Pro image, but we call it now by its nicknames. So do they, they've kind of gone all in on the nano banana nickname.

[00:18:27] This model is built on the Gemini 3 architecture and is designed to address persistent technical challenges regarding text rendering and subject consistency in image generation. So according to an announcement from Google, the model leverages enhanced reasoning to generate legible accurate text within images across multiple languages.

[00:18:46] It also integrates with Google search to visualize real world data that allows users to create context rich infographics and diagrams based on real-time information for advanced control. Google says [00:19:00] the model can maintain visual consistency across up to 14 input images and five distinct characters.

[00:19:07] Users can also perform localized edits to adjust camera angles, lighting, and depth of field. Nano Banana Pro is rolling out now to the Gemini app, Google ads and workspace tools like Google Slides. So Google confirms that all generated media is embedded with the imperceptible syn ID watermark they've been using.

[00:19:27] So if you generate something with nano banana, there is a way to tell it was AI generated and there's a new verification tool available in the Gemini app to identify AI generated content. Now Paul, from the examples I've seen, I mean that ability to nail down text representation and you know that vis the visual educational explainers, people are making infographics diagrams like these seem like a pretty big deal.

[00:19:51] It seems like it does that very, very well. 

[00:19:54] Paul Roetzer: Yeah. I mean, go back a year, six months, like if you asked it to [00:20:00] make a logo and give it your company name, like it would misspell it half the time. Yeah, like, I mean, we are not far from these things, not being able to spell. To all of a sudden they do infographics based on a PowerPoint or a long Word doc or whatever, with like near perfect spelling and understanding.

[00:20:18] I saw one, I think it was John Woodward maybe, from Google. he had shared something on LinkedIn where he basically gave it his resume and it like told this visual story of his career. I, it just, I don't know. It's wild. And then this morning I was, having breakfast with my, my wife and my daughter and I saw the, PJ Ace, who I've talked about before.

[00:20:45] Mike. We've shared his video stuff. He does, he, is an AI filmmaker, spoke at our Mayon event this year and he tweeted Nano Banana Pro is the most cinematic model on the planet. I asked Gemini to generate photorealistic leaked [00:21:00] images from the new Legend of Zelda movie. And this will change Hollywood.

[00:21:05] The detail in these 20 photos will blow your mind. They are, I mean, they're insane. Like, cinematic is a good word for it. It looks like something straight out of a movie. So two things jumped out at me here. One, I don't know what the IP guardrails are on this, like the intellectual property. They, they don't seem to have much.

[00:21:23] Yeah. Like you can, it's kinda like when Sora two came out and you could make anything. I saw ones of like politicians and AI leaders doing weird things together and like, there just doesn't seem to be the limits on this. I don't know what the safety measures are and stuff, but the intellectual property side seems very limited, in terms of what they put up into guardrails.

[00:21:45] but the second piece is, I showed this to my wife and my daughter, Zelda, happens to be my wife's favorite game. And so I said, look at this. and she just like, stared at me like, what, what is this? And so we had this conversation over [00:22:00] breakfast about what happens to designers and like digital creatives.

[00:22:04] So. The thing that I talk with them about. And then, you know, the thing I'll share with you all is like, it's what we said before, like the digital creators, the designers, the videographers, like the people who lean into this are gonna have these insane superpowers. Like they'll, they'll be able to do things with these models that you, and I like the non, you know, creators, in this world couldn't do.

[00:22:28] But it also means we're just gonna need fewer designers and creators. And I, I, I, again, I don't understand the counter argument to this, like, yeah, if you have 10 graphic designers on your staff and you go in and you play around with this and you realize what it's capable of, the reality is you could do the current level of work you're doing and maybe even do it better with probably like two or three of those designers, if they're trained and bought into using these tools.

[00:22:52] I am not saying that's a good thing. I'm saying that is just straight up a fact. these things are good enough to [00:23:00] do web pi, webpage design, infographics, brochures, logos, um. PowerPoint slides. Like anything you can imagine, we're basically at the point where these things can do those outputs. Now you're gonna need technical design capability to like take it to a final product and things like that.

[00:23:18] But you give a creative these tools, they're gonna 10 x their creativity like immediately. So that's wild. And then for the non-designers, like welcome to the world where you can bring any idea to life. Like I did this, earlier this week with a book cover. So I have this idea for a book. I'm, I may or may not write.

[00:23:35] and I wanted to like create a visual of it. And having done three books, I can tell you that the process of creating a book cover sucks. Like it almost never comes back with anything like what you're envisioning. You then have to work with the publisher's designers to try and get it somewhere. And then you have to fight with the publisher about like jamming a bunch of stuff in there that you don't want, like it's not fun to do.

[00:23:59] Book [00:24:00] covers. In about three minutes, I created a book cover that I fell in love with. Yeah. To the point where I was like, I think I have to write this book. 'cause the book cover's so damn good. so I, what I would say here is, if you have some creatives, like run a hackathon internally and come up with use cases, like just quick hack it together.

[00:24:17] Like let's find some ways to use this and like experiment this. Like just have a fun interactive session where you just design some stuff and build some stuff, see what it's capable of. And then I'll also say, Claire did the nano banana one for us, right Mike? We have a, yeah. Okay. So if you're an AI Academy, mastery member, you have access to our Gen AI app reviews that we do every Friday.

[00:24:35] There is one on nano banana already. So you want like the quick introduction and then we now have in the queue to do one on Nano Banana Pro so we can show like these enhanced capabilities. 

[00:24:45] Mike Kaput: Yeah. And I also, just one last thought. I feel like image editing is kind of about to change forever. Yeah. Like I, that's always such a bottleneck for me, where it's like, I very much appreciate love working with designers, but anytime you need a tiny little edit to something or wanna [00:25:00] see an iteration, there's a back and forth.

[00:25:02] It's really hard and tedious sometimes to get that done. That's gone away now. 

[00:25:07] Paul Roetzer: Yep. And again, like I don't, I am not trying to like devalue the work of designers. Right. Yeah. I'm just gonna, I'm gonna be like straight up, honest. Mike and I don't do design work, but we're the ones who create the content for our company.

[00:25:22] And if you look at something like a research report we do, and I can, I can tell you right now, the way we'll do this moving forward is to give it like the last three covers or last three interior pages and say. Eh, like here's what we'd really like the feeling to be like, we really like these four from these other organizations.

[00:25:40] Like these kind of hit more the feeling we're going for and like, can you help us conceptualize design concepts that we can send to our designer to work with? Yeah. Yeah. That I like. I don't know. I do this with home renovations, like I do this for architects now. It's like, Hey, here's what I'm envisioning here.

[00:25:56] Like make it actually work. But like this is, that's how I [00:26:00] use it. As a non-designer. I use these tools all the time to visualize stuff I otherwise wouldn't be able to do. 

[00:26:06] Gemini-5.1 Pro

[00:26:06] Mike Kaput: Absolutely. So our next big topic here, another kind of model release is OpenAI is officially rolled out GPT-5.1 Pro to all pro users, and this promises stronger performance in data science, business tasks, and complex writing.

[00:26:21] Now we got an early analysis from, Matt Schumer, who is that, CEO of Hyper, right? And we've talked about before. He's, very deep into using these models for everything from development to content to prompting. And he characterizes this release as a slow heavyweight reasoning model that prioritizes accuracy over speed.

[00:26:41] He describes it actually as superior to Gemini 3, four, specifically backend infrastructure and tra logic, noting that the model acts less like an assistant and more like a contract engineer that follows detailed specifications. So, pause, interested to see here, [00:27:00] not much hype at all around GPT-5.1 Pro.

[00:27:02] We kind of talked about 5.1, also kind of landed without a ton of hype cycle because they kind of got burned before. But it sounds like there's some real legs to this model. 

[00:27:12] Paul Roetzer: Maybe I don't know what the use cases are for me personally, so I still. 200 bucks a month. Right, Mike? We pay 200 bucks a month for pro.

[00:27:21] Yeah. Yeah. and I love it and I've had use cases where it's been super valuable, but to be honest, like every time I go in there and try it now, it just, it takes forever. And the outputs tend to just be way more technical than I'm looking for, or like, they lose the tone and the style that I'm used to with GPT-5.1, even GPT-4, before that I live in the thinking model, like 5.1 thinking.

[00:27:47] I just, I want the reasoning, I want it to take some more time on the task, but that's sort of the sweet spot for me. I rarely use like the auto version where just picks the model. 

[00:27:56] Yeah. 

[00:27:57] Paul Roetzer: if I was doing more technical, like data [00:28:00] analysis, I would absolutely jump into pro If I wanted to analyze, like analytics data, things like that, no doubt pro is where I would go.

[00:28:07] But for most of my strategic thinking, draft outputs, things like that, assessing work I'm doing, I'm, I'm happy with the thinking model. 

[00:28:16] Mike Kaput: Yeah, and I think there's so much iteration in my experience, required to get exactly what you want from AI models, that the delay in the output can be a real limiting factor.

[00:28:28] I don't mind waiting if I know I'm on the right track, but I wait, wait, wait. And I'm like, oh, I forgot that thing. Or it seems like such an 

[00:28:35] Paul Roetzer: ungrateful thing, doesn't it, that we're like, oh my God, it took seven minutes to write a 40 page report. Too much. I feel that all the time too. I'm like, I shouldn't be getting annoyed right now.

[00:28:45] Like it's only three minutes and it's doing this insane work for me. But 

[00:28:49] Mike Kaput: it's important to note though, because expectations are changing. I know. Yep.

[00:28:54] Nvidia Earnings

[00:28:54] Mike Kaput: All right. Next up. Nvidia has released its financial results for the third quarter of fiscal year 2026, delivering another record breaking report that serves as a primary bellwether for the global AI infrastructure market.

[00:29:07] So the company reported total revenue of 57 billion, marking a 62% increase compared to the same period last year. That growth was driven overwhelmingly by the data center division, which generated $51.2 billion alone. That's a 66% year over year jump. CEO Jensen Wong addressed ongoing market debates regarding a potential AI bubble stating instead that the industry was not in a bubble, but it entered a virtuous cycle of compounding demand for both model training and inference.

[00:29:39] He noted that the sales for the company's new Blackwell platform are off the charts and that cloud GPUs are currently sold out. So Paul, do you believe him? Does this disprove all these worries we're hearing now in the market all the sentiment and narrative about an AI bubble? 

[00:29:55] Paul Roetzer: I always laugh at this conversation because it's the, it's the thing I [00:30:00] get asked about family and friends the most.

[00:30:02] from a, from a stock perspective, investment perspective, it's like, oh, what should I do with Nvidia? Like earnings or next week, what do you think? so. As I always caveat this, this is not investing advice. I'm not telling you what to do with your Nvidia stock or whether to go buy more Nvidia stock, things like that.

[00:30:18] All I'm gonna say is, there's been a lot of conversation around this AI bubble. It comes up in like every interview you hear, like if, if Sam Alzo interview, if Jensen soon to everybody asks about the AI bubble. The AI bubble premise is that we are overbuilding, that these valuations are absurd. That Microsoft and Meta and, Google and openAI's and they're spending like tens of billions, hundreds of billions of dollars building this infrastructure, buying all these chips, doing all these things, and like the demand's just not gonna be there.

[00:30:52] That is ludicrous. Like if I got asked this on a, I did a ask me anything session with our AI Mastery members like two [00:31:00] hours ago, and this, this question came up. what. If you just listen to the media, the way I think about this is like, I started investing in AI stocks around 2014, 2015. I bet everything, like all, all my, every, everything I had went into AI stocks and I basically, my theory was that Wall Street had no concept of what was about to happen.

[00:31:20] And it took like seven years for me to be right on that one. But like I had high conviction that I was correct. And so what happens is like Wall Street lives in three months cycles. Earnings report. Earnings report. So yes, like everyone waits with bated breath every time. NVIDIA's gonna report because their projections for the next 18 to 24 months tells Wall Street whether the demand is still there, whether this massive build out is still happening.

[00:31:49] That's why they also watch for the CapEx spending of Google and Microsoft and others. They're looking to say, are they building more infrastructure to deliver more [00:32:00] intelligence down the road? That's in essence what's happening. So. The way I explain this is like we are on the very leading edge of an intelligence explosion.

[00:32:09] What we are seeing is largely been generated through chatbots that take in learn from and output text. what we are heading toward is a era where AI is everywhere and in everything. So NVIDIA's rise from 430 or four 23 billion on November 30th, 2022. So the day chat, GBT came out, there was a $423 billion market cap company to they to they touched 5 trillion.

[00:32:41] They're now at 4.4 trillion a week later. It's going to ebb and flow. It's not gonna just like keep going up inevitably. But that rise, that like 10 x in the market cap valuation in a three year period is primarily driven by the training of these models. So like all these companies that are spending all this money [00:33:00] to train.

[00:33:00] Gemini 3. I mean, they're mainly using their own TPUs, but like openAI's and others, they're using to train it. The future of Nvidia and the AI world is at inference. Meaning when you and I use the tools, so we just talked about G GPT-5.1, pro Pro uses way more computing power than the standard smaller GT five model, right?

[00:33:27] and so as we start to think about increase at an exponential rates of demand for computing power, because society will demand greater reasoning capabilities, you're gonna get used to having a model that thinks that has this like very high level strategic thinking image generation we just talked about with nano banana video generation 3D, world generation, actual autonomous AI agents, humanoid robots, like none of that is at scale yet in [00:34:00] society.

[00:34:00] All of that demands way more computing power than text in text out. And so if you think we are in an AI bubble, you have to then, as a byproduct, believe that demand won't grow for these things AI enables that we don't really see at scale and society. Yet. The only other way I see it not working is some variable unknown or an unknown, like societal revolt against this stuff.

[00:34:29] Regulation, state by state regulation that like basically, you know, thwarts a lot of this stuff, all kinds of other scenarios I won't get into that could be really bad that could stop this. But if things keep moving and we keep putting AI into all of our software, all of our devices, and it has reasoning capabilities and image and video and AI agents and all these things, then people are going to keep buying Nvidia chips.

[00:34:54] Like, and so again, it's, it's not this, I'm not providing investing advice, I'm just [00:35:00] saying. I think that there will be multiple companies, AI companies that are currently in that three to $5 trillion market cap range that will be $10 trillion companies within three years, maybe five. By the end of the decade, we, we will have probably our first $10 trillion company.

[00:35:18] And probably multiple of them. And it will be because of artificial intelligence that that'll happen. Does that mean perplexity survives? Probably not. Like does, does it mean there isn't some bubble where investments are going into companies that probably won't be here in two years? does Anthropic actually worth the kind of money they're getting?

[00:35:36] I don't know. But like there will be a bubble of sorts. Like there's all kinds of investments going into companies like safe, super intelligence thinking labs, thinking machine labs, whatever. Those may be crazy investments. They might work, but there's gonna be a whole bunch of 10, 20, $50 billion companies that just fold or sell for like pennies on the dollar.

[00:35:55] Yeah. So it's not like the whole thing is not a bubble, but the. [00:36:00] Big picture of, is AI gonna dominate society and be the underlying force that drives the economy? Absolutely. It's going to be. So if you believe that, then you know you start to look at these stocks differently. 

[00:36:14] Nvidia, Microsoft Invest in Anthropic

[00:36:14] Mike Kaput: Some other Nvidia related news, Nvidia and Microsoft have announced a combined investment of up to $15 billion in Anthropic, the creator of the AI model.

[00:36:24] Claude, this agreement establishes a significant reciprocal financial relationship while the two companies provide capital. Anthropic is committed to purchasing $30 billion in cloud computing capacity for Microsoft's Azure platform. Specifically, Nvidia will contribute up to 10 billion of the total investment securing Anthropics commitment in return to deploy one gigawatt of computing capacity powered by NVIDIA's chips.

[00:36:51] So this deal brings Anthropic closer to Microsoft, which has historically served as a primary backer for its rival openAI's. [00:37:00] However, Amazon reportedly remains Anthropics primary cloud provider following its own multi-billion dollar investment. So Paul, I guess first, what does this mean for Anthropic, but also like, how are you keeping score?

[00:37:13] Like who the heck owns which stake in which AI company, these technologies? 

[00:37:17] Paul Roetzer: it's almost becoming impossible. Yeah. I, I, so when, they announce that all the companies have to announce these things, like simultaneously, everybody puts their blog post up, everybody puts their video of their CEO explaining the deal up.

[00:37:29] Like, it, it, it's the same blueprint for everybody. And so I watched the video of Satya talking about this from Microsoft's perspective, and I retweeted and I was like, I keep track and talk about this stuff for a living like these ideals. And I have whiplash, like anyone have a great visualization of all the investments in deals between the labs, cloud infrastructure companies.

[00:37:51] And I said, for example, Google owns 14% of Cro Anthropic who has a deal with Microsoft, who owns 27% of openAI's, who has a hundred billion dollar [00:38:00] deal with Nvidia, who has deals with every lab. Everyone has a deal with Nvidia. It is the craziest situation, and I am, I don't pretend to be smart enough to like comprehend all of these deals and how they're working.

[00:38:15] it does seem like it's just a lot of a shell game. Like all like, we'll, 10 a hundred billion dollars deal. You buy a hundred billion in Nvidia chips and okay, we'll do this deal with you and then we're gonna spend 5 billion in on cloud computing next year with you. And like, it's just moving billions, tens, hundreds of billions of dollars between like seven companies.

[00:38:36] It's crazy. 

[00:38:37] Mike Kaput: Yeah. It's also like, there's plenty of big personalities and feuds and drama, but never, never sink that they're not gonna work with each other just because they're not Totally. 

[00:38:48] Paul Roetzer: Well, and that's, I tweeted, um. Oh yeah. I think I tweeted this on like Tuesday. So there was a dinner at the White House with the Saudi prince or something.

[00:38:57] Yeah. And Elon [00:39:00] Musk and Greg Brockman are in multiple photos in different rooms together. So it wasn't like they randomly were standing at each other, say, Hey, let's take a picture. They were obviously in close proximity to each other during this party, and so I half joined, I was like, I wonder if the lawsuit came up.

[00:39:16] Like Elon is suing Greg Brockman and openAI's for the founding of the company, and they're just like chilling with Jensen Wong and whoever the soccer player is. Chris. Oh, Aldo Renado. Yeah. Yeah. Like it just, he's randomly in there with like Bezos and everybody else's. Yeah. I have no idea what, like, it's so funny how, how this is all playing out and the other thing, like all the researchers that work at all these companies hang out at the same parties together.

[00:39:44] So it's like, yeah. It's presented as this, this insane competition when in reality, um. It. They're all, they all know each other. I actually saw this hilarious clip. I'll, I'll try and find it. Put it in the show notes. from the All In conference where Sergey Brin was talking about coming back to Google after [00:40:00] he like semi-retired.

[00:40:01] And he said the like, he's like, the main reason I came back was 'cause I was at a party and this dude from openAI's came up to me. He's like, what do you doing retired man. This is like the greatest moment in human history for computer science people. And he's like, oh shit. You're, you're right. Like any, going back to Google, I started working on ai.

[00:40:18] So it's just like somebody jokingly tweeted like, this dude deserves at least a billion in stock from Google, from being sur to come back and work on ai. It's funny. 

[00:40:28] OpenAI Board Member Controversy

[00:40:28] Mike Kaput: All right, next up. Former US Treasury secretary, Larry Summers has resigned from the board of directors at openAI's. This ends his tenure at the AI company after just two years.

[00:40:38] Summers originally joined the board in November, 2023. He is intended to bring some Wall Street and Washington experience to the startup. This resignation follows the release of documents by the US House Committee on Oversight and Government Reform, which made public a series of email exchanges between summers and the late sex offender, Jeffrey Epstein.

[00:40:58] According to the report, [00:41:00] the correspondence included discussions regarding public attitudes towards sexual harassment allegations. So Summers has said he is deeply ashamed of his misguided decision to continue communicating with Epstein and announced he would be stepping back from his public commitments.

[00:41:15] So, Paul, I'm curious, what, if anything, does this mean for openAI's? 

[00:41:19] Paul Roetzer: It means this is the one and only time we're gonna talk about the Epstein files on this show. Yeah. No intention of getting into the Epstein files on this show. yeah, I, again, like, I don't know, at a high, kind of zoom out macro level, there are a whole bunch of people involved with these companies, investors, board members, leaders of the companies.

[00:41:43] Who make decisions that affect all of us. And there's only about five of these companies that are basically determining the future of humanity with what they're building. And sometimes the people that are in very powerful positions, don't deserve to be there. [00:42:00] And, I don't know. I mean, that was why I even put it on the show notes was like, I didn't want to address it until I assumed he was gonna leave Open Eyes board.

[00:42:09] I I figured it would maybe take a week. It took like 24 hours. Yeah. it, we're not gonna get into the explanation Mike gave is the very surface level of how awful, the, correspondence were and the connection between the two people. We will not get into the details of that. If it is something that is of interest to you can go read it elsewhere.

[00:42:30] But for me it means there is a board seat now open Yep. At one of the most powerful companies in the world. And we'll keep track of who fills that board seat. And we will probably never speak about this part of this again. 

[00:42:43] AI Gets More Political Over State-Level Regulations

[00:42:43] Mike Kaput: All right, next up some more, AI political drama. So, US President Donald Trump is preparing a new executive order that would authorize the Department of Justice to sue individual US states over their AI regulations.

[00:42:58] So a draft of this order was [00:43:00] viewed by Bloomberg and it directs the Attorney General to form a task force to challenge state level AI laws that allegedly interfere with interstate commerce or conflict with federal rules beyond legal action, the draft proposes financial consequences. So states with AI laws deemed burdensome or restrictive could lose access to certain federal broadband funding.

[00:43:23] The administration and major tech companies, especially some of their allies, have argued that a unified federal standard is necessary to protect innovation, whereas a patchwork of state laws creates compliance hurdles. Interestingly, this is actually conflicting with state leaders from within the president's own party, including Republican governors, Ron DeSantis and Sarah Huckabee Sanders, who have been pretty vocal about enacting their own protections in the absence of congressional action.

[00:43:52] So Paul, this seems like the issue of state versus federal regulations could be actually a pretty big flashpoint, both within and between [00:44:00] parties in the lead up to us midterm elections. 

[00:44:03] Paul Roetzer: Yeah, so at a very high level, and I know I'm probably more in tune to this than most people given who I follow on X, but the amount of political conversation and tweets is like an exponential at the moment versus where it was three months ago.

[00:44:19] So there are a lot of politicians who are now getting involved from a regulation and safety standpoint, and they're picking their, their battles. You have like child safety as a hot one, jobs as a hot one. They're trying to find those wedges for the midterms next year. again, like opinion, my opinion on this, I am not an expert in federal and state regulation, so I'm gonna give this from, of an AI perspective.

[00:44:49] the state by state approach seems like an absolute train wreck waiting to happen. I completely understand why the tech companies wouldn't want that. [00:45:00] the complexity of managing and building and in theory, helping America lead in AI and being innovative, you do not want state by state, level legislation.

[00:45:12] So I totally understand that. the alternative would require the federal government to do something. They have zero intention of doing anything. So the whole point of why this is happening is because the federal government is trying to prevent, basically any regulation of the AI industry right now.

[00:45:34] And that starts at the environmental side of like, build, build, build. And don't worry about what the EPA says, all the way up to like, you know, preventing these people from innovating and accelerating. So that's the problem we're in is like, okay, state by state isn't the answer. What is, what, what are you going to do as an alternative?

[00:45:54] And so I would find myself, if I was forced to like, take positions on these things, it would be a very [00:46:00] difficult spot because I understand both sides. I also understand that the federal government isn't gonna do shit if they're not forced to do it. Yeah. So these states are gonna do it and they will absolutely be punished by the current administration for doing it.

[00:46:13] it, and like you said, this is, this is both sides. Like this is Democrats, Republicans, everybody in between. Everybody's kind of taken a spot on this. And sometimes it's actually like, oddly enough that question I asked last week about truth versus beliefs and like, what's one thing everybody would like, agree on.

[00:46:28] You're getting a whole bunch of like opposing politicians who are like on the same page when it comes to ai. It's like the one unifying thing right now in a weird way. So I don't know. It'd be interesting to watch this one play out. 

[00:46:40] Mike Kaput: Yeah. It's making some, for some very strange bedfellows. I think I saw post from actor Joseph Gordon Levit.

[00:46:46] Yes. Like vocally agreeing with Ron DeSantis his policy, which I don't know all of Joseph Gordon Levitt's politics, but I'm guessing they don't talk often as, based on the way he framed this post and people giving him [00:47:00] backlash for even talking to DeSantis. It was really interesting to see. 

[00:47:03] Paul Roetzer: Yeah, and this is, we're not gonna have to wait too long.

[00:47:05] Like, I think this is gonna start coming to a head by spring of 26 when we enter midterms. I think you're gonna have battle lines drawn on ai, so it'll be really fascinating to see what happens. And then if these state laws go through it, it's gonna affect it lot. 

[00:47:22] AI Super Pac Flexes Its Muscles

[00:47:22] Mike Kaput: Yeah, well the, here's another way This battle is shaping up.

[00:47:25] So in this next topic, this new bipartisan super pac, we've discussed it, that's backed by AI executives, called Leading the Future, has launched its first targeted campaign against a democratic congressional candidate in New York. So this group is opposing state assembly member Alex Boris, who is a co-sponsor, co-sponsor of the State's proposed raise Act.

[00:47:48] If signed into law, this bill would require major AI companies to publish safety protocols regarding misuse of their technology and mandate the disclosure of serious incidents to [00:48:00] the State Attorney General. So, leading the Future is actually funded by significant industry players like Venture Firm, Andreessen Horowitz, Palantir co-founder, some people at the search startup Perplexity.

[00:48:12] In a statement on Monday, and Greg Brockman and Greg Brockman, the PAC argued that Boris is pushing ideological legislation that would create a regulatory patchwork in handicap American innovation against global competitors like China. Now, Boris interestingly, is a computer scientist by training and responded by claiming the industry is targeting him because he understands their business too well and is in charge of regulating it.

[00:48:38] So this one even seemed like a pretty big deal as well, Paul, because like your position as a politician on AI now can determine how hard it is for you to get elected. 

[00:48:47] Paul Roetzer: Actually think, I know why Elon Musk and Greg Brock more hanging out now because Brockman is one of the leaders of this pack, and Elon obviously would side with this idea.

[00:48:58] Yeah. And Elon has obviously [00:49:00] become political in the last couple years. I wonder if that isn't the unifying thing. It's like, yeah, I'm suing you, but, but another a hundred million for your pack. I don't, right. Yeah. Again, this is, it's gonna get really messy. this is, this is what happened in crypto.

[00:49:15] They, like, rich people put their money together into these packs, and then they attack people who are anti crypto. You're gonna see this played out. And a hundred million is just the beginning of this. They're, they're gonna go after anybody that's gonna try and slow down AI for whatever reason. yeah, the administration hates it because they're going after Republicans and Democrats.

[00:49:34] Doesn't matter. They'll go after any of them. Yeah. So another interesting wrinkle on the political side. 

[00:49:40] Karen Hao Book Controversy

[00:49:40] Mike Kaput: Next up, author Karen Hao is addressing some controversy online about a statistical error in her new book, empire of ai, which is about openAI's. And this is following a viral fact check regarding the environmental impact of data centers that she talked about in the book.

[00:49:57] So. writer Andy [00:50:00] Masley posted about, empire of AI saying it was actually wildly misleading on AI's water usage. He identified a critical unit conversion mistake in one of the books' chapters on how much water data centers at Chile were using, and he demonstrated that the text appears to confuse cubic meters with liters.

[00:50:19] So a miscalculation that unfortunately inflated the reported relative water consumption of a data center by three orders of magnitude. So how actually publicly acknowledged, acknowledged the mistake on X? She said she is verifying the data with her sources and coordinating with the editors to issue a correction.

[00:50:39] This incident sparked though a really big debate kind of regarding the accuracy of AI critics. So Masley argues the error distorts the public's understanding of the industry's environmental footprint. Whereas people like AI researcher Tim, Nick Gabriel defended the work, contending that the statistical flaw does not [00:51:00] invalidate the book's broader critique of corporate power.

[00:51:03] So Paul, how big a deal is this? Like this blew up on Twitter, like what's actually Yeah. Going on here? 

[00:51:09] Paul Roetzer: So I saw this as soon as it happened, and I think I flagged it in our sandbox. And then as the week progressed, I was like, holy cow. Yeah. Like, this is just going crazy. Yep. So, yeah, I mean, in essence what happened is, Karen had taken data from a government agency that didn't include the metrics in it, like it didn't include whether it was leaders or cubic meters, whatever it was.

[00:51:34] and she wrongfully made an assumption. So Karen made an error. Now Karen is a phenomenal journalist and a, a wonderful person. She's spoken at Macom before. Um. She wrote an incredible book that shouldn't be diminished because of this error. It was, it was a mistake. the guy Andy, who did the research is like, damn, like it is an incredible analysis.

[00:51:59] Like [00:52:00] he, he was very cordial about it, but he was like, you screwed up like big and here's how you screwed up. And like, you gotta fix this. 'cause this changed. Like this drove the conversation around environmental impact. This isn't some small misleading thing. It changes the whole hypothesis of a big part of your book, basically.

[00:52:19] But he did it in a very appropriate way. I felt the problem came in with all of the accelerationist using this as the jumping off point to say everything Karen has done is basically worthless. Like the book was worthless. It, it's based on a false premise. They started attacking Karen personally, which is where Tim showed up.

[00:52:41] Tried to defend Karen. That did not go well. She did not handle it well. she basically glossed over the fact that Karen had made a mistake that Karen had already admitted to and said, I, you know, this was a problem. This is on me like I shouldn't have. And Tim shows up and just like, leave her alone.

[00:52:59] Like [00:53:00] she, you know, it's not even a big deal. Like it's still AI is still awful. Basically. Like, yeah, yeah. It was, it did not help at all. And then that just fanned the flames and then every, it just became this horrible personal attacks. And, and, long story short, though, it does appear a key part of her argument about the environmental impact was false.

[00:53:22] and that it is nowhere near the impact that it was presented to have that people are going to continue use as a, it's like the MIT study, that 95% of people don't get any value. Yeah. See that stupid study showing up like every week in headlines that congressmen, congresswoman using it in a hearing. So once you have that data point, it doesn't, she can correct the book.

[00:53:43] It doesn't matter. Like the damage is done. She can do her best to like fix it moving forward, if the publisher will allow her to reprint like the next reprint, the way it'll work is like this book is really popular, sold hundreds of thousands of copies, I'm guessing. So if you don't know how the publishing industry works, they will print runs in advance.

[00:53:59] So a book [00:54:00] selling like that probably is printing 20 to 50,000 copies in advance. I haven't met a publisher yet that's gonna eat those 50,000 copies. So you're gonna have all these books out in the world with that misrepresentation and they're just gonna have to do their best on the next run to fix it if the publisher will let her do it.

[00:54:19] New Jeff Bezos AI Startup

[00:54:19] Mike Kaput: All right, next up, Jeff Bezos is returning to a direct executive role in the technology sector to lead a heavily funded AI startup venture. So according to reporting from the New York Times, the Amazon founder will serve as coach chief executive officer. Project Prometheus, a startup developing AI systems for the engineering and manufacturing of computers, automobiles, and spacecraft.

[00:54:45] The company is launched with $6.2 billion in funding, partly provided by Bezos himself. This establishes it as one of the world's most well capitalized early stage startups, and he shares the CEO title with co-founder Vic [00:55:00] Bajaj, a physicist and former expert at Google's X Division. This Marks Bezos's first operational post formally at least since leaving helm of Amazon in 2021.

[00:55:11] And their focus, it seems, is not text-based language models. They aim to build systems that learn from the physical world to aid in scientific discovery and hardware production. The firm has already recruited nearly a hundred employees, drawing talent from major labs, including openAI's, DeepMind, and Meta.

[00:55:28] So Paul, I'm curious, like what is the play here? It's a pretty crowded space. Sounds like maybe there's a connection here to world models like we've talked about in the past. 

[00:55:36] Paul Roetzer: Yeah, I mean, robots, space, manufacturing, those are all things that Bezos is an expert in, has keen interest in. And yeah, I don't know.

[00:55:46] I mean, the fact that he's kinda, it's like the Sergei thing I mentioned earlier, like, yeah, you like Bezo doesn't need to work. Sergey doesn't need to work. but how do you miss this moment? Like, you can, you can build anything now. and [00:56:00] so to not to see him basically come off the sidelines in theory, if it's truly a COO where he is gonna be working, you know, 50, 60 hours a week, whatever that is.

[00:56:08] I think it truly is just that these people just can't watch this moment go by and not be a part of it. And they see probably better than anybody, like what could be built. And when money isn't an issue and you can jump in and put 6 billion into start, like. I, I, I, I, if you put yourself in their shoes, I could see why you can't just sit on your yacht, you know, and live your life.

[00:56:32] you gotta get back in it because it's, it's just too exciting to miss 

[00:56:38] Amazon Gives Rufus Rave Reviews

[00:56:38] Mike Kaput: some more Amazon related news. Amazon has actually attached some specific financial value to the performance of Rufuss. Its AI shopping assistant embedded directly into its mobile app and website. During the company's third quarter earnings call, CEO, Andy Jassy projected the tool will generate more than $10 billion in annual incremental sales.[00:57:00] 

[00:57:00] According to jassy, 250 million shoppers have used the assistant this year to research products with monthly interactions rising 210%. So the company is crediting Rufuss with driving significantly higher conversion rates. Amazon reports. Customers who engage with Rufuss during the shopping journey are 60% more likely to complete a purchase compared to those who do not.

[00:57:23] It is designed to answer queries ranging from broad category comparisons to specific item details. And it has now expanded to markets that include the uk, India, and France. There's also a recently added help me decide feature to guide customers through complex purchases. So Paul, this is pretty solid proof that AI done right can have an enormous impact.

[00:57:45] And I think it might also explain why Amazon, which we talked about on episode 1 79, is so hesitant to have agents from like third parties like perplexity, roaming around the site outside it's control and ecosystem. They are printing money [00:58:00] using their own kind of assistant here, 

[00:58:02] Paul Roetzer: right? And they don't want an AI agent talking to Rufuss.

[00:58:04] No. They want the human. Yeah. This tech's been in the works for a while. and you know, I, I've, I actually don't like when I go into Amazon, I'm usually like very specific. I have a thing, I don't even go in there conversations with Rufuss. But I could totally see how, you know, that could be a valuable thing.

[00:58:22] But I see this being very similar to like meta's embedding of meta ai. It just becomes the search window. It becomes the way you interact. I think that's how, you know, Google's thinking about, you know, AI mode is just, you know, anytime you're searching, you're basically interacting with their ai, which again goes to my point about it's going to be everywhere and in everything.

[00:58:40] And, you know, the demand for computing power to service things like rufuss is going to be enormous. So I don't know, but I do think there's also this, that the evolving buyer behavior, way consumers seek out information, the way they make buying decisions, the way they actually purchase, [00:59:00] whether it's them or their agent that's doing it.

[00:59:02] You know, I think this is the kind of thing we'll keep monitoring as like e as that buying process evolves and people's behavior evolves, it's fascinating to see the data, especially for things, Mike, that you and I don't really use every day. So it's hard for us to like, provide that perspective otherwise.

[00:59:18] AI Gets Further Embedded Into Education

[00:59:18] Mike Kaput: Next step, openAI's and Google are deepening their integration into education with some new tools and credentials. So first, openAI's has launched ChatGPT for teachers, which is a secure workspace design specifically for us, K to 12 educators. So this platform is free for verified teachers through June, 2027.

[00:59:38] It's a version of ChatGPT, that includes admin controls for districts and adheres to compliance standards to protect student data. So teachers can use it for their own use cases, totally, free of worry that it is unsafe or that they have to spend budget on it. Meanwhile, Google has introduced three new Gemini certifications and the new tracks target [01:00:00] educators, university students, and high schoolers, offering digital certificates to those who pass assessments on using Google's AI tools.

[01:00:08] And so basically this, Paul kind of just reiterates this trend we're seeing of both companies making pretty big education plays here. And I'm curious where you see kind of the end game with this. Are they just hoping that they become the norm in every high school, college, et cetera? 

[01:00:23] Paul Roetzer: Yeah. I assume it's a battle for that next generation of users.

[01:00:26] my kids use Google Classroom at their school. So that's kind of their go-to Gemini is not baked into it, which is an interesting thing. I wonder at what point that, that is, or maybe it's an option and like it's up to the school whether or not to turn that on. I don't, I don't actually know that, the openAI's one, I immediately think like, okay, so this is part of what you do with all that, you know, a hundred billion or whatever, you have of the equity in the for-profit company, right?

[01:00:53] So if you talk about like, you know, the well most well-funded nonprofit in history, eventually. Those are the kinds of things you can do. It's like, let's [01:01:00] just provide education to everybody. And, so yeah, it's way I look at this is I would encourage you, like if you're a parent and you have students at, especially at like high school, university, age, have them get these certificates.

[01:01:13] Like have 'em go do this stuff because yeah, it'll differentiate. Like if somebody applied for a job at SmarterX coming straight outta school and they're like, yeah, I took the Gemini certificates. I've gone through those, that training, I've gone through some open eye training. It's like, oh, okay, like that resume goes right to the top.

[01:01:28] So take advantage of this stuff. Use it to build on whatever else you're getting your, AI education from. 

[01:01:34] Mike Kaput: Yeah, no kidding.

[01:01:34] Entry Level Unemployment

[01:01:34] Mike Kaput: Because our last topic is about exactly that because US Senator Mark Warner has introduced some bipartisan legislation requiring companies to report job losses caused by ai. Now, the reason for this, as he said, on an appearance on CNBC, squawk Box, the Virginia Democrats said.

[01:01:52] He is warning that AI is poised to eliminate a lot of entry level employment roles. He predicts that unemployment for [01:02:00] recent college graduates could rise from what he says is a current 9% to 25% quote very shortly. So this bill is actually co-sponsored with Republican Senator Josh Hawley and would mandate that both private businesses and the federal government submit AI related employment data to the Department of Labor.

[01:02:21] Now, he also in this appearance argued for a public-private retraining model suggesting the AI industry should make a financial contribution to help displaced workers. So, Paul, like a rise from 9% unemployment for recent college grads to 25% is crazy. Like, is that possible? 

[01:02:42] Paul Roetzer: Oh, yeah. Look, I don't, I mean, I.

[01:02:46] I'll just be straight up. Like, I don't even know that it's possible. I think it's potentially probable, like talk to somebody trying to find a job right now, straight outta school. It's, it's tough. it's interesting just saying that I think one of the ways [01:03:00] the government might have having to do something is this might be a prelude to a secondary unemployment tax if you lost your job due to ai.

[01:03:11] Like you could, you could basically force the AI companies to pay what amounts to a penalty, a tax on getting rid of human workers, to supplement standard unemployment. 

[01:03:25] Mike Kaput: Yeah. Interesting. 

[01:03:26] Paul Roetzer: I have no idea if that's something we talked about that I, but now that I'm thinking about it, like that just seems like almost an obvious thing that would have to happen.

[01:03:32] Somebody's gotta pay the bill. Like somebody's gonna have to pay, like providing re-skilling and up-skilling is great if there's other jobs out there for them. Yeah, we're, we're gonna go through a phase where there won't be, and I think the entry level, the younger employees are probably gonna be the ones that suffer the most in the near term.

[01:03:49] There's emerging data showing that. and yeah, I mean, universal basic income is certainly like the [01:04:00] favored way to talk about this if you're an AI lab leader that Yeah, seems like their go-to. But yeah, I don't know. And this validates that thing we were saying, like it's the unifying thing. You got republican democrat who probably would never work together on stuff and they're right.

[01:04:14] Co-sponsoring a bill about this stuff. I think you're gonna see a lot of that go reaching across the aisle. And because what's gonna happen is like, it doesn't matter if you're Republican or Democrat, you lose your job. You lose your job. Yeah. Like, and so all of these senators, all the house representative, like they're all gonna represent people who are gonna be impacted by ai and.

[01:04:35] If they're gonna have very unhappy voters if there isn't something happening about it. So that is a, it is a truly bipartisan, like, unifying thing. 

[01:04:43] Mike Kaput: Yeah. It's super interesting to think about where that's gonna go. 

[01:04:46] Paul Roetzer: Yeah. 

[01:04:47] Mike Kaput: Alright, Paul, well thank you for breaking down a, very busy week, still in ai, some big releases.

[01:04:53] Just a quick, reminder for everyone. If you have not yet left us a review on your podcast platform of choice, [01:05:00] we would very much appreciate it, helps us improve the show and get into the headphones or in front of the eyeballs of more people. So please go ahead and do that for us if you haven't already.

[01:05:11] So Paul, thanks again for breaking everything down for us. 

[01:05:14] Paul Roetzer: Yeah. And again, take 60 seconds, go to SmarterX ai slash pulse and take those two questions survey this week. We'd love to have more responses to share with everybody. it is fascinating stuff. Alright, again, if you're listening to this Thanksgiving week and you're celebrating Thanksgiving, enjoy the time with your family and friends.

[01:05:33] Hopeful everybody gets a little time away and stop thinking about ai. Don't, don't spend the entire Thanksgiving dinner talking about it. I will do my best to avoid talking about it at Thanksgiving dinner. and, yeah, just have, have a great week everyone. Mike, I will probably talk to you when I get back from vacation.

[01:05:47] Right? Sounds great. All right. Talk to everyone later. Thank you. Thanks for listening to the Artificial Intelligence Show. Visit SmarterX.AI to continue on your AI learning journey and join more than [01:06:00] 100,000 professionals and business leaders who have subscribed to our weekly newsletters, downloaded AI blueprints, attended virtual and in-person events, taken online AI courses, and earn professional certificates from our AI Academy and engaged in the SmarterX Slack community.

[01:06:16] Until next time, stay curious and explore ai.
Claire Prudhomme

Claire Prudhomme is the Marketing Manager of Media and Content at the Marketing AI Institute. With a background in content marketing, video production and a deep interest in AI public policy, Claire brings a broad skill set to her role. Claire combines her skills, passion for storytelling, and dedication to lifelong learning to drive the Marketing AI Institute's mission forward.

Recent Posts

[The AI Show Episode 182]: Gemini 3, Nano Banana Pro, GPT-5.1 Pro, Nvidia Earnings, Karen Hao Book Controversy & Entry-Level Unemployment

Claire Prudhomme | November 25, 2025

In Episode 182 of The Artificial Intelligence Show, we cover Gemini 3, Nano Banana Pro, GPT-5.1 Pro, Nvidia earnings, AI book controversy, and more.

Podcasts

60% Express Concern Over AI Avatars of Deceased Loved Ones (Informal Survey)

Mike Kaput | November 25, 2025

Our latest survey reveals a significant degree of apprehension among professionals about using AI to create interactive digital avatars of the deceased.

[The AI Show Episode 181] AI Answers - Measuring AI Skills, Aligning Leaders, AI Literacy Frameworks, Overcoming Resistance & Preparing for AI Agents

Claire Prudhomme | November 20, 2025

In Episode 181 of The Artificial Intelligence Show, we answer questions about AI literacy, responsible AI guidelines, enterprise AI readiness, and more.

Podcasts