Everyday AI Podcast – An AI and ChatGPT Podcast - EP 438: AI News That Matters - January 13th, 2024

Episode Date: January 13, 2025

Is Microsoft laying off thousands because of AI? How did one small box from NVIDIA change the future of work? What are Google’s big AI shakeups? And why is OpenAI getting into humanoid robots? So many AI questions. We’ve got the AI answers with the AI news that matters.  Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan questions on AIUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:1. OpenAI Robotics Department2. NVIDIA's AI Projects and Tech3. Google's AI Updates and AGI shift4. Microsoft's Open Source Model5. Meta speaks on AI and Software EngineeringTimestamps:00:00 Daily AI news, podcast recaps, expert episodes.04:34 NVIDIA's CES keynote: Major AI GPU announcements.06:52 NVIDIA uses generative AI to enhance GPUs.12:19 Local powerful AI models enhance data security.16:10 NVIDIA forks Meta's Llama for enterprise AI.20:44 Google aims for AGI using advanced world models.22:34 Phi 4: Efficient, powerful, open-source AI model.25:35 Microsoft prioritizes retaining AI talent with bonuses.31:34 OpenAI revives robotics department for versatile robots.35:18 OpenAI urges US to secure AI investments.39:21 Observations connect over time; predictions often accurate.40:01 Prediction on AI agent numbers was impactful.Keywords:OpenAI, robotics, humanoid robots, adaptive robots, AI models, AI supercomputer, NVIDIA GPUs, DeepMind AI, Microsoft's open-source model, AI automation, Meta software engineering, US AI leadership, AI Predictions, AI industry news, RTX 50 series GPUs, Project Digits, NVIDIA's Grace Blackwell superchip, local AI computing, Cosmos, Isaac Robot Simulation, Nemotron Models, Enterprise AI, DeepMind's World Models, Google’s Artificial General Intelligence, Google AI projects, Microsoft layoffs, Microsoft Phi-4 model, Hugging Face, Coding automation, Meta's AI advancement.Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All In One Creative AI Studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. Is Microsoft laying off thousands of people because of AI?

How did one small box from Nvidia change the future of work? What are Google's big AI shakeups that they just announced? And why is Open AI getting into humanoid robots? Y'all, so many AI questions that might be swirling around in your head, right? now, we've got your AI answers. So every week, every day, you could spend hours trying to answer these questions, right? And more importantly, how they impact your company, your career, your growth. Don't do that. Don't spend hours a day trying to keep up, trying to worry about it. Join us. This is what we do. Well, we do it every day, but on Mondays, we bring you the AI

news that matters. What's going on, y'all? My name's Jordan Wilson and welcome to Everyday AI. This is your daily podcast live stream and free daily news that are helping everyday people like you and me not just keep up with AI, but how we can be the smartest person in AI at your company and grow your company and your career. So if that's kind of what you're doing, right, trying to keep up, trying to get ahead, trying to be the smart person in AI at your company, welcome. This is your new home. All right. So if you haven't already, please go to your everyday AI.com.

There, you can sign up for our free daily newsletter. In that daily newsletter, we not only recap each and every podcast slash live stream every single Monday through Friday, but we also break down everything that you need to know that's going on in the world of AI in that newsletter. Also on our website, if you didn't already know, there's like 430-some back episodes that you can listen to from some of the smartest people in AI. all sort of by category. So if you care about legal, tech, enterprise, marketing, whatever, go click that category, and you can go listen to the world's leading experts, dish it all, all for free on our website. All right.

So I'm excited to announce this, though, before we get into the AI news that matters. Next week, all right? So starting January 20th, all week, we got something special for you. So we have our 2025 AI predictions and roadmap series. So five straight days, five banger episodes. Y'all, I can't tell you how excited I am for this. So I did this last year. I did it in December, actually, with kind of our 2024 AI predictions.

And I was still getting people reaching out in October, November, nine to 10 months after the episodes, saying how helpful it was. So what we decided to do was go even more in depth, double down on not just predictions, right? That's not necessarily what this is about, but really the roadmap and how you can plan and make use of all of these predictions.

So we're going to be doing 25 of these and just five a day. Quick shows, we're going to keep them. They're not going to be an hour long for those of you that, you know, are out walking your dog or on the treadmill listening to this show. So they should be fast. action-packed. I cannot wait. So mark your calendars right now. If you're a listener of the podcast, make sure you listen to every single one. So that's next week starting Monday, January 25th, or sorry, Monday, January 20th through Friday, January 24th. Keep an eye out for that.

All right. Enough chit-chat. Let's get into the AI News that matters for January 13th for this week. And hey, live stream audience, thank you all for tuning in. Michael, joining on YouTube, Brian, Jackie, Christopher, Zofia, Gene, Joe, Michael, big crew today, Tim. Let's go, Michael says. All right, let's do it. So, Nvidia, Nvidia, All right, going to have a lot of that today. So while Open AI and Google had a straight up slug fest in December that ultimately is going to change the way we work, Nvidia said, all right, y'all, hold my AI beer.

And Nvidia straight up went to town during their CEO, Jensen Wong's keynote address at the CES, so the Consumer Electronics Show conference in Las Vegas. So this is now it seems like old news, but it happened on Monday night after our last AI News that Matters. So we've been covering a lot of this in our newsletter, but we have to bring it to you on our AI News That Matter show because it's a lot of AI News that matters. Some a little dorky, some not so much. So first, let's talk GPUs, all right? This is, if you didn't know, GPUs is what makes everything go around in this AI world, right?

So the big companies, right, that are, you know, running your AI chatbot or running whatever AI your company is using, they're using GPUs. And then people locally, right, like if you want to run local models, if you want to have the best tech, the best hardware on your computer to take advantage of this AI, you probably need one of it. Vidias GPUs actually. So, Nvidia unveiled its highly anticipated RTX-50 series GPUs at the CES keynote. So the new lineup includes the RTX 5090, RTX 5080, the 5070, and prices starting at $549 for the base new GPU. All right. So these GPUs boast pretty impressive speed increases, promising. to double the performance of their predecessor,

thanks to some updated technology like DLSSS4. So the series introduces a more compact design, making high-performance GPUs accessible also for smaller PCs. So, Nvidia is extending its RTX-50 series to laptops. Yes, you don't have to have a big at-home box, and that's the availability for that is beginning in March. So basically it's this. their new $549 R-TX 50 series GPU, $550.

Wild, because it offers pretty much the same performance as their previous Nvidia 4090 GPU, which was $1,600. All right, you might be wondering like, okay, how is that even possible? Well, I don't know. Nvidia's some sort of, I don't know, a bunch of wizards running around there. But one thing is they're using generative frames now. So, you know, this is something that is going to be a growing trend for, you know, GPUs in gaming, video animation, all of those things is generative frames.

So essentially, you know, you might have previously, you might have needed more kind of on board power to generate more. But now we're using generative AI to essentially generate frames between or to upscale frame. So that's one of the reason this is even possible and that we're able to get these huge, hugely powerful GPUs on smaller devices. A lot of Nvidia digits. I'm freaking excited about this. I don't personally have $3,000 to burn, but I might have to find $3,000 to burn. So that's because that's the price tag for the new Nvidia project digits. So this was announced also at CES. this is essentially the first,

Nvidia's first personal AI supercomputer. So it starts at $3,000 and it's designed to bring AI supercomputing capabilities to the desktop, making it accessible to developers, data scientists, and smaller organizations. So the new device is powered by Nvidia's GB10, which is its new Grace Blackwell super chip, combining a powerful Blackwell GPU with a 20 core grace GBA. CPU, all right? So essentially a bunch of crazy hardware tech all in one box. 128 gig of memory if you're a dork like me and care about those things. And it also boasts one petaflop of AI performance.

Yes, that's a petaflop. We're in a new, a new tier of flops, apparently. We're in the petaflop tier now. So this essentially allows users to handle complex AI tasks. and large language models locally. That's the big thing here that were previously limited to larger supercomputers. So, yeah, you put together, like, as an example, two of these things, because you can chain them, interconnect them.

And you can run as an example, like Lama 405B, right? A 405 billion parameter model. In theory, on one of these, you could run GPT-40. Yeah, let me repeat that. On one of these, in theory, you could run GPT40. The world's probably one of the world's top, top two, top three, most powerful large language models. Obviously, you can't download a proprietary model like OpenAI's GPT4O. But what this means is the largest.

I don't think large language models, especially open source ones, are getting larger. They're actually getting smaller. So there will be no large language model. open one that you cannot run on this first version of digits. And this is going to be the worst this thing is ever going to be. And a year or two, it's going to be cheaper. It's going to be more powerful. But that's what this signifies.

This changes how we work. Here's something important to know as well that I think a lot of people are overlooking because right now, yes, project digits can function as a standalone Linux-based workstation. So yeah, you can, you know, plug in a keyboard and a mouse and all that stuff. But it can also connect to a different primary computer for additional AI computing power for your current machine. So you don't have to be running on NVIDIA's kind of Linux powered operating system. You can hook it up to your computer, your company's computers, whatever, and it can run alongside any Windows or Mac computer.

So if you are running it just as its own computer, it runs on Nvidia's DGX operating system, which like I said, is Linux based. But it can operate and extend your current Mac or PC system. That's the thing that people are kind of overlooking. But y'all, think of this. I talk about AI every day. Yeah, I'm a big Nvidia fan. All right. I told you all like two years ago.

when hardly no one had heard of in media. I said they're the most important company in the world. No one's ever heard of, right? And they've obviously risen to prominence. And at times, you know, they're usually top three. But at times, they were the number one biggest company in the world by market cap. One from relatively unknown to biggest company in the world. I told you all that that was going to happen.

All right. It's because every company, every person needs GPUs. The future of work is generative AI, large language models. And you need power. You need GPUs. That's why this digits is extremely impressive, right? Live stream audiences, anyone going to go drop three grand on this? Michael says,

NVIDIA needs to send you one. Yeah. Hey, friends at NVIDIA, I know a couple of you are listening. Yeah, you have my address. Go ahead and send me one. I'll review it on the show. But it's wild. It's, it's how do I even?

Do you understand this? Right? two, three, four years ago, no one would have thought this is possible to run, in theory, it has the capability to run the world's most powerful large language models locally. So then you don't have to worry as much about data security, sending everything to the cloud, extended inference times because you're running everything on your own machine, right? I think even as edge AI or on-device AI was more of a trending topic a year and a half ago, two years ago maybe, we were talking about small language models, right?

We were talking about the many, many versions. Now we have the capability in the future, at least open source models that you can download, right? So when we talk about meta's Lama or Nvidia's Nemotron, which we're going to be talking about here, running it locally. Wow. All right. More Nvidia news, Cosmos. So, Nvidia has unveiled Cosmos. All right.

So that is designed to enhance capabilities in humanoids, industrial robots, and self-driving cars. So according to Nvidia, Cosmos is capable of generating images and 3D models of the physical world, unlike language models that focus more on text generation. So Jensen Wong in his keynote, which if you haven't listened to the keynote yet, I highly recommend you do if you want to know where AI is going and AI is heading. So he was talking about this demonstrating Cosmos' use in simulating warehouse activities, showcasing its training on 20 million hours of real world footage. Let me say that again.

Cosmos was trained on 20 million hours of real world footage. Wow. All right. So the goal of Cosmos is to enable AI to better understand and interact with the physical world more effectively, rather than just creating content. So companies like agility, figure AI, Uber, Wave, and all these others are already employing Cosmos to advance their robotics in autonomous vehicles technology. NVIDIA also announced some enhancements and updates to its Isaac Robot simulation platform, which is essentially a gym or a workout place for AI robots. They go into NVIDIA's cloud and go work out and learn.

So that will now help robots better, learn tasks more efficiently by generating synthetic data from, limited examples. So the introduction of Cosmos and updates to Isaac are expected to appeal the businesses aiming to develop and deploy humanoid robots in various settings. Yeah, we're going to be hearing a lot about humanoid robots in the very near future. Adobe just introduced an entirely new way to create, bringing the power and precision of its creative suite into one conversational experience. Meet Firefly AI assistant now live in the Adobe Firefly app, the all-in-one creative AI Studio. Powered by Adobe's creative agent, Firefly AI assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the assistant.

The assistant orchestrates multi-step workflows, drawing on 60 plus prograde tools across Adobe Creative Cloud apps, including Photoshop, Illustrator Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks, like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director. Adobe Firefly AI assistant now in public beta.

See it today at firefly.adopi.com. One last. One last, invidia news story, y'all. Nvidia also launched Nemotron models. So they unveiled their new Nemotron large language family of models. So previously, Nvidia essentially sent out a blog post when they kind of announced or unveiled their one variation of Nemotron. But now it got super official at CES.

It's a whole family now. So Nemotron, which you're going to be hearing a lot more about, is built on Madas Lama Foundation models. All right. So these large language models are designed to power robust enterprise grade AI solutions. So Jensen Wong highlighted the transformative potential of these new models in software coding, predicting that AI assistance will become essentially the same thing as coding, right? So I'll leave.

Well, actually, I have another new story on that later. So I won't give too much of my hot take on that now. So the Lama Nemotron models are available. in small, medium, and large sizes, suitable for deployment on PCs, edge devices, and cloud environment. So why is Nvidia doing this? Why are they essentially forking Mata's Lama models? Well, they're really trying to get even more of a hold in Enterprise AI, right? You know, they had their chat with RTX platform, which I think was good, but you had to have only a sales. specific GPU on your computer to run it, and it was somewhat limited.

So by essentially, Nvidia is saying they're going all in on forking Meta's Lama, but I think it's smart because they can now more quickly enter the enterprise AI space with powerful tools customized for businesses' needs. And if you follow things like, I don't know, model rankings, benchmarks, right? actually Meta's Nemotron, originally, their first variation of it, kind of ranked or scored higher than the model it was based off. So all the smart engineers over there in the San Jose area in California somehow whipped up a better version of Lama themselves. Also, this boost demand for their hardware. So these AI models, especially the medium and large one, require significant computational power to run effectively.

So businesses using Lama Nemotron will likely need NVIDIA's GPUs, right? Or as an example, their new digits, right? So smart here. InVIDIA really doubling down on their investment on this open, it's not truly open source, but open source-esque variation of Lama. I do expect once we get benchmarks, I expect some pretty impressive benchmarks. And I do express some enterprise companies to see this and be like, yeah, we're going to be at least putting this in our LLM stack almost immediately. So it does make sense for Nvidia to probably try to improve existing open source software that's out there. Because if companies latch onto it, they're probably going to need even more GPUs.

All right. We got away from the Nvidia news, y'all. And live stream audience, podcast audience, too. Let me know what you think. Were you impressed with what Nvidia announced at CES? The thing I'm more shocked by, if I'm being honest, they're going to be doing this. And video is going to be doing this all again in like three months at their GTC conference,

which I attended last year. I may or may not attend this year. We'll see how things go with my schedule. But the fact that they're probably going to be announcing a slew of new updates in like three months, I don't comprehend how they can innovate this quickly. All right. Speaking of innovating Google, DeepMind is creating a new team of AI researchers focused on developing world models to simulate physical environments, according to a recent announcement from Tim Brooks.

So Tim Brooks talked about on the show last week or maybe it was the week before previously was a co-lead. This is why it's important. He was previously a co-lead for Open AIs Sora, their AI video tool. So he joined DeepMind in October and is leading this new initiative with DeepMind doubling down even more on world models, essentially trying to advance Google's AI capability and simulating real world scenarios. So world models are cutting edge development. in AI, trying to revolutionize, yes, things like video games, movies, and realistic training environments for robots. But even more so than that, this is ultimately for all of these companies that are pushing

toward artificial general intelligence. These world models are a big gap in what they need, right? Essentially, all these large language models have been trained on the entirety of the internet, scraped, copyrighted content, everything, but not what actually happens out in the real world, right? So essentially, large language models have never went out and touched grass, right? I'm sure there's some people that tell me that internally. They're like, Jordan, you should probably go outside and touch grass.

You're in front of the computer too much, talking about AI too much, right? So this is the equivalent, right, of large language models going outside and touching grass. It's world models, right? So real world physics, the relationships between things in the real world, right? This is what these world models are for. So this initiative, Google's kind of new forming team, is part of their broader strategy to achieve artificial general intelligence, AGI before its competitors, emphasizing the importance of scaling,

pre-training on video and multimodal data, not just text only. So the project is in competition with other AI advancements, like we talked about, SORA, InVIDIA's Cosmos platform like we talked about in World Labs. So DeepMind's new team will collaborate with existing Google AI projects, including the Gemini AI models, Vio, Video Generator, and Jeannie, which simulates 3D environments in real time. So, according to job descriptions,

DeepMind is seeking research engineers and scientists to tackle challenges related to training at massive scale in integrating world models with multimodal language models. It was also announced that the teams at, or sorry, the teams that develop Google AI Studio and the Gemini developer API will be moving under Google DeepMind. So little footnote there, but extremely important. More large language model news, Microsoft has released its open source Phi4 model.

So Microsoft has released its 5-4 model as a fully open-source project on Hugging Face. So the 5-4 and, you know, make sure to go check it out in the newsletter, but if you're just wondering, if you want to look it up, it's P-H-I-4. It is a 14 billion parameter model offering powerful reasoning capabilities while still being efficient in resource usage. So according to Microsoft, this update, so from the 5-3 makes the model accessible to a wider audience, including commercial applications under the popular permissive MIT license. So 5-4 excels in benchmarks, scoring over 80% in tests like math and math.

MGSM outperforming larger models such as Google's Gemini Pro. So, you know, you're not going to see smaller models like 5-4 outperforming the world-leading models like Google's Gemini Pro, Ultra, Open AIs GPD-4-O on all benchmarks, but it is pretty impressive that a much smaller model is already competing with the world-class models on certain metrics in an open model nonetheless. So Microsoft's decision to open source the model aligns with an ongoing trend of fostering innovation and transparency in AI development. So the move could impact the AI landscape, right, by making advanced AI capabilities more accessible to organizations with more limited resources. Joe says, can't wait to get my hands on 5-4 to test.

It is available now. We'll be leaving the link in our newsletter. You have to have a pretty powerful. set up right now to run it locally, but it is available. Actually, no, it shouldn't be that powerful if you have a pretty, pretty new computer, 1632 gig or RAM, you should be able to do it. I don't think any of my computers will do it. Well, maybe my new co-pilot plus PC.

I'll have to see if that can run it. All right. Next, Microsoft, according to reports, is already reporting, reportedly planning to lay off. about 1% of its workforce, which could result in thousands of jobs. So the layoffs are said to be performance-based and span across the security division and other departments. So despite these layoff, Microsoft tends to backfill roles vacated for performance reasons, which means the overall employee count at Microsoft might remain stable. So as of last reports, which I believe was in June,

Microsoft had approximately 228 full-time employees. So the move follows previous workforce reductions at Microsoft, including a notable cut of 10,000 employees in January 2020, which at the time was less than 5% of their workforce at the time. So the layoffs come as Microsoft continues to invest heavily in the emerging artificial intelligence market, aiming to maintain its leadership position. Microsoft is also reportedly focusing on retaining key talent, particularly in its AI initiatives,

by offering retention bonuses in the form of stock or cash reward. So an internal document, according to reports, shows that Microsoft managers are prompted to evaluate the potential harm of losing specific employees, especially those critical to AI efforts. So a question about an employee's contribution to AI projects has reportedly been added to performance evaluations, targeting a specific large group within the company. The focus on AI talent retention highlights Microsoft's strategic shift toward AI, with some employees being moved from the team's chat app to AI projects like co-pilot. We're going to see this all across the board, y'all. I'm not going to be the AI doom and gloom guy, but I don't know. I literally said this back in, uh, 20, 23, early 2023 that I said quarter four,

2024, uh, and thereafter. I said, that's when the layoffs are going to start to hit. Uh, because what happens is first big tech companies like Microsoft, like Google, like Amazon, like meta, they're going to stop hiring new employees. as the technology grows, right? I think they're going to not replace people that leave and they might hire some, but I think we're going to see at the big tech company, we're going to see overall headcount kind of go down, right?

Yes, a lot of them were a little overinflated thanks to post-pandemic or mid-pandemic over-hiring. But I do think that we are going to see a lot of layoffs. So 2024 was a record-breaking year for. tech layoffs. And I think we are going to see that trend to continue even more so in 2025. And unfortunately, it is a lagging effect. So the rest of the business enterprise world, especially here in the U.S., is going to follow suit later in the year. Speaking of that, meta. Yeah. So meta, CEO Mark Zuckerberg announced that meta plans to automate the work of mid-level software.

engineers using AI, potentially changing the entire landscape of the tech industry. Yeah, Zuckerberg has gone full all, like 180, it seems in the last month or two, right? So Matt is also doing some things that some people are looking at as controversial, right? Kind of deemphasizing some initiatives internally that were previously emphasized, getting rid of human fact checkers, and instead using a community note system. But Zuckerberg straight up said, yeah, we're going to start to automate a lot of the work that are mid-level software engineers we're doing, and we're going to use AI to do that. What's interesting here is he emphasized mid-level engineers, not entry-level, right? So that shows the people building the technology know what large language models are capable of.

This isn't always just the low hanging quote unquote entry level easy jobs. It's sometimes mid-level jobs. So according to Zuckerberg, Meta, and other tech giants, as the key there, are aiming to have AI that functions as mid-level engineers capable of writing code effectively by 2025. So the transition to AI-driven coding might initially be called. possibly, but Zuckerberg believes it will eventually lead to all coding in Meta's apps being done by AI. Currently, mid-level software engineers at Meta earn close to mid-six figures, not close to six figures, close to mid-six figures.

That means most of them are earning about a half million a year. So interesting, right? And I do think as we see, it's just a wind of change that we've seen in the coding space. Right. I actually think it was InVitya CEO, Jensen Wong, that kind of started this conversation last year saying like, hey, yeah, I wouldn't teach kids how to code, right? And everyone's like, whoa, what do you mean, right? Shouldn't everyone learn how to code?

Isn't coding becoming more accessible, right? And it's like, yeah, but it's also becoming natural language, right? You can literally code an app, quote unquote, code an app without knowing anything about code. You can just use something like Microsoft GitHub copilot, something like cursor, something like lovable, windsurf, et cetera, and just be like, yo, go code me this. Or, hey, here's a screenshot of something I use. Here's what it does. Go build me this, right?

And you can, if you know what you're doing, one shot, an entire app, right? you could clone something, something that you need. That's the future, right? And I know I've been saying it for a while. You know, some of the tech leaders have been saying it for a while, but I think we have to actually pay attention here. When we get official and companies and leaders at these companies are literally saying, yeah, AI is going to be doing most of our coding, even at a kind of mid-engineer level, right?

Not just entry-level stuff. Michael saying I've used Replit and it's incredibly simple and intuitive. Dennis says, as Gardner predicted in recent trend reports, you will begin to see knowledge workers unionize. Yeah, I do think that's coming, Dennis. And I think we're going to have an AI researcher from Gartner on the show here in the coming weeks, FYI. All right. Open AI.

Going back all in on robotics. So we talked about initial rumors on the show a couple of weeks ago, but now Open AI has made those rumors official and has announced the revival of its robotics department, aiming to develop a general purpose, adaptive, and versatile robotic fleet. So according to a social media post by OpenAI's hardware director, the company plans to create robots with a custom sensor suite and AI developed internally. So job listings reveal that OpenAI's robotics team will focus on integrating cutting edge hardware and software to explore various robotic form factors. The company is looking to employ contract workers to test robotic prototypes, suggesting the potential inclusion of robots with limbs. Yay. I don't know. I don't. I don't, I don't, I'm weird. I don't really want a bunch of AI humanoid robots, but it doesn't matter if any of us want them or not. They're clearly coming. So Open AI has explored building humanoid robots with aspirations for full-scale production in the future.

The robotics sector, yeah, obviously been booming recently and has seen some significant investments, raising over $6.4 billion from VC firms last year, highlighting the growing interest in robotics technology. So companies like X1 and Figure, backed by OpenAI and Vigr, backed by OpenAI and VE, and just about everyone else are working on creating humanoid robots, though, the challenges remain significant. And, y'all, AI, not a new sector. Robotics, not a new sector. Autonomous, you know, humanoids, not a new sector.

Why it matters now, it's the intersection of all of the work that has been done in the prior decades, but obviously with the emergence and the ever-growing capabilities of large language models, right? So large language models are no longer text base. Right. So previously, like I said, this is not new. There's been great autonomous robots used in factories before chat. GBT even was a thing. But this is how it changes.

It brings the cost down. Capabilities go up. So the possibilities expand, right? Because these large language models in 2025 and beyond are multimodal by default. Right. Whereas before, as an example, let's just look at GPT4, right? Technically, it was three different models kind of working together under the hood. So, you know, things like processing, you know, a photo or producing an audio output, essentially the way it worked is three different models would kind of pass information to each other. And that created more latency. It created sometimes

errors or an increased hallucination rate. So now models are multimodal by default. What that means is they are much faster, much more capable, and less error prone when dealing in multimodal facets. All right. So if you're like wondering, like, okay, what's all this about robotics? Robotics are new. Well, robotics plus multimodal gen AI is extremely new, right?

And that's why the sector is exploding. All right. Our last piece of AI news for the day. Open AI has released a 15-page document titled Economic Blueprint, emphasizing the need for the U.S. to secure investments and supportive regulations to maintain its lead in AI over China. So the document highlights the significance of chips, data, and energy as crucial elements for winning the AI race, urging immediate action to establish nationwide rules. This announcement comes just before President-elect Donald Trump takes office here in the U.S. with expectations of a tech-friendly administration. OpenAI CEO, Sam Alton, did contribute approximately $1 million, as a lot of the leading tech, you know,

tech guys out there in Silicon Valley did to Trump's inaugural fund, indicating a strategic move to strengthen tides with the new administration. So the blueprint, which we will be linking to in our free daily newsletter warns that an estimated $175 billion in global funds are ready for AI investments, which could potentially flow to China if not attracted by the U.S., therefore increasing China's global influence. Open AI proposes export controls on AI models to prevent access by adversary nations that might misuse the technology. An event is planned in Washington, D.C. later this month by OpenAI to discuss these

proposals further. All right. All I'm going to say is a lot of today's news stories aligned perfectly with the series that we're launching next week. All right, but thank you for tuning in. You all. I'm going to quickly recap everything, the AI news that matters. So, Nvidia with a lot at CES and their keynote from Jensen Wong, four big stories we cover today, their new 50 series GPUs, their new project digits, a personal AI supercomputer. They unveiled Cosmos, kind of a new era, what they hope to be a new era for AI robotics and automation. Then their Lama Nemotron models, small, medium and large.

Then we saw DeepMine assembling a team to develop AI world models, as well as a shakeup in their structure there. Microsoft released their open source by four model. Microsoft also reportedly are planning to lay off up to a thousand employees amid AI investments and a kind of refocus on AI talent internally. Mattas Mark Zuckerberg said there's a push toward AI automation in their software engineering. And then Open AI is reviving their robotics efforts and just unveiled their strategic blueprint to maintain USAI leadership.

All right. I hope this was helpful. If it was, I'm just going to say this. Mark your frigging calendars. If one of your big goals in 2025 is to better understand the AI landscape, maybe you want a raise at your company. Maybe you're trying to land a new AI job. Maybe you just feel overwhelmed, right?

and you want to secure your career. You want to lead your department. You're a small business owner. You're trying to make sense of everything. I don't know if anyone knows this. This is what I do every day, right? And I've been doing this for, I don't know, a long time. I think more than two years, almost two years now, right?

So I've talked to on this show, hundreds of AI leaders from big tech companies like Google, Microsoft, IBM, OpenAI, right? And then small startups and every day, hundreds of just enterprise business leaders. I've learned a lot from them. And sometimes it's not until later when I can start connecting all these dots, right? It's like I get these little breadcrumbs from people. And then eventually, as days become weeks and as weeks become months, as months become quarters, I start to piece these things together.

It pick up on trends that maybe no one else is picking up on, if I'm being honest. I'm not trying to toot my own horn. But go back and look at my 2024 predictions from December, 2023. They were eerily spot on, right? I gave 24 of them, depending on how you look at it, about 21 of them technically came true or didn't go false. Maybe I was slightly off on one. I said we would have more AI agents in 2024 than humans. Although we got a lot of reports that even single people had tens of thousands of AI agents.

we didn't see like an official number, but for the most part, did a very good job of predicting what 2024 would look like. And I literally had people reaching out in September, October, November, who had just listened to that episode and said, this changes my perspective on everything. And all of a sudden, I am understanding things at a much higher level and telling me the impact that has had on their career or their company. So 2025, I'm changing it. five episodes, you need to tune in.

I don't say that a lot, right? More or less, I literally tell people all the time. People will reach out and they're like, oh, Joe and I listen to the show every day. I'm like, why, right? I made this. Maybe you tune in two or three times a week, maybe two or three times a month, and that's okay.

I never say you need to listen to every single episode. Now, listen when you want. Listen when something catches your eye. All of you, if you're listening to this right now, do not miss a single episode. Period. I'm going to say that. All right. If you want to get ahead in 2025, literally, we are laying out the blueprint for you, the roadmap for you, based on hundreds of conversations, thousands of hours of research. We're laying it all out for you. So I hope this was helpful. If so, tell someone about it. If you're listening on the podcast, make sure to subscribe so you don't miss our new series launching January 20th. Please leave us a rating if you find it helpful. If you're on social media, tag someone that needs to hear. this, repost this to your network. We'd super appreciate it. Thank you for tuning in. Hope to see you back tomorrow and every day for more, everyday AI. Thanks y'all. Meet Firefly AI Assistant. Now live in

Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us.

If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next. time.