AI Daily News March 19 2025: 🥰 Disney's robots will be powered by Nvidia and Google ✏️ Gemini gets new coding and writing tools 🤖Nvidia unveils future AI chip lineup 🧠Nvidia unveils a foundation model for humanoid robotics 🤖Mistral
15:33 Share

Right, kicking things off. This is AI Unraveled, the podcast created and produced by Etienne Newman, senior software engineer and passionate soccer dad from Canada. If you're digging these deep dives, we'd super appreciate it if you hit that like and subscribe button over on Apple. Back again, ready to explore the AI landscape and see what really matters. Absolutely.

And today is, well, buckle up, folks. It's a wild ride. We've got a whole stack of AI news from March 19th, 2025. We're going to break it down, pull out the important stuff, the stuff that matters for, you know, for you guys, for you to understand where AI is going and hopefully have some, you know, those aha moments along the way. Sounds exciting. It is. We're talking new AI chips, robots that are practically here, the legality of AI art, even AI tackling huge societal problems. So let's jump in. Let's do it.

All right. First off, felt like everyone in tech was glued to NVIDIA's GTC conference. Oh, yeah. Big one. Huge, right? Yeah, definitely. And the star, no question, was Jensen Wang, the CEO unveiling the Blackwell Ultra AI chip.

The fascinating thing here is it's not just about more power. It's about Nvidia's vision for this chip. They call it the chip for the age of AI reasoning. Okay. Age of AI reasoning. That sounds pretty big. What does that actually look like? So imagine AI that can take in and understand all kinds of information at once, text, images, video, you name it, seamlessly. Melody modal, like they say. Exactly. But it's also about making decisions in real time, like really fast.

and running the complex systems that can control things autonomously. Think about delivery robots all over a city making decisions on the fly based on a ton of data. That's the kind of thing we're talking about. It's not just faster AI, but a whole new type of AI with applications we can't even imagine yet. Right. Well, NVIDIA's goal is even bigger. They see Blackwell Ultra as the foundation for research into artificial general intelligence.

AGI. The big one. Yeah. AI that's as smart as a human across the board. Still theoretical, but that's where they're aiming. And this is where it gets really interesting.

What could this mean for us, for people listening in the near future? Well, the immediate thing you might see is AI models getting much faster and more efficient. Right now, hardware limits how complex and data heavy AI can be. Blackwell Ultra could change that. Breaking those bottlenecks. Exactly. So better AI in everything from virtual assistants that actually understand you to self-driving cars that are, you know, even safer. And beyond those everyday things. Yeah.

Think about medical research. Being able to process complicated biological data in real time could totally change how quickly we discover new drugs. Wow. OK, so from super smart chips to, well, let's talk about AI art. Yeah. There was a pretty important legal decision about copyright.

Yeah. The U.S. appeals court ruled that if AI makes art without a human being directly involved, it can't be copyrighted. They're really emphasizing that human creativity has to be there for something to be protected by copyright. So an AI paints a beautiful picture all on its own and nobody owns it. Basically. Yeah. According to this ruling, that's the situation. Interesting. So what does this mean for, you know, for artists?

for the whole AI and creativity thing. It's a big deal. It clarifies the legal side of AI generated content, which will probably lead to more discussions about ownership and fair use. It might also make artists see AI as a tool to work with rather than something that replaces them. Collaboration, not competition. Right. If the artist is guiding the AI, shaping the output,

then they can still copyright the work. It's all about human intent, human direction. That's the key. Okay, from the digital world to the real world, it looks like AI robots are getting seriously advanced. Definitely. NVIDIA, again at GTC, showed off Blue,

It's an AI robot they developed with Disney Research. And this isn't your factory robot. This is designed to interact with people naturally, like in entertainment or service roles. Disney and AI robots. That's not something you hear every day. What can this bling robot actually do? From what they showed, it moves really smoothly, lifelike. And it's got expressive facial expressions. The goal is for it to interact with people in a way that feels natural.

you know, natural and engaging. So you're picturing robots at Disneyland having conversations with visitors. Yeah, exactly. And it goes deeper. NVIDIA is also working with Disney Research and Google DeepMind on Newton.

It's like a physics engine, but for robots. A physics engine. Okay. Like teaching robots how the world works. Yeah, pretty much. Disney's going to be the first to use Newton in their robots. They even showed some Star Wars droids running on it. Imagine those droids moving realistically, reacting to the environment. It's like teaching a robot how gravity works, how to not fall over. In a simplified way, yeah. It allows them to move in a much more predictable way.

That's pretty amazing. And it sounds like NVIDIA wants to share this technology. Yeah, they're releasing an open source version of Newton later this year. This is huge for anyone developing robots. It'll let them fine tune how robots interact with different materials, combine it with other AI tools. Including those from Google DeepMind. Exactly. It's like giving robot developers a head start instead of starting from scratch.

So the takeaway here is humanoid robots are coming and they're going to be in places like theme parks and hotels. Definitely. It's a glimpse into a future where robots aren't just doing tasks, but interacting with us in a more, you know, human way. All right.

Switching gears a bit, let's talk about a big societal problem. AI is being looked at as a solution to the caregiver shortage in Japan. Yeah, Japan's population is aging fast. The demand for elder care is huge and there simply aren't enough human caregivers. So companies are turning to AI. What kind of AI applications are we seeing in this area?

Companies like Cyberdyne and Mugen are using AI-powered exoskeletons to help caregivers lift and move patients. It's physically demanding work, and these robots can assist with that. - Take some of the strain off. - Exactly. And there are smart home systems that use AI to monitor patients. They can analyze movement, detect falls, and even provide personalized assistance. - Like a watchful eye and a helping hand. - Exactly. The potential benefits are huge. Less burden on caregivers,

better safety for patients, and even more personalized care. Now let's talk about open source AI models. There's a new player on the scene making some big claims. Mistral AI just released a new open source model that's supposedly outperforming even OpenAI's GPT-4 mini. What's interesting is it does this with fewer parameters. Fewer parameters usually means a simpler model, right? Yeah. So how is it beating a bigger model? It means the architecture is more efficient, the way it's trained is more effective,

The result is a model that's powerful, but also faster and doesn't need as much computing power. Perfect for things like smartphones and smaller devices. Exactly. This could be a game changer for AI development. It gives more people the ability to build sophisticated AI without needing huge computing resources or relying on big companies. Democratizing AI. Exactly. And it puts pressure on companies like OpenAI and Google to rethink how they price their models and who has access to them. Competition and open access.

Always good things in tech. Speaking of big tech, Amazon's making changes to Alexa that are raising some privacy concerns. Yeah. The next generation of Alexa is going to be much smarter, more conversational. It'll offer real-time personalization. But there's a catch. What's the catch? Before, you could opt out of having your voice recordings stored and analyzed. With the new Alexa, they need access to all your recordings all the time. It analyzes them in real time to understand what you're saying and respond better. So...

Better AI, but less privacy. That's the trade-off. And it's definitely going to get attention from privacy regulators. For sure. Okay, this next one is a bit different. It looks like AI is being used in the oyster farming industry. Yeah, there's a story out of Marin County, California. An oyster farm there is using AI to monitor and automate their operations. AI and oysters. Now that's something you don't hear every day. How does that even work? They're using AI to track water conditions, the health of the oysters, even market trends.

So they can make smarter decisions about how to run their business. Exactly. Minimize waste, predict yields more accurately. It's a cool example of AI moving beyond the tech world and into more traditional industries. This kind of AI-powered decision-making could become a big thing in agriculture and food production in general. It's amazing to see how wide-ranging AI's impact is becoming. Now, back to NVIDIA.

They've got big plans for AI and robotics beyond the Blackwell Ultra chip. Oh yeah, they're not slowing down. At GTC, Jensen Huang laid out their vision. They unveiled Isaac GR00TN1, which is...

get this, an open source foundation model for humanoid robots. Okay, break that down for me. What is a foundation model for a robot? Think of it as a starting point for developers building robots. It's pre-trained, open source, so anyone can use it and modify it. It's designed to let robots do complex tasks, learn, and adapt. It even uses a dual system architecture, kind of like the human brain, for both fast reactions and more deliberate thinking.

So this is a big step towards really advanced robots. It is. And they're also releasing simulation tools and ways to generate synthetic training data that's going to speed up robot development significantly. And on top of that, what about NVIDIA's plans for future AI chips?

Blackwell Ultra is just the beginning, right? Right. They announced the Vera Rubin AI super chip platform coming out in late 2026, then an even more powerful version, Vera Rubin Ultra in 2027. What kind of power are we talking about with those? The projections are insane. They're talking about 100 petaflops of FP4 precision per chip for Rubin Ultra.

and 15 exaflops of inference compute per rack using NVL 576 configurations. So basically going from a regular computer to like a supercomputer times a thousand. Uh-huh.

Uh-huh. Yeah, something like that. It's a massive amount of power for AI. So NVIDIA is really putting themselves at the center of AI and robotics. No question. They're pushing the boundaries of both hardware and software, and that's going to have a huge impact on everything from humanoid robots to AI applications in all sorts of industries. Okay, let's switch gears to Google and their AI model, Gemini. Google's been busy making Gemini even better at coding and writing. They want it to be a tool for everything, technical and creative. What kind of new features are they adding?

They've got this new thing called Canvas. It's on web and mobile. You can upload documents, and Gemini will generate content based on them. Like, imagine uploading meeting notes and having it write a summary or even a draft email. That sounds super useful. Right. And it can handle coding, too.

You can generate code, edit it, even preview the changes in real time. So Gemini is becoming more and more integrated into how we work. Definitely. Google's making it a practical tool for everyday tasks. Now let's talk about Tesla. They're making some big moves in the self-driving car world.

They just got their first permit in California to operate a robo taxi service. It's a big win for their plans for autonomous ride hailing. But there are some limits to this permit, right? Yeah, for now they can only use it for employee transport, pre-arranged routes, that kind of thing. They still need more approvals before they can open it up to the public. So baby steps. But they're moving fast. They're planning to launch robo taxi services in California and Texas this year.

They're even going to use their new cyber cab for unsupervised rides in Austin, Texas.

And they want to expand to more states by the end of the year. So while the California permit is limited, their overall robo-taxi plans are moving forward quickly. Absolutely. Autonomous ride hailing is coming and it's going to change how we get around cities. All right. Let's talk about Adobe. They had some interesting AI announcements too. Yeah. At their Adobe Summit, they showed off a bunch of AI tools for businesses. Like what? One big thing is Agent Orchestrator.

It's part of their Adobe Experience platform. It lets businesses build and manage AI agents from Adobe and other companies. These agents can do all sorts of things like create content, analyze data, and interact with customers. Automating business processes. Exactly, making things more efficient and personalized. They also announced Brand Concierge. It's an AI agent that creates custom brand websites. They adapt to how visitors interact with them, providing personalized content and recommendations.

So Adobe's all about using AI to make customer interactions better. Exactly. More efficient, more personalized. It's a big shift in how businesses think about customer experience. Now, it seems like voice integration is becoming a big thing in AI language models. Yeah. Anthropic, they're the ones behind some really advanced language models. They're testing voice features. What's the advantage of adding voice? It makes...

Interacting with AI feel more natural, more like a conversation. It's also good for accessibility. People who have trouble with text based interfaces can use voice instead. And it opens up a lot of new possibilities for conversational AI. Making AI feel more human. OK, here's another interesting one. AI is making its way into the legal system. Yeah, the Arizona Supreme Court is using AI powered avatars to help people access legal information.

AI avatars giving legal advice. Yeah. That's pretty wild. What can they actually do? They can give real-time legal guidance, explain court procedures, and answer common questions about things like filing documents. So it's like having a virtual legal assistant. Kind of, yeah. It's great for people who can't afford a lawyer. But of course, it's super important that these AI systems are accurate and transparent. We don't want them giving out bad legal advice. Absolutely. Accuracy is key in legal matters.

OK. March 19th was a busy day for AI. What else happened? Oh, tons of stuff. We already talked about Canvas and Gemini, but Google also launched audio overviews. Meta's Lama open source model hit a billion downloads. OpenAI's VP of research is leaving to start a new AI material science company, and OpenAI is going to invest in it.

Google unveiled TX Gemma, a collection of Gemma-based models for drug discovery. Tencent's Hunyuan released new 3D generation models. Stability AI launched stable virtual camera for making 3D videos from single images. And Anthropic-backed Graphite launched Diamond, an AI code review tool, and raised $52 million in funding.

Wow, that is a lot. AI is moving fast. It is. From new chips and robots to new applications and open source models, it's a really exciting time. So that wraps up another deep dive into AI. We covered a ton today, from chips and robots to AI art and even AI in elder care. AI is everywhere now. It's shaping our present and it's going to be a huge part of our future. Absolutely.

If you're enjoying these deep dives and want to help us keep this podcast free, consider donating. You can find donation links in the show notes. Every little bit helps. And if you're a business looking to reach a smart and engaged audience, consider advertising on the show. It's a great way to connect with thousands of professionals who are interested in AI. You can find more information about advertising in the show notes too.

And before we go, think about this. We've talked about AI reasoning, robots, AI in law, and elder care. What does this all mean for society? What kind of norms and ethical considerations will we need to address as AI becomes more powerful and more integrated into our lives? It's a question we're thinking about. Maybe do some research, explore some of the companies and technologies we've talked about. Absolutely. The conversation about AI's impact is important, and we want to hear your thoughts. Thanks for listening. Until next time.