Tesla Bot Optimus Gen 2 is Stunning, but it's just the beginning... | NVIDIA Eureka and Google RT-X

Today:

Tesla Bot Optimus Gen 2 is Stunning, but it's just the beginning... | NVIDIA Eureka and Google RT-X

Tesla's latest buzz is about their new humanoid robot, Optimus. No tricks, the video's real. Robotics ain't what it used to be – we're moving from simple, one-trick bots to ones with brains, kinda like Google's RT2, now the RTX. 

This shift is massive. Forget the slow robot news of the past. We're entering a world where robots are common, thanks to AI brains like GPT-4. These robots think and learn, outsmarting human programming. 

WATCH THE VIDEO ON YOUTUBE

Gemini API and more new AI tools for developers and enterprises

Last week, Google dropped Gemini, their smartest AI yet, and gave everyone a sneak peek at what it can do, including a test run in Bard. Now, they're rolling out Gemini Pro for the tech crowd and big companies, alongside a bunch of other cool AI tools. Here's the lowdown:

• Gemini Pro's up for grabs through the Gemini API for the coding wizards in Google AI Studio and big-time players using Google Cloud's Vertex AI.

• They're also throwing in other models in Vertex AI, like an upgraded Imagen 2 for awesome image creation, and MedLM, specially designed for healthcare, but only for U.S. Google Cloud users right now.

• Plus, Duet AI is now ready for both developers and security teams.

Developers can mess around with Gemini Pro through AI Studio for free, then level up to Vertex AI for the big league features. Imagen 2's got a facelift, now nailing even the tricky stuff in image creation. MedLM's setting up shop in healthcare AI, and Duet AI's making life easier for coders and security pros. Google's stoked to see what folks will create with Gemini, and they're hinting there's more cool stuff on the way.

READ THE ARTICLE ON GOOGLE

Google Cloud partners with Mistral AI on generative language models

Google Cloud is teaming up with Mistral AI, a Paris-based startup, to host its AI language tools on Google's tech. Mistral AI, created by ex-Google and Meta pros, recently bagged a huge $415 million investment. They'll use Google's fancy AI gear and benefit from its top-notch security. 

This partnership means Mistral AI can grow its big AI models that spit out text and stuff, using tons of data. Basically, it's a big win for both, with Google providing the muscle and Mistral bringing its brainy AI ideas to the table.

READ THE ARTICLE ON CNA

Microsoft Phi-2

Microsoft's Phi-2 is a big deal in the AI world. It's a Transformer model with a hefty 2.7 billion parameters. It's been trained with a mix of the same stuff as its older sibling, Phi-1.5, plus some new synthetic texts and clean websites. When put to the test, Phi-2 almost nailed it in areas like common sense, language understanding, and logic, especially for its size.

It hasn't been fine-tuned with human feedback. The whole point of Phi-2 is to help researchers tackle big issues like cutting down toxicity, understanding biases, improving control, and more. It's not for just anyone to use in their day-to-day projects; it's more like a tool for deep research.

Phi-2 works best with question-and-answer formats, chat scenarios, and coding problems. As for the technical side, it's a beast. It was trained on a massive dataset using high-end GPUs for two weeks straight. 

READ THE ARTICLE ON HUGGING FACE

Worldcoin is bringing its eyeball-scanning tech to Reddit, Discord, and... Minecraft?

World ID 2.0 is like a digital passport for the internet, letting folks prove they're real people while keeping their personal stuff private. It's getting a big update with new features like World ID Apps, where you can link your online accounts, like Reddit or Shopify, and use your World ID as a login. They're also rolling out three levels of security, including face authentication for extra important actions.

Over 2.6 million people globally are already using it, and it's especially popular in places like Chile, Argentina, and Portugal. This upgrade makes it faster and easier for developers to create apps, and it's all about keeping your identity safe and under your control. You can reset your ID at an Orb station if you lose it, and everything's open source. The roll-out starts now and will keep going into early 2024. 

READ THE ARTICLE ON WORLDCOIN

Google Ventures adds general partner to back AI, open source startups

Google Ventures, now called GV and backed by Google's parent company, just hired Michael McBride as a new big cheese. McBride, who used to work at GitLab, a company that creates tools for software developers and is also part of GV's investments, is all set to focus on startups that are into open-source and artificial intelligence. 

He's got his eye on young companies aiming to sell to big businesses. Despite the money world being a bit shaky these days, GV's still going strong, investing in lots of companies, including some like Uber and Slack. They even bought stocks in some of their own companies that have gone public. GV's got a big pot of money, around $8 billion, from Alphabet, and their team is split between life sciences and tech stuff, covering a wide range of areas in North America and Europe.

READ THE ARTICLE ON REUTERS

OpenAI Startup Fund launches second Converge startup cohort

Despite some buzz around OpenAI, their startup program is moving forward without a hitch. The OpenAI Startup Fund, which is its own thing and not directly a part of OpenAI, just kicked off Converge-2. This is their second round of a six-week program supporting top-notch engineers and creators who are using AI to shake things up. 

They're picking 10-15 startups for this cohort, each getting a cool $1 million from a pot that's over $100 million, partly filled by Microsoft. That's at least $10 million going into Converge-2. The chosen ones also get to join tech talks, social events, and rub elbows with big names in the field. They don't have to use OpenAI's tech, but it's likely they'll find it super appealing. The program is open to founders of all stripes, from anywhere, and you don't need to be an AI whiz to apply. But, if you get in, you gotta commit 4-6 hours weekly from March 11 to April 19, with a couple of trips to San Francisco on the house. 

READ THE ARTICLE ON TECHCRUNCH

Announcing Our Latest Open Source AI Grants

Andreesen Horowitz just dropped news on their latest open source AI grants. They're backing the second crew of folks who are deep into crafting the nuts and bolts of modern AI. Unlike throwing cash for a stake in the game, these grants are all about supporting the little guys - developers and small teams. 

This round, they're big on tools for training, hosting, and checking out language models, plus all the jazz around visual AI. It's a neat follow-up to their first round, which was all about making language models sharper. 

READ THE ARTICLE ON a16z

Instagram now lets you make AI-generated backdrops for your Stories

Instagram's got a cool new feature: AI-generated backdrops for your Stories. It's pretty neat, and the backgrounds have a distinct AI art look. You can spice up your pics by just tapping a few times and giving a prompt like 'chased by dinosaurs' or 'surrounded by puppies.' It totally changes the vibe of your image. To try it, hit the backdrop button when you're making a new Story. Plus, when you share your AI-backed image, a "Try it" sticker pops up, so your buddies can easily give it a whirl too.

READ THE ARTICLE ON THE VERGE

AI Helps Chipmakers Design the Very Processors That Speed Up AI

Intel's getting crafty with AI, using it to design their new Meteor Lake processors. This tech's not just for boosting laptop smarts, but also for making the chips faster and more efficiently. Intel's cooking up these processors, their first with multiple "chiplets" in one package, by letting AI catch bugs early and manage manufacturing, which means fewer chips end up in the trash. 

This AI edge means they can sell more chips, a big deal for Intel as they hustle to beat Taiwan's TSMC and beef up the US chip game. For us regular folks, this could mean better, cheaper laptops. Plus, processors are super complex now, loaded with billions of tiny switches. AI helps sort this mess, a game-changer in the chip design world. Intel's mix of in-house and external AI tools is helping them figure out the best way to sell these new chips, ensuring they get the most bang for their buck.

READ THE ARTICLE ON CNET

Mapping the micro and macro of biology with spatial omics and AI

Spatial omics and AI are cracking the code of human biology. Imagine understanding every one of the 37 trillion cells in our body and how they mesh together - it's like the ultimate biology puzzle. Spatial omics is letting us peek into cells in their natural habitat, tissues, and suss out what's happening at a crazy detailed level. 

We're talking huge leaps in understanding diseases like cancer and Alzheimer's. Companies like Owkin are spearheading this, especially with their massive cancer project, MOSAIC. They're not just collecting data but using AI to dig deep, finding new ways to tackle diseases. It's a whole new world for biology, with spatial omics leading the way, making us rethink everything from biomarkers to drug development. In a few years, we'll map every bit of our body's cells and molecules, and that's just the beginning. 

READ THE ARTICLE ON MIT

Google is rolling out new AI models for health care. Here’s how doctors are using them

Google's getting into health care big time with their new AI, MedLM. It's this set of fancy AI models made just for doctors and researchers. They're all about making tough medical studies easier and jotting down what goes down between docs and patients. 

MedLM's got a couple of different models. One's big and brainy for heavy-duty tasks, and the other's more nimble for on-the-fly stuff. It's all based on Med-PaLM 2, Google's medical language whiz they showed off in March. If you're a Google Cloud user in the U.S., you can get your hands on it now. Price tag varies, but the smaller model's cheaper to run.

Google's also planning to beef it up with Gemini, their latest and greatest AI. The idea is to not just have one tool for everything but a whole toolbox, each tool good for different medical tasks.

Real-world, this stuff's already in action. Big health groups like HCA Healthcare are testing it in ERs to cut down on paperwork. Google's AI can take a transcript of a doc-patient chat and turn it into a neat report. This could save docs a ton of time on admin stuff, which is a big deal.

On the future front, companies like BenchSci and Deloitte are already eyeing MedLM for things like drug discovery and patient care. But again, they're testing the waters first, making sure it does what it's supposed to.

READ THE ARTICLE ON CNBC