Ilya Sutskever on AI mental models | Hearing AI Voices | Visualizing Neural Nets | GNOME makes mats

Today:

Ilya Sutskever on AI mental models | Hearing AI Voices | Visualizing Neural Nets | GNOME makes mats

This video is a quick tour of various cool AI developments. First, it shows an AI interpreting actions in real-time, demonstrating its potential in understanding and reacting like a human. Then, it dives into large language models like GPT-3, offering a peek into their complex neural networks with detailed visualizations. It highlights how these models, though seemingly just predicting the next word, actually build a sophisticated understanding of the world.

Augmented reality gets a spotlight too, with an impressive demonstration of realistic, non-existent objects seamlessly integrated into a real environment. It showcases an autonomous robot synthesizing new materials, a breakthrough in AI and robotics collaboration, leading to advances in various fields like semiconductors and medicines.

WATCH THE VIDEO ON YOUTUBE

Snapchat+ subscribers can now create and send AI-generated images

Snapchat+ just got cooler for its 7 million subscribers! Now, they can create and send their own AI-generated images. Think of fun stuff like "a cheese planet" or "beach day," and Snapchat turns it into a picture. Plus, they added a sweet update to their Dream selfie feature – you can now include your pals in these funky AI selfies. And if you've ever taken a too-close pic of your dog and wished for more background, their new AI tool can fix that by zooming out and filling in the details. 

These fresh features are rolling out, but availability might differ by region. Snapchat's really upping its AI game, and it's paying off – they made a cool $20 million in November from Snapchat+ alone, which costs just $3.99 a month.

READ THE ARTICLE ON TECHCRUNCH

Meta’s AI for Ray-Ban smart glasses can identify objects and translate languages

Meta's teaming up with Ray-Ban to test some cool AI features on their smart glasses. They're trying out this tech that lets you know about stuff you see or hear, like identifying objects or translating languages. Mark Zuckerberg showed how it works on Instagram, like picking out pants that go well with a shirt. The AI can even translate text and come up with captions for images. 

boztank

View more on Instagram

Zuckerberg mentioned this back in September, saying it's like having a helpful buddy you can chat with all day. Andrew Bosworth, the big tech guy at Meta, also demonstrated how it can describe things and help with photo captions and translations. This test is just starting in the U.S., and only a few folks who sign up can try it out.

READ THE ARTICLE ON THE VERGE

Essential AI Raises $56.5M Series A to Build the Enterprise Brain

Essential AI, a San Fran startup, just bagged a whopping $56.5 million in a Series A funding round, led by March Capital. This cash infusion, along with an earlier $8.3 million seed round, brings their total haul to nearly $65 million. Big names like Google, NVIDIA, and AMD are in on it too. 

The company, founded this year by ex-Google pals Ashish Vaswani and Niki Parmar, is all about making computers and humans work better together. They're cooking up AI products to streamline boring tasks and boost productivity. With this cash, they're set to shake up how businesses use AI, pushing the boundaries of what's possible with their tech know-how and a strong team of pros.

READ THE ARTICLE ON BUSINESSWIRE

A new old kind of R&D lab

Jeremy Howard and Eric Ries just kicked off Answer.AI, a fresh AI R&D lab that's all about turning big AI research breakthroughs into real-world stuff people can use. They got a sweet $10 million from Decibel VC to make it happen. The team's gonna be a bunch of top-notch tech whizzes working remotely, picked for their skills, not their resumes.

It's kinda like what Thomas Edison did with his "Invention Lab" way back, making electricity useful for everyday folks. Answer.AI's aiming to do the same with AI – figure out the hardcore science stuff and then make it something everyone can use.

Howard's a big name in AI with fast.ai and all, and he's bringing his A-game in making AI easy to get for everyone. Ries, the other head, is the brains behind Lean Startup, focusing on how to turn a bright idea into a killer product. Their goal? Use AI to cook up new products that really matter.

READ THE ARTICLE ON FAST.AI

Meet NANA, Moonshine Studio’s AI-Powered Receptionist Avatar

Moonshine Studio, a VFX company focused on animation, faced a dilemma: their receptionists in Taiwan were too busy greeting guests to do other important work. Eric Chiang, a senior 3D artist, tackled this issue by creating NANA, a realistic AI-powered virtual assistant. 

NANA can welcome guests and share basic company info, easing the receptionists' workload. Chiang used his NVIDIA Studio-badged PC with a GeForce RTX 4090 graphics card, leveraging its GPU-accelerated features for this project. His workflow was enhanced by the GPU's Tensor Cores, boosting AI tasks and overall work quality. The project also highlights NVIDIA's December Studio Driver and various creative contests. Chiang sees AI as a game-changer for creators and society.

READ THE ARTICLE ON NVIDIA

Relevance AI’s low-code platform enables businesses to build AI teams

Relevance AI, an Aussie startup, is making waves in the tech world. They've designed a low-code platform so companies of all sizes can create their own AI teams, like having a digital co-worker. This platform is a game-changer, especially for smaller firms that don't have big tech teams. 

They've just bagged $10 million in funding to grow this cool tool. Already, around 6,000 businesses are on board, using it for stuff like customer service and market research. It's all about making repetitive tasks a breeze and letting teams focus on the big picture. 

The founders, Daniel Vassilev and his buddies, started this in 2020 and now they're planning to beef up their team and set up shop in San Francisco. They're betting big on AI, predicting that by 2025, every team will have an AI buddy, and by 2030, full AI squads.

READ THE ARTICLE ON TECHCRUNCH

Limewire Is Back, But AI’s Making The Music This Time

LimeWire, a former file-sharing giant sued for copyright issues, is making a comeback with a twist. Now, they're diving into AI music creation. Their new tool lets users generate AI-made tracks simply by describing what they want or uploading images for inspiration. 

This new LimeWire is cautious about legal troubles, using only licensed music for training their AI. The brand, bought last year for its infamous name, initially eyed NFTs but shifted to AI. Their AI music studio, while innovative, still struggles to match human creativity. The company aims to enhance its capabilities, including a text-to-voice feature, aspiring to make AI music creation accessible to everyone.

READ THE ARTICLE ON ROLLING STONE

Startup Armada Is Bringing AI To Remote Places, Using SpaceX Starlink Satellites

Armada, led by former DataRobot CEO Dan Wright, is shaking things up by bringing AI to remote places like oil rigs and mines, using SpaceX's Starlink satellites. It's a game-changer, especially in areas where data's been collecting dust. They've raised $55 million, hitting a near $250 million valuation. 

Their big idea? A tech platform, Commander, that links remote sites to the internet using Starlink. Plus, they've got this cool mobile data center, the Galleon, that's packed with powerful GPUs for crunching AI models. Despite no customers yet, investors are pumped, betting big on Armada's potential in energy, manufacturing, mining, and defense. Wright's making a comeback with Armada, aiming to make a mark by leveraging AI in places where it's never been before.

READ THE ARTICLE ON FORBES

Google’s New AI, Gemini, Beats ChatGPT In 30 Of 32 Test Categories

Google's new AI, Gemini, is making waves, outdoing ChatGPT in almost all test categories. It's more than just a chatbot - this thing handles text, images, videos, and audio. It's like the Swiss Army knife of AI, especially with its top-notch version, Gemini Ultra. This beast even beats human experts in a tough test covering a wide range of subjects.

But here's the catch - Gemini and other AI tools are stirring up some legal drama. They use loads of material from writers, artists, and thinkers without asking or paying them. Think about it like using someone else's work to build your own project, without giving them credit or a slice of the pie. Now, big names like John Grisham are taking a stand, suing these AI companies for what they call "mass scale theft." 

READ THE ARTICLE ON FORBES

Deploy Mixtral 8x7B on Amazon SageMaker

Mixtral 8x7B, a new AI model from Mistral AI, is now easy to set up on Amazon SageMaker. This model is like a team of 8 expert models rolled into one. 

First, you set up your coding environment with some Amazon tech and Python. Then, you grab the special container Amazon made for large models like Mixtral. Next, the article talks about the hardware you need – basically, a beefy computer setup since this model's pretty big. After that, it walks you through deploying Mixtral on SageMaker, which means setting it up on Amazon's servers. 

Once it's up, you can actually chat with the model and see how it responds. The blog wraps up by showing how to build a demo app for a better user experience and reminds you to clean up your digital workspace when you're done. If you're stuck or curious, the writer's open for a chat on social media.

READ THE ARTICLE ON PHIL SCHMID

Durable cements $14M to build bots and other AI tools for small businesses in service industries

Durable, a Canadian startup, just bagged $14 million to boost its AI tools for small businesses like builders, bakers, and personal trainers. These tools, including an AI website builder, help those with little to no online presence. In just a year, Durable's tech has crafted over 6 million websites. The funds will expand their services, like an AI bot for business advice and an upcoming proactive assistant that'll manage tasks automatically. 

This Series A funding, though not the largest, includes notable investors like Spark Capital and Altman Capital. Durable's goal? To empower small business owners with affordable, cutting-edge AI solutions, simplifying tasks like branding and scheduling. Their use of AI aims to bring small-scale businesses up to speed in the digital age.

READ THE ARTICLE ON TECHCRUNCH

The US has a new plan for wielding AI to fight climate change

The U.S. is ramping up its AI game to tackle climate change. AI's got a bunch of uses, like making better batteries, keeping an eye on the power grid, studying weather patterns, and checking out emissions. But, it's not all smooth sailing – there's stuff like data leaks, privacy issues, and it chews through a ton of power. The Department of Energy (DOE) is on top of this, dishing out grants to AI startups, using super-fast computers, and tapping into the smarts at national labs.

Fu, the new head, is lining up all these pieces. She's aiming to get more folks involved, like agencies, scientists, and entrepreneurs. Her to-do list includes using AI for nuclear fusion power, making supercomputers more energy-efficient, protecting AI from hackers, and sharing tons of DOE data with other scientists and AI companies.

READ THE ARTICLE ON SEMAFOR

Federal watchdog finds more than 1,000 ways government could use AI

The Government Accountability Office (GAO) dropped a report saying the U.S. government could use artificial intelligence (AI) in over 1,200 ways, and 200 are already in play. The report's like a heads-up to handle AI smartly to dodge risks and unintended mess-ups. It's for Congress to keep an eye on AI use. 

Most AI action is in science and internal stuff, with NASA leading big time with 390 uses. Surprisingly, Housing, Nuclear Regulatory, and Small Business folks reported zero AI plans. Already, 282 AI applications are working, like NASA’s volcano watch and Commerce's fire prediction. 

The GAO checked out 23 federal agencies plus two big offices, focusing on how they're meeting AI-related rules set by past and current presidents. Ten agencies aced the compliance test, 12 are getting there, and one's off the hook. 

READ THE ARTICLE ON THE HILL

A peek inside the black box that is OpenAI’s finances.

OpenAI, known for ChatGPT, had its non-profit parent's 2022 finances made public. The big news? Sam Altman, the big boss, got fired last month. The report shows Altman earned about $73K in 2022. Co-founders Greg Brockman and Ilya Sutskever bagged $113K and $335K, respectively. 

But here's the kicker: the three board members who helped give Altman the boot didn't earn a dime. This report doesn't include the dough from OpenAI's commercial side, so we're not getting the full money story.

READ THE ARTICLE ON THE VERGE