INSANE AI Image and Video | Forbes Doxed AI Developer | MagicAnimate and Magnific.AI 🔥

Today:

INSANE AI Image and Video | Forbes Doxed AI Developer | MagicAnimate and Magnific.AI 🔥

There's this AI that's off the hook in revamping old video games, making them look brand new without losing their original charm. But, it's got some quirks – like randomly stripping people in the crowd or hiding creepy severed heads in images.

Forbes' exposé on Gil has stirred the pot. Some see it as revealing the man behind the curtain, while others feel it's a privacy breach. The fallout's just beginning, and it's shaping up to be a real doozy in the AI world. 

WATCH THE VIDEO ON YOUTUBE

Elon Musk is looking to raise $1 billion for xAI

Elon Musk is now aiming to bag a cool $1 billion for his new AI gig, xAI. He's already pocketed $135 million from some investors, but he's still short by about $865 million. Musk's plan is to roll out Grok, xAI's own AI chatbot, which promises to be edgier than the likes of ChatGPT or Google's Bard, handling the kind of bold questions others shy away from. 

Grok is still a newbie, just two months in the making, but it's gearing up for a beta launch exclusively for X Premium+ folks. One of its snazzy features? It's supposed to stay up-to-date with the latest scoop on X, but the catch is it's gotta be sharp enough to tell real news from fake.

Musk, who helped kickstart OpenAI, split from them in 2018. He's been throwing some shade at them, especially after some drama with their CEO, Sam Altman. Raising another $865 million ain't gonna be easy, but Musk's got the chops for it. The only catch? Investors might be jittery about his recent stumbles with X.

READ THE ARTICLE ON TECHCRUNCH

AI Alliance Launches as an International Community of Leading Technology Developers, Researchers, and Adopters Collaborating Together to Advance Open, Safe, Responsible AI

The AI Alliance is this big new team-up of folks from all over – companies, universities, and government agencies. They're all about making sure AI grows in a way that's good for everyone and doesn't step on any toes.

Big names like IBM and Meta are running the show. Then there's a whole bunch of others, like AMD, NASA, universities from Cornell to Tokyo, and some AI-focused groups like Hugging Face and MLCommons. These guys are the brains behind stuff like AI tools, benchmarks, and ethics in AI.

Their game plan? Create working groups that focus on all things AI – from how it's made to how it's used and governed. They also want to make sure AI stays open and helpful to everyone, not just a few big players.

READ THE ARTICLE ON META

Bing’s GPT-4-powered Deep Search takes its time with AI questions

Microsoft cooked up a feature called Deep Search for Bing, and it's like having a mind reader for your online searches. This baby is powered by GPT-4, so it's pretty smart. You throw a question at Bing, and it digs deeper, expanding your question to cover all bases.

For example, say you're curious about loyalty card programs in Japan. Bing doesn't just give you the basic gist. It goes full Sherlock Holmes, asking about different types of loyalty cards, their perks, and even how they stack up against other payment methods. It might even throw in some stuff about immigration policies and public transport fares, all to really get what you're after.

The cool part? All these expanded topics show up on your search results page. You pick the one that hits the spot, and Bing narrows down the search, focusing on the most relevant stuff. But patience is key here – this deep dive can take up to 30 seconds. If you're in a rush, the standard Bing search is still there for you.

READ THE ARTICLE ON THE VERGE

AMD To Launch AI Chip Wednesday That Could Heat Up Competition With Nvidia

AMD's gearing up to drop a new AI chip this Wednesday, looking to give Nvidia a run for its money. They're calling it the Instinct MI300 and it's all about beefing up AI systems. Word is, they might even throw some shade at Nvidia's H100 chip during their "Advancing AI" shindig.

Analysts are buzzing, saying this new piece of tech could help AMD get a bigger piece of the AI pie. They've already got big plans, eyeing a cool $2 billion in sales by 2024. Nvidia, on the other hand, is playing it cool with their new H200 chip, but they've hit a snag and pushed its launch to 2024. Meanwhile, Microsoft's in the mix too, planning to use AMD's new chip in their Azure cloud computing deal.

READ THE ARTICLE ON INVESTOPEDIA

AI Revolutionizes Neuron Tracking in Moving Animals

Scientists have whipped up a smart AI method to keep tabs on brain cells in critters that don't hold still or keep their shape, like wiggly worms. It's a big deal for brain science. This AI, a type of whiz-bang tech called a convolutional neural network (CNN), is like a shortcut for spotting and tracking neurons in animals, especially those tricky ones that keep squirming around.

This AI is trained to do a lot of the grunt work by itself. It's like teaching a computer to pick out the brain bits in images, without someone having to point them out every time. That's a huge time-saver and a big leap forward.

They tried this out on a roundworm, Caenorhabditis elegans, which is pretty much a neuroscience superstar because of its 302 neurons. The AI showed that these neurons can have some pretty complex dances, changing up their moves when different stuff happens, like smelling different scents.

And hey, they're not keeping this AI to themselves. They've made it user-friendly and up for grabs, so other brain explorers can jump on this and see what's up in their own critters.

READ THE ARTICLE ON NEUROSCIENCE NEWS

Boston-based Dotmatics helping scientists make better drugs in less time using AI

Dotmatics, a company out of Boston, is shaking things up in the drug-making world. Usually, getting a new drug out there takes a decade and costs a boatload - like $2.5 billion. But Dotmatics is using AI to speed up the process. They've got this software that lets scientists put all their research in one spot and then slap some AI on it to find patterns and predict the next moves.

Stephen Tharp from Dotmatics explains it like giving scientists a super flashlight to scan loads of data and focus on the good stuff. The big game-changer could be in biologics, where they use proteins to fight diseases. AI's helping to create protein structures that might tackle tough illnesses like Alzheimer's and cancer.

Right now, there are about 15-20 drugs out there discovered with AI's help, but in 5-10 years, that number's expected to skyrocket. This AI magic could mean big breakthroughs in less time than we ever thought possible. For more scoop, check out Dotmatics.com.

READ THE ARTICLE ON CBS NEWS

A New Trick Uses AI To Jailbreak AI Models—Including GPT-4

A group called Robust Intelligence, teamed up with Yale folks, has figured out a way to mess with big AI language models like GPT-4. They've come up with these "adversarial" AI tricks that can make these models go haywire.

Even though OpenAI got a heads-up about this, they haven't said much back yet. Yaron Singer, the big cheese at Robust Intelligence and a computer science prof at Harvard, is saying there's a big safety issue with these models that we're not tackling.

This new method uses extra AI systems to come up with and check prompts to find ways to crack the model. It's not the first time this has happened, and it's making some experts, like Zico Kolter from Carnegie Mellon, worried about how easy it is to break these models.

Robust Intelligence showed some ways to sneak past these checks. They even found ways to make phishing messages and hide on government networks. Another group, led by Eric Wong at UPenn, came up with a similar method, but the Robust Intelligence team's approach is more efficient.

READ THE ARTICLE ON WIRED

Playground v2: A new leap in creativity

Playground's rolling out a new version of their graphics model, Playground v2, and it's up for grabs. You can give it a whirl at playground.com or snag it from HuggingFace. The cool part? You can even use it for business stuff.

People seem to like Playground v2 way more than Stable Diffusion XL, like 2.5 times more. They ran a test with a bunch of prompts and thousands of people, and Playground v2 won the popularity contest.

They've cooked up a new way to see if the images are top-notch, called MJHQ-30K. It checks out the artistic vibe of images using a fancy dataset with 10 types of pics, each with 3,000 samples. They're all about quality and making sure the image matches what you're talking about.

They're also sharing some starter weights to help researchers, especially where computer power is a bit tight. You can get these in 256px and 512px versions on HuggingFace.

READ THE ARTICLE ON PLAYGROUND

Meta's top AI scientist reportedly warned Mark Zuckerberg that Facebook and Instagram could go extinct if they didn't catch up with ChatGPT

Meta's top AI brain, Yann LeCun, gave Mark Zuckerberg a heads-up: if they don't step up their game and compete with ChatGPT, Facebook and Instagram might bite the dust. This chat went down while they were in line for lunch at Meta HQ.

This chat happened at Meta's HQ while they were grabbing lunch. Zuckerberg didn't say much then, but by dinner, he seemed to get it and agreed with LeCun.

Meta was already using AI in their apps, but Zuckerberg decided to fast-track their own AI model, named LLaMA. But things got messy quick. Right after they released LLaMA to researchers, its code leaked online. This sparked some serious security concerns and even got some senators worried about the risks.

Despite the hiccups, Meta's not backing down. They kept pushing AI, releasing a new version of LLaMA and some fancy Ray-Ban smart glasses. Plus, they've got a bunch of AI chatbots mimicking celebs like Paris Hilton and Snoop Dogg. It's a wild ride in the AI world, for sure.

READ THE ARTICLE ON BUSINESS INSIDER