Mistral 7B | Raw Open Source Power Dethroning OpenAI

Today:

Mistral 7B | Raw Open Source Power Dethroning OpenAI

Mistral dropped their own AI model, but they did it old school – through peer-to-peer sharing, like how we used to snag music and games back in the day. This Mistral model is a big deal because it shakes up the AI scene, no longer just OpenAI's playground. It's open-source, meaning it's free and customizable.

This model uses a cool trick called a 'mixture of experts' – it's like having a bunch of specialist brains in one. People are saying it's already nipping at the heels of older models like GPT-3.5.  It's versatile, doing everything from code to customer service. If you're into DIY, you can run it yourself, but you'll need some serious computer muscle. 

WATCH THE VIDEO ON YOUTUBE

AFL-CIO and Microsoft announce new tech-labor partnership on AI and the future of the workforce

The AFL-CIO and Microsoft are teaming up to shape the future of work with AI. They're focusing on three things: educating workers about AI, including their input in AI tech development, and shaping worker-friendly AI policies. 

This partnership builds on earlier agreements with video game workers and sets a new standard for how tech companies and labor unions can work together. The plan includes AI education for workers, direct feedback from labor to Microsoft’s AI team, and joint efforts in policy and skills development. 

They're looking to use AI to boost, not replace, workers' roles, addressing concerns from both sides – 70% of workers worry about AI taking their jobs, while the same percentage would use AI to reduce their workload. This collaboration is a big step in making sure AI development considers workers' voices and needs.

READ THE ARTICLE ON MICROSOFT

Audiobox Responsible Generation Grant

Meta's FAIR team introduced Audiobox, a cool new tool for making custom voices and sound effects easily. It's super user-friendly, letting folks easily make custom voices and sound effects just by using voice inputs or typing in what they want.  Just feed it voice clips and what you want to say, and it does the magic.

Now, they're offering a sweet deal for researchers digging into AI and audio tech. If you're in the academic or research world, you could get your hands on Audiobox models and the code to play with, plus a chance to snag up to $50,000 to boost your work. This is all about pushing the boundaries in responsible AI and audio generation.

READ THE ARTICLE ON META

US in talks with Nvidia about AI chip sales to China - Raimondo

The U.S. is chatting with Nvidia about selling AI chips to China, but they can't sell the really high-end stuff. Commerce Secretary Gina Raimondo says it's cool for Nvidia to sell regular AI chips there, mainly for regular business stuff. But the U.S. doesn't want China getting their hands on the super powerful chips that could boost their tech game big time. Nvidia's CEO, Jensen Huang, is all about following the rules and just needs to know what they are.

Raimondo's also sending a heads-up to chip makers: don't try to sneak around the rules by making chips that are just a tad under the limit. The U.S. will catch on and clamp down on that too. Nvidia wants to do right and sell as many chips as they can, but it's a balancing act.

READ THE ARTICLE ON SCMP

EU agrees tough limits on police use of AI biometric surveillance

The EU's cracking down on AI-driven biometric surveillance by police. Cops can't use real-time AI surveillance without a judge's okay, except in rare cases like serious crimes or terror threats. 

This rule applies everywhere, from parks to stadiums. They've got 24 hours to get the green light, or they gotta turn off the AI and delete any data collected. This move's meant to stop "predictive policing" that could unfairly target folks based on race. 

There's a specific list of 16 serious crimes where this rule might not apply. Plus, the EU's banning other risky AI stuff, like tech that messes with people's free will or does "social scoring" like in China. They're serious about not turning into a Big Brother state.

READ THE ARTICLE ON THE GUARDIAN

A financial news site uses AI to copy competitors — wholesale

A financial news site, Investing.com, is accused of using AI to copy articles from competitors, like FXStreet. FXStreet, which has a 60-person team analyzing global currencies, noticed that Investing's stories were eerily similar to their own. 

This copying is harming FXStreet's competitive edge. Their head, Pere Monguió, sees this as a major threat to journalism and original content. While some rivals, like The Motley Fool, aren't as worried, noting the lower quality of Investing's AI-produced articles, they acknowledge that AI is now smart enough to turn good content into something average.

READ THE ARTICLE ON SEMAFLOR

Sports Illustrated publisher fires CEO after AI debacle

The Arena Group, running big names like Sports Illustrated, canned their CEO Ross Levinsohn and other top dogs after a mess-up where Sports Illustrated used fake names and AI-generated pics for some stories. Levinsohn, in charge for three years, got the boot without much explanation. He's been swapped out for Manoj Bhargava, the big shot behind 5-Hour Energy and the main investor in The Arena Group. 

This shake-up happened just after three other execs were shown the door. Despite this drama, it's business as usual at Sports Illustrated, but the company's staying mum about the whole AI article thing.

READ THE ARTICLE ON CNN

A spectral device using Generative AI could detect bad microbes in food factories in real time

A French startup, Spore.Bio, just raised €8 million for its cool new device that spots nasty germs in food factories using Generative AI. Instead of the old-school way of checking for germs with petri dishes and waiting forever for lab results, this gadget uses a special light to find the bad bugs right there and then. It's like a super-fast germ detective, comparing what it sees with what it already knows about typical factory floor microbes. 

The team behind it includes some smart folks who used to work at big companies. But hey, they're not alone in this game – there's a U.S. rival, PathogenDX, which has already scooped up $11.6 million. So, it's not going to be an easy ride for Spore.Bio.

READ THE ARTICLE ON TECHCRUNCH

AI predicts age of newly discovered supernovae within milliseconds

This article discusses a groundbreaking artificial intelligence system. This AI can instantly figure out the age of supernovae – those massive, exploding stars – just moments after telescopes spot them. This tech is a big deal for upcoming projects like the Vera C. Rubin Observatory’s Legacy Survey of Space and Time, set to start in 2025. 

This 10-year sky survey will capture a whopping 15 terabytes of data nightly, possibly identifying over 10 million cosmic events each evening. That means we could see a hundredfold increase in spotting new supernovae. In short, this AI is a game-changer in how we understand and discover cosmic events.

READ THE ARTICLE ON NEW SCIENTIST

AI made from living human brain cells performs speech recognition

Scientists at Indiana University Bloomington, led by Feng Guo, have cooked up a wild idea. They're using mini-brains, basically tiny balls of human brain cells, to do some basic speech recognition. These mini-brains, grown from stem cells, are hooked up to computers and they're learning to pick out voices from audio clips. They started off kinda iffy, getting it right just 30-40% of the time, but after a couple of days, they upped their game to 70-80%.

Regular AI guzzles energy, but these mini-brains are more like energy sippers. Plus, they're trying to get around the limits of standard computer chips. There's a catch, though. These mini-brains only last a couple of months before they check out, and they're not exactly chatting up a storm – they're just figuring out who's talking, not what's being said. 

READ THE ARTICLE ON NEW SCIENTIST

AI and EEG Transform Silent Thoughts to Text

Some smart folks at the University of Technology Sydney whipped up a cool gadget that turns your silent thoughts into text – no jabber needed. They use this snazzy cap that reads your brain waves and an AI named DeWave to do the heavy lifting. 

This gear is a game-changer because you don't need any surgery or to be stuck in a giant MRI machine. It's pretty good at understanding what you're thinking and could be a big help for folks who can't talk due to health issues. It's also got potential for controlling stuff like robotic arms. 

They tested it on 29 people and it worked pretty well, though there's room for improvement. The big brains behind it are hoping to make it as good as the usual language translation tools. This nifty invention got the spotlight at a big AI conference, and it's part of a bigger effort to connect our brains with machines.

READ THE ARTICLE ON NEUROSCIENCE NEWS

Nvidia emerges as leading investor in AI companies

Nvidia, the top dog in chipmaking, is now a big player in the AI startup game. They've thrown cash into over two dozen companies this year, big and small, in areas like healthcare and energy. They're leading the pack, even beating major Silicon Valley venture firms in AI investments. 

Nvidia's not just splashing cash for fun; they've sunk over $870 million into these deals, which is way more than last year. Their main thing? They back companies that use their tech, especially their H100 GPU, a hot item in Silicon Valley for powering advanced AI services like ChatGPT. 

While they're keen on getting good returns, they also want strategic partnerships. Some startups are betting that teaming up with Nvidia means getting the inside track on new tech. Nvidia insists they're not twisting arms for chip use, and they're playing fair with their partners. Even with the big bucks and strategic moves, they're all about keeping it friendly in the high-stakes world of AI.

READ THE ARTICLE ON FINANCIAL TIMES

MIT group releases white papers on governance of AI

MIT's brainiacs dropped a bunch of papers about keeping AI in check. They're telling U.S. bigwigs how to use old rules to manage new AI tricks. The plan? Use what we've got, like laws for risky stuff, and tweak them for AI. The main idea: figure out what AI's meant to do first, then decide how to keep it on a leash.

For example, if an AI's pretending to be a doc, that's a no-go, just like if a person did it. They're also looking at who's responsible when things go south, especially when AI's a part of bigger tech stacks. 

They're also suggesting new ways to keep an eye on AI, like audits and a special AI-watchdog agency. Plus, they're tackling tricky legal stuff, like copyright and AI doing things no human can, like mass-surveillance.

The MIT squad believes that managing AI right needs both tech smarts and social know-how. They're all for AI progress but insist on keeping it under control. It's like they're saying, "AI's cool, but let's not let it run wild."

READ THE ARTICLE ON MIT

The AI Content Tools Entrepreneurs Can’t Live Without

This article's all about how AI tools are a game-changer for entrepreneurs in content creation. Forget spending hours making content; now, there's a bunch of AI options that whip up tons of stuff in no time. It's a no-brainer to try them if you're into making content.

There's Copy.ai for generating marketing copy, Tome for slick presentations, Unriddle for research assistance, Midjourney for AI-generated art, and Pika for video creation from text prompts. These tools not only save time but also boost creativity and impact. 

They're a game changer for entrepreneurs, helping them ditch the tedious stuff and focus on the big picture. It's all about finding the right AI tools to supercharge your creative process.

READ THE ARTICLE ON FORBES

How to use ChatGPT to write code

ChatGPT, a tool by OpenAI, can be a handy coder's assistant, but it's not perfect. It's great for specific coding tasks and choosing libraries, but not for building entire apps. You need to give clear, detailed prompts for the best results. Although it can generate code, it might also spit out unusable stuff. Always verify the AI-generated code and don't expect it to handle complex or nuanced problems requiring deep experience. 

ChatGPT is better suited for breaking down big projects into manageable chunks and coding those pieces. Remember, it’s more of a conversation, refining your questions based on its responses. The tool is useful but has its limits; it's not a replacement for human coders or deep coding knowledge.

READ THE ARTICLE ON ZDNET