Eureka ChatGPT designs robots

Today:

Eureka ChatGPT designs robots 🤯 Open Source project that uses GPT-4 to train robots in simulations

Robotic hands usually ain't got the moves, but this new tech is changing the game. The same brainiacs who created Voyager, that Minecraft AI that keeps getting smarter, used GPT-4 to power it, and now they’re applying that magic to robot hands.

The new project, dubbed Eureka, is like Voyager on steroids, but for robotic dexterity. These folks, now with Nvidia, are pushing the boundaries of what AI can do, teaching robot hands to do things that'll blow your mind, like spinning a pen just right.

WATCH THE VIDEO ON YOUTUBE

Inside Apple’s Big Plan to Bring Generative AI to All Its Devices

Apple kinda snoozed when big AI trends like ChatGPT hit the scene. Now they're hustling to catch up. They didn’t hop on the AI train fast enough and were left in the dust by companies like Google and Microsoft. Apple's only recent AI move was a better auto-correct on their phones.

But don’t count them out! They're pouring big bucks (about a billion a year!) into getting their AI game strong. They’ve got their big shots, John Giannandrea and Craig Federighi, heading this mission. They're tweaking Siri to make her smarter, and they're packing the next iPhone software with AI features. They also want to help app creators by putting AI in their development tools.

READ THE ARTICLE ON BLOOMBERG

Spotify’s AI Chief Wants to Remix Our Relationship With Discovery

Ziad Sultan, Spotify's big cheese for AI, is shaking up how we find new tunes and podcasts. Back in the early 2000s, when he was just a teenager fresh from Beirut, he dove headfirst into AI at MIT, even though it wasn't the hot topic it is today.

Fast forward nearly 20 years, and he's at the helm of Spotify's personalization efforts, using AI to give us tailor-made playlists, chatty AI DJs, and podcasts in different languages. This guy saw the potential of AI way before it was cool, and now he's using it to change the game at Spotify.

READ THE ARTICLE ON THE INFORMATION

How should creators be compensated for their work training AI models?

AI companies are using artists' work to train their software. But how do you pay these artists? In the future, many creatives might use AI to help with their work, but right now, it's their art that's helping AI companies get better. This has led to a big question: How should they get paid?

Big companies like Adobe and Canva are coming up with ways to pay artists when they use their art to train AI systems. This helps them avoid legal problems, since some companies have been sued for using art without permission. Plus, for these AI tools to keep improving, they'll need lots of good-quality art to learn from.

READ THE ARTICLE ON QUARTZ

Google Pixel’s face-altering photo tool sparks AI manipulation debate

Smartphones these days are quick to edit photos, enhancing colors and lighting. But Google's newest smartphones, Pixel 8 and Pixel 8 Pro, have upped the game. Using AI, these phones can swap people's expressions in photos. Say someone didn't smile in a group pic? These phones can grab a smiling face from another photo and pop it in. Google named this feature "Best Take". Plus, users can now erase or shift things around in a photo, like people or buildings. This magic touch-up is possible with a tool called "Magic Editor".

Not everyone's thrilled. Some tech critics call it "icky" or "creepy", concerned about trust issues online. A pro photographer, Andrew Pearsall, warns this could lead us into murky waters. But Google's Isaac Reynolds defended the tech, saying these tools aren't about faking, but making the best memory. For him, the final image becomes a "blend of moments".

READ THE ARTICLE ON BBC

ChatGPT will ‘help us with extremely hard problems’ in the future, says OpenAI’s chief tech officer

Mira Murati from OpenAI spilled the beans on ChatGPT's future at the Atlantic Festival, saying it's gonna be a game-changer. Back in 2018, when she first joined OpenAI, she wasn't sure this chatbot was going to make waves. Fast forward a few years, and ChatGPT is helping folks left and right, in personal and work stuff.

Murati says we're on the brink of chatting with ChatGPT like we do with our buddies, no keyboard needed. She dreams of high-bandwidth, natural conversations with AI, maybe even sorting business cards with a snap of a photo. Jony Ives might be cooking up some fancy hardware for this, aiming to make "the iPhone of artificial intelligence."

Next up, ChatGPT's gotta start thinking deep. Right now, it's great for a back-and-forth, but Murati wants it to cook up new ideas, even solve crazy hard problems. She's aiming high, wanting AI's answers to be even sharper than Wikipedia.

READ THE ARTICLE ON CNBC

Roche is using AI to find the hardest-to-find lung cancer patients, with a potential blockbuster drug on the line

Roche's lung cancer drug, Alecensa, outdid standard treatments big time in a recent study. The kicker? It works best for folks with a rare gene error found in only about 5% of lung cancer patients. These patients are mostly younger, probably haven't smoked much, and often slip through the cracks without an early diagnosis.

Roche's teaming up with Israeli tech firm Medial EarlySign to use AI and figure out the best time for CT scans. The goal? Spot these specific patients and their tumors early on, ideally before the cancer spreads. Roche plans to pitch the Alecensa study to the big decision-makers for a thumbs-up. Even though Alecensa's for a small patient group, it's still expected to rake in a cool $1.75 billion this year. Charlie Fuchs from Roche says it's proof that a drug doesn't need a massive audience to be a game-changer, both scientifically and cash-wise.

READ THE ARTICLE ON FORTUNE

AI and 10 seconds of voice can screen for diabetes, new study reveals

Klick Labs did this cool study where folks just talked into their phones for a bit, and some smart tech figured out who might have type 2 diabetes. Turns out, this method is right about 89% of the time for ladies and 86% for gents.

They had 267 people, some with diabetes and some without, chat into their smartphones a couple times a day for two weeks. After listening to a whopping 18,000+ recordings, the tech squad noticed that voices of people with diabetes kinda sounded different. But not in a way we'd notice – they used some fancy science stuff to catch it.

This voice check thing might be a game-changer. Normally, folks have to haul themselves to a doc to get tested. But this? Just chat into your phone and boom! Plus, there are tons of people walking around not even knowing they have diabetes. This tech might help catch them.

READ THE ARTICLE ON MEDICAL XPRESS

American Federation of Teachers partners with AI identification platform, GPTZero

The big teacher union, American Federation of Teachers, teamed up with a company called GPTZero. Why? To catch students who use robots to do their homework. But it ain’t all about catching cheaters. The union's head, Randi Weingarten, says AI can be good for classrooms if used right. But they gotta make sure it's safe and respects privacy. She compares it to riding a wave: you can’t stop tech from coming, but you can learn how to ride it.

GPTZero ain’t just about catching robot-made work, though. They got tools to help teachers and students work together. They even have a feature for students to swear their work's all human, or to ‘fess up if they used AI. Tian, the dude who started GPTZero, thinks AI ain’t the enemy. He wants to help everyone get the best from it without causing problems. And Weingarten? She thinks AI can help teachers with stuff like paperwork and making lesson plans. But she says they gotta be real about it, not expecting it to fix everything.

READ THE ARTICLE ON CBS NEWS