Train, tune, and customize Large Language Models like never before, leveraging NVIDIA’s DGX Cloud supercomputing capabilities seamlessly within the Hugging Face ecosystem


NVIDIA and Hugging Face to Connect Millions of Developers to Generative AI Supercomputing

NVIDIA and Hugging Face are teaming up to give a bunch of developers access to super powerful AI computing. This partnership will help folks build smart AI stuff, like chatbots that can chat better or systems that can search and summarize info really quickly.

NVIDIA has this thing called DGX Cloud, which is a beast of a computer system that’s super good for AI tasks. Hugging Face, on the other hand, is a platform used by loads of organizations to make and use AI models. By putting the two together, developers can use NVIDIA’s power within the Hugging Face platform. It’s like putting that race car engine into your family car, but for AI.

The two companies have a new service cooking up called Training Cluster as a Service. It’s all about making it easy for businesses to create their own AI models really fast.

For the techy stuff, the DGX Cloud has a ton of power and memory. If any issues pop up, NVIDIA experts are there to help out. And all of this is expected to be up and running in a few months.


NVIDIA AI Workbench Speeds Adoption of Custom Generative AI for World’s Enterprises

NVIDIA just dropped a new toolkit called the AI Workbench. It lets folks mess around with AI models right from their computers or fancy workstations, and then, when they’re ready, push them to big-time data centers, cloud systems, or even NVIDIA‘s own cloud.

Before this, getting started with a major AI project felt like trying to assemble a jigsaw puzzle with pieces from ten different boxes. But now, with the AI Workbench, developers can easily grab and tweak models from popular spots like Hugging Face and GitHub. Plus, they can share these models with the click of a button.

NVIDIA is also launching an updated software platform, the NVIDIA AI Enterprise 4.0. This bad boy is packed with tools that make it even easier to roll out AI, ensuring it’s safe and stable for businesses.


Google launches Project IDX, a new AI-enabled browser-based development environment

Google’s dropping a new tool called Project IDX, which is like a fancy online workspace for building web and app stuff. It works with popular coding tools and languages, and more are coming soon. Instead of making something totally new, they beefed up Visual Studio Code with their own AI, Codey, to give smart coding suggestions. It’s like having a buddy chat with you about your code, give advice, and even tell you what your code does. Google says AI can help coders work faster and better.

This new thing runs online and hooks up with Google services. Plus, you can pull in your stuff from GitHub. Other big players like GitHub and Amazon have their versions, but Google’s spin leans into building complete web and app projects. While Project IDX showcases Google’s AI smarts, we gotta wait and see if coders will actually use it for real-deal work.


Microsoft AI Red Team building future of safer AI

Microsoft’s cooking up a plan to make AI safer. They’ve got a special team, called the AI Red Team, that acts like the bad guys to find weaknesses in AI systems before real troublemakers can exploit ’em. Since AI’s spreading like wildfire in the tech world, Microsoft’s been putting a lot of muscle behind this effort, doing everything from teaming up with universities and industry partners to creating tools that help everyone improve AI security.

What’s this red teaming all about? It’s like a game where you put on the villain’s hat and try to find ways to break into your own systems. It’s not just about guarding against hackers and such, but also figuring out how things might go wrong even for regular users.

Microsoft’s really upped the ante on this, working on ways to help folks understand and control new risks that come with AI, like tricky things that might make AI say or do something harmful. They’re even taking a serious look at fairness and privacy and making sure everyone’s on the same page about what’s right and what’s not.


Spotify expands its AI-powered DJ feature globally

Spotify’s adding a new twist to music streaming by taking its AI-driven “DJ” feature worldwide. Think of it as a personalized radio DJ that knows your musical tastes and even chats about your favorite tunes. After testing the waters in North America and parts of Europe, this feature’s now rolling out to about 50 countries, but only in English. So, if you’re a premium Spotify user in places like Australia, Sweden, or Ghana, get ready for a more chatty playlist experience. Europe’s gotta wait a bit longer, though.


Google and Universal Music negotiate deal over AI ‘deepfakes’

Google and Universal Music are chit-chatting about a deal. They want to license songs and voices of artists for tunes whipped up by artificial intelligence. Why? Because there’s been a boom in “deepfake” songs where AI makes songs sound just like big-name artists. 

This tech stuff is causing a ruckus because an artist’s voice is their bread and butter. Jacking it ain’t cool. But, if Google and Universal can work something out, fans could make these songs the right way and artists get their fair share. Sounds like a win-win, right?


Disney has a task force exploring ‘interesting opportunities’ in A.I., as writers and actors strike over concerns it could steal their jobs

Disney’s got a team checking out how A.I. might be used across the board. This has actors and writers worried because some big shots in Hollywood and the tech world are looking at A.I., like ChatGPT or Google’s Bard, to whip up movie and TV scripts. They’re also thinking of keeping a library of actors’ faces for future projects. That’s got the writers’ and actors’ unions all riled up.

Writers are scared they’ll be out of jobs if A.I. does the storytelling. Actors? They don’t want their faces used forever without a say. The big movie and TV studios tried to calm things down, saying they’d only use an actor’s face for the movie they signed up for. But then, Disney goes and gets a patent in July for tech that can put a human face on a cartoon. That didn’t help.


Apple Planning to Test AI Services Using Foxconn Servers

Apple’s got an event for the iPhone 15 either on September 12 or 13, with the launch on September 22. Foxconn, the big dog that already helps Apple with their data centers, is gonna make some special servers for Apple in Vietnam. Why Vietnam? Apple’s trying to rely less on China. These servers are for some fancy artificial intelligence (AI) stuff.

Bloomberg’s dude, Mark Gurman, hinted Apple’s working on an AI project similar to ChatGPT, but doesn’t seem like they have a solid game plan yet. CEO Tim Cook says AI is a big deal for Apple, and they’ve been poking at it for a long time.


A.I. can identify keystrokes by just the sound of your typing and steal information with 95% accuracy, new research shows

New research says A.I. can now pick out what you’re typing just from the sound of your keystrokes with a 95% success rate. In the study, folks managed to nab what was typed on a MacBook Pro using only a recording from a nearby phone or a Zoom call.

The bad guys use another gadget, say a cell phone next to your laptop or even a mic on a Zoom call, to record your typing sounds. Then, they play that recording to a super-smart A.I. that’s been trained to figure out which key makes what sound, piecing together your typed message.

The paper points out that the more we use and improve mics and this A.I., the more likely these sneaky sound attacks will become. And because many folks often work in public places like cafes, anyone could secretly record the click-clack of your keyboard without you noticing.


Mental Health App Tests Limits of Using AI for Medical Care

An app from Mexico called Yana, short for “You are not alone,” is trying to see if computers can boost our mental mood. This Spanish app gives users some to-dos like writing in a gratitude journal or doing grounding exercises. Plus, there’s a chatbot to talk out your feels. Yana’s brainchild, Andrea Campos, had the idea in 2016 when she wanted to track activities that lifted her spirits on her phone. Fast forward, and now it’s a hit with 13 million users from places like the US to Spain and even some in Asia. Campos made the chatbot herself, making it feel like you’re chit-chatting with a real person. But if the magic’s gone, users might ditch it.

The new twist? They’re adding some AI smarts to the chat. Campos thinks it can help users think deeper and get more personal advice. But making money off this chatbot, powered by OpenAI, is tricky. They’re thinking of charging daily or monthly fees. And hey, there’s also the big ethical elephant in the room. If someone’s talking about harmful stuff, the chatbot will cut off and point them to real help. Campos is also fine-tuning how unpredictable the AI replies are. But she’s got worries, especially for young folks. She’s stressing about the AI saying something harmful since not everyone gets that robots goof sometimes.


What is a ‘AI drift’ and why is it making ChatGPT dumber?

ChatGPT seems like it’s got a few screws loose lately. Even though it’s supposed to get smarter with more use, that’s not always the case. This weirdness is called “drift.” Think of it as when you tweak one part of a car to make it better, but then the brakes stop working right. A couple of brainiacs from UC Berkeley and Stanford took a peek at how ChatGPT’s brains, GPT 3.5 and GPT-4, are doing over time. They tested them with math, coding, doctor exam questions, and more.

Turns out, the newer version of GPT-4 from June wasn’t as sharp as the one from March. It goofed up especially with simple math. James Zou, one of the science guys, said the speed of this “drift” thing surprised him. But hey, it’s not all bad news. Both versions did get better at some stuff. The takeaway? Keep using these chatbots but check under their hoods regularly to make sure they’re running right.