- NATURAL 20
- Posts
- AMD Boosts Open-Source AI with Nod.AI
AMD Boosts Open-Source AI with Nod.AI
AMD and Nod.AI join forces, setting new benchmarks for open-source artificial intelligence tools and applications
Wes Roth
October 11, 2023
Today:
AMD To Acquire Nod.AI, Boosting Its Open-Source AI Software Capabilities
AMD's buyin' the AI software startup Nod.AI, lookin' to up their game in open-source AI stuff. They ain't spillin' how much they paid, but it's all about making their tech, like their data centers and chips, even cooler. If you're using AMD's gear, you'll now get to play with some fancy AI tools from Nod.AI. This move's gonna make AMD's AI offerings slicker, especially for their hardware.
Nod.AI's been around since 2013, got cash from some big backers, and has helped both big companies and the little guys. With AMD stepping in, more folks around the world can tap into Nod.AI's offerings.
Adobe is upgrading Photoshop’s generative AI model — and releasing more for Illustrator and Express
Adobe's doubling down on AI, launching three beefed-up AI models for Illustrator, Adobe Express, and Photoshop. Their big show-off was the Firefly Image 2 model for Photoshop, which they say makes way better-quality images, especially the detailed stuff like skin and hair. Plus, this model lets folks tweak photo settings, kind of like using camera controls. There's also a feature called “Generative Match” that lets you make generated content look like a specific image, but there's some chatter about how it might be misused, so Adobe's trying to play it safe with lots of rules
Next, for Illustrator, they rolled out the Firefly Vector model, bragging it's the first-ever AI model for vector graphics. In simple terms, it helps make sharp images that don't get blurry, no matter how much you resize them, which is a big deal for folks making logos and stuff. Adobe says this one’s good to go for commercial use once it's out of testing.
Lastly, there's a model called Firefly Design that whips up ready-to-use design templates. Think typing "beach party poster" and getting a bunch of cool designs to start with, kind of like what Canva does.
Tesla Builds a New Home for ‘Dojo’ Supercomputer as AI Ambitions Rise
Tesla's cooking up something big in Austin, Texas. They're building a new spot for their Dojo supercomputer, which will power the brains behind their self-driving cars. Down the line, Elon Musk's hinted they might even jump into the cloud services game, kinda like Amazon's doing with AWS.
Tesla’s been buying more chips for Dojo and some big-wigs think it could make the company a whole lot more valuable. CEO Elon Musk even hinted that Dojo might not just be for cars - it could help other companies with their computer stuff
Introducing Skills in Microsoft Viva, a new AI-powered service to grow and manage talent
Microsoft had a big reveal at the HR Tech Conference, spicing things up with their employee system, Microsoft Viva. They’ve thrown in some AI magic called "Skills in Viva" and also launched "Viva Amplify" for better workforce communication. The main aim? Keep employees in the loop and make businesses thrive.
"Skills in Viva" is like a smart assistant that can tell what skills an employee has, and where they might need a boost. With most jobs getting techy and 60% of folks feeling like they're missing some skills, Microsoft believes AI can bridge that gap. They use info from Microsoft and LinkedIn to figure out the skills landscape.
"Viva Amplify" is all about talking the talk. It helps bosses and teams communicate better, sending out the right messages to the right folks at the right time.
Cleanlab Raises $25 Million To Help Solve AI Models’ Data Mess
Cleanlab, a startup that cleans up messy data for AI models, just bagged $25 million in funding. The company, built by three MIT brainiacs, offers software that can auto-correct a lot of bad data. The whole idea is based on the old saying, "garbage in, garbage out." If you feed an AI model a ton of pics, say, of bananas, and sneak in a picture of Curious George, the model might think bananas have tails. That's because if you feed an AI junk data, you get junk results.
Now, big companies like Chase and Tesla have been using Cleanlab's free tools for a while, but they’ve recently rolled out a fancier paid version, Cleanlab Studio. They caught the eye of Databricks, a major cloud company, which tested Cleanlab and found it slashed errors and boosted accuracy. One firm saved $30 million using Cleanlab Studio.
GitHub Copilot Loses an Average of $20 Per User Per Month
Microsoft's GitHub Copilot, which helps programmers with their code, is kinda bleeding cash. Even though folks pay $10 a month to use it, Microsoft is dropping about $20 on each person. Some are even costing the company a whopping 80 bucks a month! It's no shocker then that they're planning to charge more for an upgraded version called Microsoft 365 Copilot. And get this, Google's doing the same with its own AI service.
Microsoft is trying to make their own chips to cut down on those costs. They're pushing for new computer chips called NPUs to handle AI tasks right on our computers, so they don't have to spend as much on the backend. They're also thinking about using cheaper services for free stuff like Bing Chat.
Microsoft announces new AI tools to help doctors deliver better care
Microsoft just dropped some new AI tools to help doctors give even better care. Turns out, hospitals are sitting on a mountain of data, but they can barely use any of it 'cause it's all over the place. Microsoft's new system, called Fabric, is like a magic box that pulls all that info together. It's like cleaning up a messy room and finding you had a treasure chest under all the junk.
On top of that, Microsoft's also rolling out some fancy AI tools in its Azure platform. One's a chatbot that can answer medical questions. They've also got a tool that can sift through doctor's notes to pull out the important stuff. Plus, there's this cool tool that gives doctors a quick look at a patient's health history like flipping through a photo album.
AI Uncovers Secrets of Brain Aging
Scientists at Mount Sinai used some cool AI tech to come up with "HistoAge", a tool that can guess how old you are by looking at sections of your brain. They took a peek at 700 samples from older folks' brains, particularly a spot called the hippocampus, which is a hotspot for age-related issues. The results? This AI can predict your age at death with an error margin of just over 5 years.
Not only is it a neat age guesser, but HistoAge is also good at spotting early signs of brain issues like Alzheimer's. This could be a game-changer, as we might be able to understand and maybe even slow down brain aging and diseases.
Using digital and AI to meet the energy sector’s net-zero challenge
Clean energy like wind and solar is key to hitting net-zero emissions by 2050. Even though renewables have been kickin' it out of the park, they're facing some roadblocks like slow red tape and limited land.
Companies in the renewable energy game are now using fancy tech like AI to speed things up. This article chats with big shots from a renewable energy company, an energy provider, and a tech giant to get their two cents.
The first open source alternative to ChatGPT
Zephyr is a new series of language models designed to be helpful buddies. The first model in this lineup, Zephyr-7B-α, is a fancy version of an earlier model called Mistral-7B-v0.1. They tweaked it using a special method called Direct Preference Optimization, and guess what? It performed better!
What's inside this model:
It's like GPT but has 7 billion parameters.
Speaks mainly English.
It's free but with some strings attached (a specific license).
It's evolved from the Mistral-7B-v0.1 model.
They first refined the model using some chat-based data and then gave it a tune-up using another dataset and a training tool.
Modal is now generally available
Modal, a platform to make cloud functions a breeze, just launched for everyone. Started in 2021, its goal? Help data teams run stuff in the cloud without a hitch. They saw that data and tech folks were getting bogged down with the techy side of things, like wrangling with setup files, instead of just getting their code up and running. So, Modal came into the picture.
What's it do? Modal lets you:
Whip up Python code and see it work in the cloud pronto.
Handle big tasks with fancy tech gear.
Set up recurring tasks or web stuff with just a line of code.
And the best part? No tricky setups. Everything's code.
Some big names are already using Modal for their heavy lifting. Ramp for their big data projects. Substack to turn spoken words into written ones super-fast. Suno for creating music with AI. And Sphinx Bio for science-y stuff like protein folding.
What'd you think of today's edition? |
What are MOST interested in learning about AI?What stories or resources will be most interesting for you to hear about? |
Reply