- NATURAL 20
- Posts
- Microsoft LOSES Sam Altman PLUS a theory of what is ACTUALLY happening at OpenAI...
Microsoft LOSES Sam Altman PLUS a theory of what is ACTUALLY happening at OpenAI...
Dive into the latest developments as Sam Altman makes a surprising bid to return to OpenAI amidst Microsoft negotiations
Wes Roth
November 21, 2023
Today:
Microsoft LOSES Sam Altman PLUS a theory of what is ACTUALLY happening at OpenAI...
Sam Altman, once booted from his gig as OpenAI's big boss, is making moves to come back. Yeah, you heard that right. This ain't no fake news; it's all over The Verge. The twist? The deal where he was heading to Microsoft ain't set in stone yet. Now, Ilia Sutskever, one of the OpenAI honchos, has flipped sides, backing Altman's return. This means they gotta convince a couple more board members to switch their tune.
This whole mess started getting more tangled when Microsoft, led by Satya Nadella, stepped in, offering to shelter the OpenAI team under its wing, kinda like how DeepMind chills under Google's Alphabet umbrella.
OpenAI Approached Anthropic About Merger
OpenAI's bigwigs, the board of directors, hit up Dario Amodei, the top dog at Anthropic (another big player in the AI language model game) to talk about possibly joining forces.
This all went down right after OpenAI gave their CEO, Sam Altman, the boot. The board seemed keen on getting Amodei to take over Altman's spot as the CEO of OpenAI. That's a pretty big deal, considering these are two heavy hitters in the AI world.
Amodei wasn't having any of it. Despite the big offer, he turned it down flat because of his commitment to Anthropic. So, as far as we know, the merger talks didn't really go anywhere after that.
In short, OpenAI was looking to shake things up by merging with Anthropic and getting their CEO to lead the charge. But Anthropic's head honcho decided to stick to his guns and keep doing his own thing.
Amazon aims to provide free AI skills training to 2 million people by 2025 with its new ‘AI Ready’ commitment
Amazon's stepping up big time with their "AI Ready" promise. They're gonna teach 2 million folks how to use artificial intelligence (AI) for free by 2025. This is huge 'cause AI's the big deal in tech right now. They're rolling out new stuff for adults and kids and making their current AI training programs free, so money won't stop anyone from learning these top-notch skills.
They've got three fresh plans:
Eight free AI courses, including the fancy generative AI stuff.
The AWS Generative AI Scholarship for over 50,000 high school and college kids, with a new course on Udacity.
Teaming up with Code.org to teach students about generative AI.
There's a massive need for AI-smart workers. A study shows 73% of employers are hunting for AI talent, but three-quarters can't find what they need. Upskilling in AI could mean up to 47% higher salaries. Plus, 93% of businesses reckon they'll be using AI in all sorts of ways in the next five years.
Amazon's got eight new, free AI courses for different job needs. Some are basic, some are advanced, and they're for everyone from business leaders to tech nerds. This adds to their existing 80+ AI courses.
And for the young crowd, Amazon's working with Code.org on this cool "Hour of Code Dance Party: AI Edition" where kids code a music video with AI. They're doing this globally during Computer Science Education Week in December.
Microsoft Shares Reach Record High After Hiring Former OpenAI CEO Sam Altman
Microsoft's stock prices went through the roof, hitting an all-time high on Monday. Why? They just brought on board Sam Altman, the ex-big cheese of OpenAI (yep, the folks behind ChatGPT) to run their new AI squad. Microsoft's been throwing big bucks into AI, especially with their cozy partnership with OpenAI, and it's paying off. They're in the big leagues now, rubbing shoulders with other tech titans like Amazon, Google's parent Alphabet, and Meta, all duking it out in the AI arena.
As of Monday, Microsoft's value was a whopping $2.81 trillion, second only to Apple in the U.S. The boost in their stock came not just from hiring Altman, but also from some cool AI stuff they've been unveiling, like a new AI chip and the Copilot features. Plus, OpenAI's fresh off announcing their latest brainchild, GPT-4 Turbo.
Spotify partners with Google’s AI to offer better recommendations to its users
Spotify and Google Cloud are teaming up big time. Spotify's diving deep into AI to figure out what you like to hear, not just in tunes but also in podcasts and audiobooks. They're using these fancy AI tools, like OpenAI's ChatGPT and Google's own smarties, to get smart about what you're into.
Google's got a bunch of these AI brains - think PaLM 2, Codey, Imagen, Chirp - that are sharp in text, code, images, you name it. Spotify's been into AI for a while, using it to suggest music. Now, they're getting into using it for all the other stuff they offer. They're hoping this AI move will make them more dough, especially with podcasts and audiobooks. They've been spending big on these areas, expecting good returns.
Gustav Söderström, the big brain at Spotify for products and tech, says this is a big leap for them, and they're psyched about Google Cloud's gear for their AI adventures. Also, they're thinking about using AI to keep things safe and filter out bad stuff.
NVIDIA’s New Ethernet Networking Platform for AI Available Soon From Dell Technologies, Hewlett Packard Enterprise, Lenovo
NVIDIA's got this hot new tech for AI, the Spectrum-X Ethernet stuff, and the big players like Dell, HP, and Lenovo are jumping on board. This gear is built just for AI, especially the kind that creates stuff, and it's way faster than the usual internet gear.
They're mixing this Spectrum-X with some fancy NVIDIA GPUs and AI software to help businesses really get into using AI. Jensen Huang from NVIDIA is all about this, saying it's gonna be a game-changer in data centers for AI workloads.
Spectrum-X got this super-fast Spectrum-4 Ethernet switch, a BlueField®-3 SuperNIC for handling intense AI work, and some software to make it all sing. This Spectrum-4 switch is a beast, handling data like a champ and keeping things smooth in busy AI cloud environments.
The BlueField-3 SuperNICs are perfect for heavy-duty computing, keeping things fast and secure, and they fit nicely in enterprise servers. The software side of Spectrum-X is loaded with NVIDIA tools to make AI work run smoothly.
NVIDIA's also showing off this setup with their Israel-1 supercomputer, a beast built with Dell and rocking this Spectrum-X tech.
Mastercard doubles down on effort to detect and tackle crypto fraud with AI tie-up
Mastercard's going all-in on stopping crooks using cryptocurrencies. They've teamed up with this tech company, Feedzai, that's real smart at spotting shady money moves online using AI. Together, they're gonna keep an eye on over 6,000 crypto exchanges for any fishy business, like fraud and money laundering.
Feedzai's gonna soak up all the data from Mastercard's CipherTrace Armada platform, which helps banks keep tabs on crypto transactions. The cool part? This setup can sniff out bad transactions in a blink and also knows when a transaction is on the up and up.
A big chunk of scammy crypto transactions start right out of a regular bank account. With this new deal, Mastercard gets to tap into Feedzai's AI brainpower, which looks over transactions worth a whopping $1.7 trillion every year.
This move by Mastercard isn't just about stopping fraudsters. It's also about making crypto more legit, like regular money stuff. Banks are curious about crypto, but they're kinda nervous about jumping in because of the sketchy side of it. Last year, crypto scams and thefts went through the roof, with $14 billion going to illegal addresses.
UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage, lawsuit claims
Two families, who lost their loved ones, are taking on this big health care company, UnitedHealth. They're saying UnitedHealth used some messed-up AI to cut off care for older folks, care that their docs said was super important. This went down in a Minnesota federal court last Tuesday.
UnitedHealth offers Medicare Advantage plans. These are like an alternative to the regular government Medicare, but run by private companies. The suit says UnitedHealth used this shaky AI, made by NaviHealth and called "nH Predict," to stop paying for elderly patients' care unfairly and too soon. This left some old folks either kicked out of care facilities or draining their savings to keep getting the help they needed.
NaviHealth's guy, Aaron Albright, told CBS MoneyWatch that their AI isn't really calling the shots on coverage, it's just there to give advice on what patients might need. He says final decisions are based on Medicare's rules and the patient's insurance plan. Albright thinks the lawsuit doesn't stand a chance.
Former GitHub CEO Friedman, Scale AI Founder Wang Declined OpenAI Top Job
Nat Friedman, the big cheese who used to run GitHub (you know, that's Microsoft's code hub), got tapped on the shoulder by someone from OpenAI's team. They wanted him to step in as the interim head honcho over at OpenAI, but Friedman was like, "Nah, I'll pass." This info comes straight from someone who knows the deal.
Now, OpenAI, they're the brainiacs behind stuff like ChatGPT, and they were looking to fill the big seat after Sam Altman decided he wouldn't be coming back as CEO. So, before they got Twitch co-founder Emmett Shear to agree to be the interim CEO, they had their eyes on Friedman, but he wasn't biting.
OpenAI board faces growing revolt over Sam Altman’s ousting
OpenAI's been in hot water lately. Basically, their CEO Sam Altman got the boot, and it's caused a big mess. A ton of employees and some investors are up in arms, demanding that three big-shot directors step down. The number's pretty wild: 747 out of 770 employees are saying, "Enough's enough."
There were some talks about bringing Altman back, but he was like, "Only if the board quits." They didn't bite, so they brought in Emmett Shear from Twitch to take the reins temporarily. Meanwhile, Altman and his co-founder, Greg Brockman, got scooped up by Microsoft to lead a new AI unit.
OpenAI's chief scientist, Ilya Sutskever, is now on the employees' side after regretting his part in firing Altman. And these employees, they're ready to jump ship to Microsoft's new unit unless the current board members quit and they get some new independent leaders.
Some big-name investors are holding out hope for Altman's return. They're talking about a massive stock sale that could value OpenAI at $86 billion, but that's all up in the air now. Other companies are trying to cash in on this chaos. Like, Salesforce's CEO is practically asking OpenAI researchers to join his team.
South African university students use AI to help them understand – not to avoid work
A bunch of college professors in South Africa got curious about how students use AI, like ChatGPT, in school. They ran a survey with over 1400 students from five universities to see what's up. Turns out, students aren't just using ChatGPT to skip out on homework. They're actually using it and other AI tools to get a better handle on tough concepts, brainstorm for essays, and polish their writing.
The survey showed that most students do their schoolwork on laptops or phones. They're using a bunch of different AI tools, not just ChatGPT, for stuff like translations and fixing up their English. Less than half have used ChatGPT for essay questions, and they're aware of the risks like plagiarism.
Students think these AI tools are pretty helpful for understanding their courses, especially when something isn't clear or English isn't their first language. They're using AI as a sort of digital tutor, helping them get their heads around tough topics.
Synthetic imagery sets new bar in AI training efficiency
MIT's brainiacs are breaking new ground in AI training by using fake pictures instead of real ones. They've got this thing called StableRep, which makes images using super cool text-to-image tech like Stable Diffusion. It's like they're painting pictures with words.
The trick up StableRep's sleeve? It's called “multi-positive contrastive learning.” Lijie Fan, a smart cookie at MIT, says they're teaching their AI to grasp big ideas from different angles, not just by stuffing it with data. They use a bunch of images made from the same words and treat them like they're showing the same thing. This helps the AI get the gist of what's in the pictures, not just the pixels.
StableRep is so good it's beating the big guns trained on real-life photos. It's shaking up how we train AIs, making it cheaper and easier. No more hunting down photos or dealing with messy, biased internet data. Now, you just need to type in what you want, and bam! You've got your training data.
Reshaping protein design with function-first, AI-guided engineering
So, these brainy folks at Generate Biomedicines in Massachusetts have cooked up a slick AI they're calling Chroma. This bad boy can whip up new proteins – those tiny workers in our bodies – and figure out what they can do. Why's that a big deal? Well, proteins are like the VIPs of our body's party, and we use them in loads of medicines.
Chroma's like a high-tech Lego set for proteins. You tell it what you want – like how the bits of the protein should be arranged – and it builds it for you. This isn't just about making new proteins; it's about making them do specific stuff, which is huge for medicine.
They put Chroma to the test, and even though it's early days, it's showing some promise. It's hitting the mark about 3% of the time, which might not sound like much, but in this game, that's pretty neat. It's nailing complex designs and could lead to drugs we haven't even dreamed of yet
What Elon Musk is really building inside his ChatGPT competitor xAI
Elon Musk, after leaving OpenAI in 2018, has jumped back into the AI game with his new venture, xAI. Musk's goal with xAI is pretty ambitious – he wants it to understand the universe's true nature. He's got an all-star, all-male AI team to kick things off, each owning a bit of the company. Musk's idea of AI safety is different from others; he thinks an AI that's super curious and truth-seeking is the way to go, instead of trying to align it with human values.
In November, Musk unveiled Grok, xAI's chatbot. Unlike other AI chatbots, Grok's got a bit of an attitude, not shying away from snark and even some off-color humor. It's part of X (formerly Twitter) and uses tweets for learning. But how Grok ties in with Musk's other tech ventures, like Tesla's self-driving tech or the humanoid robots, is still up in the air.
Musk seems to have a bit of a rivalry with Google DeepMind, and his new AI team, mostly math and physics whizzes from Google DeepMind, seem more focused on unlocking new truths through mathematics. Their approach is different from the typical generative AI models, aiming for reasoning and understanding, not just repeating learned material.
Musk's plan for xAI isn't crystal clear yet. So far, Grok seems more about being a cheeky chatbot than anything else. Musk hints at potential uses for both consumers and businesses but doesn't get specific. To really take on the big players in AI, xAI will need a ton of computing power, which Musk is already working on by securing server space and high-end processors.
Who is new interim OpenAI CEO Emmett Shear? All you need to know about the man replacing Sam Altman
Emmett Shear? You might not know him, but in Silicon Valley, the guy's a big deal. He's one of the brains behind Twitch, that live-streaming giant, which he and his buddies started back in 2007. Twitch got so big, Amazon scooped it up for a cool billion in 2014. Shear was leading the charge until earlier this year, but stepped down, partly to hang with his new kiddo.
So, why's he jumping into OpenAI's hot seat? Well, he's not one to shy away from a challenge, and he thinks OpenAI's pretty important – they're the folks behind ChatGPT, that chatbot everyone's talking about. Shear's got his work cut out, though. With all the buzz around AI, he's gonna be under the microscope, especially with regulators and world leaders sweating over AI safety and how it might turn our world upside down.
Shear's not blind to the risks either. On a podcast, he once said that AI's potential to go haywire is the kind of stuff that should really freak people out, talking about it being a "universe-destroying bomb" level of scary.
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