- NATURAL 20
- Posts
- MemGPT 🧠 the next BIG thing in AI
MemGPT 🧠 the next BIG thing in AI
MemGPT is a system designed to break the barriers of limited context windows in large language models.
Wes Roth
November 03, 2023
Today:
MemGPT 🧠 the next BIG thing in AI
Large language models (LLMs) are pretty much the bee’s knees in AI, but they’ve got this issue where they can’t keep a lot of info in their head at once, which makes them not so hot for long chats or digging into big documents.
MemGPT, the new kid on the block, inspired by how old-school computer systems juggle memory. It’s like having a smart assistant that can shuffle papers on a desk and keep the important stuff on hand. MemGPT helps LLMs keep track of more stuff at once, and it’s got this nifty trick of hitting pause and chatting with the user when it needs to sort things out.
Microsoft starts selling AI tool for Office, which could generate $10 billion a year by 2026
Microsoft is riding high and cashing in on AI, kicking off sales of their new tool, Microsoft 365 Copilot, for their Office apps. This bad boy doesn't come cheap at $30 a head each month, but analysts are betting it could rake in over $10 billion a year by 2026. The competition's heating up with Google trying to snag a piece of the pie, but Microsoft's not sweating it.
The folks over at Piper Sandler are putting their money on Microsoft, predicting a good chunk of users will jump on the Copilot bandwagon. There's a real fear of missing out (FOMO) factor here. If your competition's using Copilot and you're not, you might just be left in the dust.
Microsoft's CEO, Satya Nadella, is all in, saying customers are hooked on Copilot once they give it a whirl. They've been testing the waters with the big dogs, with 40% of Fortune 100 companies getting a sneak peek and loving it.
But here's the catch: it's still early days, and there's not a ton of data on how Copilot's really shaking things up performance-wise. And you need a hefty 300 licenses to get in on the action. So, it might be a slow burn to get everyone on board. While companies are setting up Copilot, they might just find themselves using more of Microsoft's Azure cloud services. So, it's a win-win for Microsoft.
AI companies agree to government tests on their technology to assess national security risks
A bunch of big-shot AI companies, including OpenAI, Google DeepMind, and Meta, have shaken hands with governments from the US, UK, Singapore, and others, promising to let them check out their latest AI tech for any national security red flags before releasing it to businesses and the general public. This all went down at a two-day shindig in the UK, wrapping up with the signing of a big deal document – though it’s more of a handshake agreement than a legally binding contract.
Governments from around the world, like Australia, Canada, the EU, and Japan (but not China), also put their names on the dotted line. They’ve got a panel of smarty-pants set to roll out a yearly report, keeping tabs on any shady AI business, like bias or spreading fake news, and even scarier stuff like helping make chemical weapons.
Rishi Sunak, the UK Prime Minister, is all pumped up, saying this is a huge win for the good guys. He's confident that they’re leading the pack in making sure AI stays safe. The US is also on board, issuing an executive order to tackle AI threats and planning to set up their own AI watchdog.
Elon Musk tells Rishi Sunak AI will render all jobs obsolete
Elon Musk, the big-shot entrepreneur, had a heart-to-heart with UK Prime Minister Rishi Sunak, telling him that artificial intelligence (AI) is shaking things up like never before. Speaking at the fancy Lancaster House in London, Musk painted a picture of a future where jobs are optional, thanks to AI's ability to handle anything and everything.
When Sunak asked for his two cents on the initiative, Musk, with a smirk, said it might be a hassle but gave it a thumbs up, emphasizing the need for a watchdog. He threw in a word of caution, stressing the need to have an "off switch" for AI, just in case things go south.
Wrapping up the 40-minute talk, Musk gave props to Sunak for including China in the AI Safety Summit, saying it’s crucial for them to be on board. He also mentioned that he signed a letter earlier this year, calling for a time-out on advanced AI development due to safety worries. But some folks think he’s just playing catch-up with OpenAI and their popular ChatGPT chatbot.
Box announces Vertex-powered generative AI tools to process and analyze documents faster
Box, the cloud storage peeps, are teaming up with Google Cloud to make their service even snazzier. They're using Google's Vertex AI to help folks quickly sift through and make sense of their data on Box. If you're familiar with ChatGPT, that's the kind of tech we're talking about – AI that can think and act like a human.
Box's new feature will help businesses automatically sort their files and find important stuff like expiring contracts or unpaid invoices. Plus, they're making it even tighter with Google Workspace, so users can do more straight from Box. You'll see stuff like creating documents, attaching files to emails in Gmail, and more.
On top of that, Box will now use more of Google Cloud's tools, offering faster and smarter service. Some experts think it's a smart move since Google's ahead of the game in this area. To sum it up, if you're using Box, expect to see some cool upgrades thanks to their buddy-buddy situation with Google Cloud.
Meet Dot, an AI companion designed by an Apple alum, here to help you live your best life
Dot's an AI buddy created by an ex-Apple designer and a team of hotshots. It's like a chat app on steroids, helping you connect dots in life and pushing you to up your game. It’s not just about answering questions; it's about building a history, a journey with you.
This thing remembers almost everything you tell it, making it like a digital diary on speed. It suggests stuff based on what you like, stores your recipes, and even helps with life's curveballs. The goal? To be more than just an assistant; it’s aiming to be a guide in your life.
The creator, Yuan, has a unique take on it, calling your chats with Dot a "living history." It's less about quick fixes and more about the long haul, helping you figure out the big picture. And it’s designed to be a good listener and a thoughtful responder, thanks to some help from a playwright.
Brave responds to Bing and ChatGPT with a new ‘anonymous and secure’ AI chatbot
Brave's stepping up its game with Leo, a new AI assistant that's all about keeping things on the down low. This new buddy won't blab your secrets, meaning it won't record your chats or use them to get smarter. And the best part? It's free for all Brave desktop users, as long as you've got the latest version.
Leo's got all the usual tricks up its sleeve, like answering questions and translating languages, but it's rolling out in stages, so keep your eyes peeled. And for the folks who like a little extra, there's a premium version with a faster AI assistant, but it'll cost you 15 bucks a month.
The folks at Brave are all in on privacy, and they're making sure Leo sticks to that promise. So while Leo's got your back, remember it's not perfect and might make mistakes. But hey, nobody's perfect, right? Just treat it like a new friend who's still learning the ropes.
Ukraine Is Using AI to Help Clear Millions of Russian Landmines
In Kyiv, officials use tech to identify and remove the massive number of landmines plaguing the country. After Russia's invasion, Ukraine's become the world's most landmine-heavy spot, with a third of its land littered with explosives. This not only slows down Ukraine’s military but also risks millions of civilian lives and messes with prime farming land, hurting the nation's economy.
Svyrydenko, Ukraine's Economy Minister, ain't waiting around. She's got big plans to get 80% of this mess cleaned up in a decade. To do this, she's getting techy with it. We're talking AI, drones, and loads of data. Ukraine’s teaming up with U.S. tech company, Palantir, to figure out the most important spots to clean first. They’re combining tons of different info sources and using young Ukrainian coders and innovative farmers to speed up the clearing process.
This won't be a quick fix. The World Bank thinks cleaning up might cost more than $37 billion. But folks reckon the methods Ukraine's trying out could change how we deal with post-war cleanups worldwide.
The Beatles’ final song is now streaming thanks to AI
The Beatles just dropped a "new" song, “Now and Then” – the first since '95. It's up on streaming and here's the kicker: Paul and Ringo used some high-tech magic to build it from an old rough recording by John Lennon. They tried to make this tune work in the '90s but got stuck because the old tape was a mess. So it just sat in a closet.
Fast forward to when Peter Jackson's working on a Beatles documentary, and his crew whips up this tech that can pick apart any old music track. Paul and Ringo see their chance, add their own flair, and boom – we got a fresh Beatles song. Some folks are side-eyeing how it came together, but both living Beatles and their families are vibing with it. It’s wild to think this might be the last Beatles jam, but hey, if old recordings can get this treatment, who knows what golden oldies might come back to life next?
Scarlett Johansson takes legal action against use of image for AI
Scarlett Johansson is throwing the book at an AI app that used her face and name to hawk their product without asking her first. The ad, which ran for 22 seconds on a platform that used to be Twitter, showed a fake version of Scarlett, complete with made-up words coming out of her mouth. She caught wind of it on October 28 and didn't waste any time getting her lawyer on the case. The ad's been yanked since.
The gist of the ad? It starts with a real clip of Scarlett on a movie set, then switches to a computer-made version of her talking up the app. Even though the ad had a tiny disclaimer saying the images were made by the app and not connected to Scarlett, she wasn't having any of it.
Microsoft launches probe after AI poll on woman's death sparks blowback
Microsoft is in hot water and looking into what went wrong after a poll about a woman's death popped up on their news aggregator, Microsoft Start, causing quite the stir. The poll, which asked readers to guess why the woman died, was called out by The Guardian Media Group, saying it was in bad taste and blaming Microsoft’s AI tech for the mishap.
In response, Microsoft pulled the plug on all AI-generated polls on news articles and is now digging into how this all happened in the first place. The company said it’s taking steps to make sure this kind of mistake doesn't happen again.
The Guardian wasn't just upset about the poll itself; they also wanted Microsoft to publicly own up to the mistake and make sure their news content isn't messed with by Microsoft’s AI tech in the future. They're also pushing Microsoft to talk about how they plan to pay news publishers for using their content to train AI models.
Processor made for AI speeds up genome assembly
A fancy new computer chip made for AI is making waves in the bio world, speeding up how we put together DNA and protein puzzles up to ten times faster than the usual tools. Giulia Guidi from Cornell and her crew tested this hot rod, called an IPU, and it rocked at lining up these tiny biological bits.
When scientists read DNA, they get a bunch of short snippets that they gotta piece together. Old school methods used graphics cards to do this faster, but this new IPU is like trading in a bike for a rocket. It’s especially good for the bumpy and unpredictable nature of DNA alignment.
The team even shrunk down the memory needed for their tools, so they could do more work directly on the IPU, making everything zip along even faster. The hope is, with a bit more tweaking and upgrading, this IPU thing could be the new go-to tool for big biology problems.
New study validates the business value and opportunity of AI
Microsoft's latest blog post is a real eye-opener about AI's place in the business world. They had IDC do some digging and it turns out, AI is not just a fad—it's making companies serious money! We're talking a $3.50 return for every buck put in. And it's not just the tech giants; businesses all over are getting in on the action, with 71% already using AI.
But it's not all sunshine and rainbows. Companies are running into some hiccups, especially when it comes to finding folks who know their way around AI. Over half of the businesses surveyed said that's their biggest headache.
Microsoft's not just sitting on this info. They're all about spreading the AI love, training millions of people and helping businesses get their AI game on point. They're even showing off some real-world examples on their Cloud blog for anyone looking to jump on the AI bandwagon.
How Apple’s thinking about AI spending and deployment differs from Big Tech rivals
Apple's been on a shopping spree, scooping up over two dozen AI companies since 2015. You might not recognize names like Emotient or Drive.ai, but Apple's been baking their tech into gadgets like iPhones, Macs, and the upcoming Vision Pro VR headset.
Now, while Apple's rivals are all about crafting stand-alone AI models, Apple's strategy is different. They're picking up companies with top talent to infuse AI into consumer products. Think of Apple as being more about the backstage tech to make your gadgets smarter, not just creating a standalone AI superstar.
CEO Tim Cook is a bit hush-hush on AI but emphasizes its importance to Apple. For instance, the new iPhone 15 lets users mimic their voice, and the camera can tell if you're snapping a pic of Fido or a friend.
But some folks think Apple's trailing in the AI race. One analyst says they're "far behind." But here's the twist: Apple's got a massive user base, over 2 billion, running on its software. So, while rivals might have a head start, Apple has a sea of users waiting for whatever they dish out.
Why Conscious AI Is a Bad, Bad Idea
AI is making big moves. You can chat with systems like ChatGPT as if they were real people, and they can even create cool photos and fake videos. But are these machines actually "awake" like we are? Some folks think we're close to having conscious AI, while others, like the author, aren't so sure. Even if it's a long shot, we shouldn't ignore the risks. If AI seemed conscious, even if they aren't, it can mess with our heads big time.
A bunch of smart people wrote a letter saying we need to think hard about whether AI can become conscious and what that means. They're calling for a big organization to keep an eye on AI worldwide. Even big names in the AI game are raising eyebrows. The author thinks we shouldn't even try making machines that are "awake".
Even if AI isn't truly conscious, if it seems like it is, it can mess with our heads. We might trust them too much or think they care about us. This can lead to all sorts of problems. Think of the mess in TV shows like Westworld.
Biden’s Elusive AI Whisperer Finally Goes On the Record. Here’s His Warning.
Bruce Reed, a big shot at the White House and a brainy Democrat, is sounding the alarm on AI, and he's finally spilling the beans about it. Folks have been getting jittery about AI, what with tools like ChatGPT making waves and Bing's chatbot pulling some crazy stunts. Reed, charged with cooking up an AI strategy for the Biden administration, got to thinking that maybe we're not worried enough.
He had a powwow with some tech ethicists, including Tristan Harris, and they laid it out straight: AI's been messing with our heads on social media for a while, and now it's about to pull even crazier tricks, like cracking computer codes in a snap and copying voices with just a snippet of sound. Reed took this to heart, realizing AI could really shake things up and mess with our grip on reality.
Reed knows the stakes are high. He's seen how social media took us for a ride, and he's determined not to let AI pull the same stunt. He's on a mission, and there’s no time to waste. As Reed puts it, it's high time for the policymakers to step up their game and make sure AI plays by the rules. No sitting on the sidelines – it's time to get in the game and shape the future of AI.
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