Meta Is Investing in New Technologies and Infrastructure to Build a Chatbot as Advanced as OpenAI’s GPT-4
Meta sets GPT-4 as the bar for its next AI model
Meta’s shooting for the stars, aiming to whip up a chatbot as slick as OpenAI’s big dog, GPT-4. The skinny? They’re buying a bunch of fancy computer chips and building mega data hubs to pull this off. Zuckerberg, the head honcho at Meta, wants this new tech toy to be free for companies to play around with. They’re also ditching Microsoft’s cloud stuff, going all DIY.
Rumor has it, Meta’s been testing an Instagram bot with a bunch of personalities, kinda like the sci-fi stuff they’re planning to unveil soon. But it ain’t all sunshine; Meta’s had some brainy folks jump ship, and they’ve got some stiff competition.
While OpenAI’s taking a breather and not working on a GPT-5 anytime soon, Apple’s throwing mad cash at their own AI project. Meanwhile, Google, Microsoft, and Amazon are also throwing their hats in the AI ring.
AI-Powered Creative Guidance In Google Ads
Google’s stepping up its ad game with a new AI tool that gives advertisers tips on how to make their video ads better. Found in the Google Ads platform, this feature checks if your ad is missing the good stuff, like a logo in the first few seconds or a voiceover. It even looks at how long your video is to see if it matches what you’re trying to sell. If you’re falling short, it’ll give you a heads-up and suggest ways to step it up.
But remember, the tool uses past data to give advice, so it might not be a perfect fit for every brand’s vibe. Still, this could be a big help for businesses looking to get more bang for their advertising buck.
Google.org to invest $20M into AI-focused grants for think tanks and academic institutions
Just before a big private meeting in Congress about artificial intelligence, Google announced a $20 million fund through its charity arm, Google.org. The cash will go to think tanks and colleges to research the good, bad, and ugly of AI—everything from job impact to national security. The goal is to get independent minds tackling tough questions, like how to keep AI fair and secure.
Some big-name orgs like the Aspen Institute and MIT are already in line to get some of that money. Google says this isn’t just a U.S. thing; they’re looking to help groups globally. The timing’s hot, with everyone from President Biden to Europe’s bigwigs trying to lay down some ground rules for AI.
Numenta launches brain-based NuPIC to make AI processing up to 100 times more efficient
Numenta, a company that’s been studying the brain for 17 years, says it’s figured out how to make AI a whole lot faster and cheaper. They’ve launched a new platform called NuPIC, which stands for Numenta Platform for Intelligent Computing. It’s based on neuroscience research and aims to make AI up to 100 times more efficient. Even better, it’s designed to run on the CPUs you’ve already got, instead of needing special graphics chips (GPUs). This could save businesses big bucks.
What’s cool is that Numenta’s teaming up with Gallium Studios, the game makers behind stuff like The Sims and Where in the World is Carmen Sandiego. They’re gonna use NuPIC to help make AI characters in their new game smarter and more adaptable while keeping your data secure and private.
NuPIC also promises to make AI more accessible for regular software folks, no PhD required. And because it’s all designed to work on a CPU, the tech’s more flexible and easier to work with than solutions tied to GPUs. This could be a game-changer, making AI faster, cheaper, and more secure for everyone.
AI app Character.ai is catching up to ChatGPT in the US
Character.ai is giving ChatGPT a run for its money, especially among the younger crowd. Launched in May 2023, Character.ai already has 4.2 million active users in the U.S., while ChatGPT sits at about 6 million. Character.ai’s doing something right because most mobile apps lose users like a sieve but it’s held onto its early fans. It’s still behind ChatGPT on a global scale, but it’s grabbing the attention of folks aged 18-24 like no other AI app out there.
While ChatGPT saw its web visits dip recently, Character.ai held strong, probably because it’s more about fun than doing your homework. Plus, as schools are back in session, ChatGPT is showing signs of bouncing back. Character.ai’s got deep pockets, thanks to a $150 million investment, and some serious brains behind it. The founders are Noam Shazeer and Daniel De Freitas, who’ve been big shots in AI for years.
Unistellar’s AI telescope lets you view galaxies in the night sky from the city
Unistellar launched an AI-powered telescope that lets folks peek at galaxies right from their backyards, even if city lights are blazing. The trick? Their telescope uses smart tech to wipe out light pollution, letting you see the night’s wonders without hiking to the middle of nowhere. Just connect your smartphone to their eQuinox 2 telescope, and you’re in control, directing it to stars, nebulae, and more. Don’t expect it to be pocket change, though – it’s $2,500.
It’s got a database of 37 million stars and can even give a mini astronomy lesson right on your phone. Started in 2017, Unistellar had a rocketing Kickstarter campaign and now has thousands of users globally. Join in, and you might just feel super tiny as you marvel at a star cluster 22,000 light-years away. Cool, right?
Spatial Biology and AI Combined by Noetik to Develop Immuno-Oncology Drugs
Ron Alfa and Jacob Rinaldi, the brains behind Noetik, are trying to make smarter cancer drugs. They believe current cancer treatments are too general and don’t consider patients’ immune biology. To tackle this, they’re using a blend of spatial biology and artificial intelligence (AI).
What’s spatial biology? It’s like a high-tech map that shows how different aspects of a tumor interact. Noetik uses this data to train AI models to understand tumor biology in ways that we humans can’t. They’re not making drugs yet, but they aim to use this info to develop medicines that target specific kinds of tumors.
In simpler terms, Noetik is using big data and AI to figure out why some tumors respond to treatments and others don’t. They’ve collected heaps of data on patients and are running their machines 24/7 to learn as much as possible. They recently bagged $14M in funding, which shows folks are paying attention.
Scientists Are Beginning to Learn the Language of Bats and Bees Using AI
Scientists are using AI to crack the code on how animals like bats and bees talk to each other. No joke, they’re actually recording these critters, then running the sounds through special computer programs to figure out what’s what. Turns out, bats and bees have a lot to say, from who they are to where the good stuff (like food) is.
This isn’t just for kicks; understanding animal talk could change how we see our furry and buzzy neighbors, and maybe even how we treat them. It’s like opening up a whole new world, kinda like how microscopes did for seeing tiny things. Super cool, right?
New benchmark tests speed of running AI models
A group called MLCommons recently released results on tests checking how fast the latest tech gadgets can run AI stuff. Nvidia’s chip came out on top for running a big AI model, but Intel’s version wasn’t far behind. This test used a huge AI model that sums up CNN news stories and looks at the AI process that powers creative AI tools.
Nvidia’s been big in training AI but hasn’t really nailed this particular part of the AI market. The bigwig at Nvidia, Dave Salvator, said they’re doing great all around. Intel’s chip was about 10% slower, but they claim it’s cheaper than Nvidia’s. Both companies are hush-hush about the actual price.
Forecasting inflation with AI
Folks from the St. Louis Federal Reserve used Google’s language model, PaLM, to predict inflation rates from 2019 to early 2023. They compared these AI-generated predictions to what human pros forecasted and the actual numbers. Guess what? The AI did a better job most of the time!
The researchers made sure the AI wasn’t “cheating” by using current info. They set it up so that PaLM would only use information available up to a certain date when making its forecasts. Still, there were some bumps. For example, the AI’s predictions could vary depending on how you asked the question, and sometimes even gave different answers with the same question.
UK researchers start using AI for air traffic control
UK scientists are testing how artificial intelligence (AI) can help or maybe even replace human air traffic controllers. They’ve teamed up with the company that runs the UK’s air traffic control, a big-shot data science institute, and a university. They’ve made a computer simulation of England’s skies to experiment in. The goal? To make flights more fuel-efficient, reduce air traffic jams, and maybe even fix the problem of not having enough air traffic controllers. They’re using a ton of past flight data to teach the AI how to do the job.
The AI and humans are currently working together in this simulated sky to see how well they do. By 2026, they plan to test the AI using real-time flight data but don’t worry, it won’t actually control real planes yet. If all goes well, we might see AI helping humans in the control towers for real after more tests.
More writers sue OpenAI for copyright infringement over AI training
Some big-name U.S. authors, like Pulitzer winner Michael Chabon, are suing OpenAI, saying it used their writing without asking to train its chatbot, ChatGPT. This isn’t the first time OpenAI, backed by Microsoft, has faced such claims. Some other big companies, like Microsoft and Meta, have also been taken to court over similar issues.
While these companies believe they’re making fair use of online content, the authors argue otherwise. They say that OpenAI used their work to teach the chatbot, making it mimic their unique styles. ChatGPT had a major growth spurt, getting 100 million users a month, before Meta’s Threads app took the top spot. The writers want money and for OpenAI to stop these “shady practices”.