To put this advancement in perspective, the previous leader MindsAI, the previous leader, MindsAI, achieved 55.5% just 1 month ago.
FYI: You can take the test yourself and see how you score compared to o3 https://t.co/iXo0Mm1fkE.
Check out the chart below to see how o3 stacks up against other leaders in ARC-AGI evaluations. 👇
(Chart: ARC-AGI SOTA, Dec 2024)
Big news in AI! 🚀 OpenAI just released o3, their latest model, and it’s breaking records in adaptability and generalization. Here’s why this matters:
• 75.7% on ARC-AGI’s Semi-Private Eval under $10k compute limits (previous leader was at 55.5%).
• A high-compute version hit 87.5%, showing how much potential this model has.
To compare, a 2024 NYU study found 98.7% of ARC tasks are solvable by humans, with average human scores ranging from 64.2% to 76.2% depending on the dataset. o3 is getting closer to human-level performance on these incredibly tough tasks!
This isn’t just another upgrade—it’s a genuine breakthrough. o3 fixes a key issue in AI: the inability to adapt and solve new problems at test time. Its ability to recombine knowledge is a game-changer and sets a new benchmark for AI systems.
I can't wait to test it to see if we can finally solve some challenges for a project I'm working on. 😀
This is an exciting step forward for AI. Want to dive into the details?
Read the full report here:
https://t.co/YJwaVDfqhk
@SoftBank’s $100B investment in U.S. AI projects is a huge move for the future of tech—focusing on AI, chips, and data centers. Announced with Trump, the plan includes creating 100,000 jobs and will be rolled out before the end of his term. Big things ahead! 🚀🤖
This investment is all about building what AI needs to grow—better chips, powerful data centers, and the systems that make it all work. With 100,000 jobs in the mix, it’s a big bet on AI leading the way for innovation and keeping the U.S. ahead in the global AI race.
AI is already changing how we live and work, and moves like this open the door to even more breakthroughs in things like healthcare, technology, and everyday life. It’s exciting to think about what’s next!
#AI #ArtificialIntelligence #Tech #Innovation #FutureOfTech
This is an interactive demo powered by @EveryAnswer! It’s completely free to try—no sign-up or anything required. Just a fun way to spread some holiday cheer and showcase the magic of AI. Let us know what you think! 🎅✨
🎅✨ Get in the holiday spirit with AI Santa Claus! Ask Santa anything about Christmas—reindeer, gifts, cookies, you name it! 🎄🎁
Try it now and share the magic with your friends!
👉 https://t.co/jnhjiGdqh5
#HolidayCheer#SantaClaus#AISantaClaus#FestiveFun#AI #ArtificialIntelligence @EveryAnswerAI
I was just asked to finish the statement "If I were @elonmusk for a day,..." Here's my answer.
"If I were Elon Musk for a day, I’d start a company dedicated to tackling the cybersecurity challenges that will arise from superintelligence and quantum computing. The technology and methods (passwords, encryption etc.) we use today will be easily beaten. These are threats not only to every business, human, connected device, and AI agent, but also to “my” technologies like Telsa cars, robots, and other innovations. Developing solutions to these challenges could not only protect society but also position the company as a market leader in what will be a massively profitable space in the near future (think recurring revenue from trillions of billable endpoints)."
What would you do if you were Elon for a day?
Wow, Gemini 2.0 is here! 🎉 @GoogleDeepMind just dropped their newest model, and I’m still processing how much AI innovation we’ve already seen this month. 🤯
We’re only 4 days into @OpenAI’s 12 Days of AI (o1 and Sora already blew my mind), and December’s already packed with huge updates like @xAI’s Grok Aurora image generation and @Meta’s Llama 3.3. 🔥
#AI #MachineLearning #Gemini2 #ArtificialIntelligence
Welcome to the world, Gemini 2.0 ✨ our most capable AI model yet.
We're first releasing an experimental version of 2.0 Flash ⚡ It has better performance, new multimodal output, @Google tool use - and paves the way for new agentic experiences. 🧵 https://t.co/4V1gJ9yngt
Just tested an early preview of EveryAnswer’s new OCR.
I uploaded Wizard of Oz (yellowed pages, fading, uneven scans, illustrations). In about 10 seconds, it OCR’d 312 pages & added them to the Expert’s training.
Our team knocked it out of the park again! This will make training Experts faster and easier—even with older, messy data. 🚀📚 #AI #OCR #ArtificialIntelligence
🎉 We’re thrilled to introduce the Pokémon Pokédex! An AI-powered tool to explore Pokémon stats, evolutions, battles, and lore—instantly! Ask it anything about Pokemon and see if you can stump it.
https://t.co/5ZIzUhglfc
#Pokemon#Pokedex#PokemonGo#AI#artificalintelligence
Our team at EveryAnswer has brought the ultimate AI-powered Pokémon Expert to life. We’ve had friends and fans test it, and they’re blown away! Test it yourself—can you stump it with your Pokémon knowledge? 🎮
https://t.co/Fd80PHQJds
#Pokemon#AI#Pokedex#artificalintelligence #PokemonGo #GottaCatchEmAll
EveryAnswer is officially live and open to everyone! (no more waitlists, finally 🥳)
Hope you think it was worth the wait—can’t wait to hear what you think! 🫶 Try it here: https://t.co/7CAFE3ED7t
#AI#futureofwork#ArtificialIntelligence#startup
We’re building on our commitment to Trustworthy AI by introducing new capabilities across our tech stack to strengthen the security, safety, and privacy of AI.