Orchestrating agents to build harnesses for personal assistance and value investing. Previously worked on Humanoid Robots & Deep Reinforcement Learning.
@karpathy Amazing interview as always! Regarding the jaggedness of LLMs: aren't humans completely jagged too? I think it makes LLMs even more human-like. 🧵👇
The war in Ukraine showed that Russia is not army Nr. 2, but it will make it army Nr. 1.
They are building a Russian #Starlink and have put their whole economy on a war footing.
#UkraineWar#DefenseTech#EricSchmidt
https://t.co/aALzZRbNUY
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
I love @Gemini_Notebook! With 70+ notebooks, I have to scroll through all of them to find the right one when sharing a webpage or YouTube video with #GeminiNotebook.
Please add a search bar to the Share View! ����🙏
@OpenAI Amazing, thanks! How does this affect Luna usage limits in OpenClaw setups when using a ChatGPT Plus subscription via OAuth? Will we get higher Luna limits there as well?
#OpenClaw#OpenAI
@karpathy@Etched@karpathy How risky do you see Etched's pure Transformers bet? Do you see Transformers dominating for the next few years, or is it likely that other architectures will evolve?
@Etched, how quickly could you pivot in this case?
Btw, what you built is pure genius! 🚀
TTS Arena is live. ��️
A blind benchmark for text-to-speech, rebuilt from the ground up - listen to two anonymous models, pick the one that sounds more human, and watch the leaderboard move.
Now with the latest models. Come vote ⬇️
Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
@karpathy LLM jaggedness is just RLHF tuning them for high-value work (like refactoring 100k lines of code) over trivial common-sense puzzles. Like humans, they get exactly what we optimize the data for!?
@karpathy Amazing interview as always! Regarding the jaggedness of LLMs: aren't humans completely jagged too? I think it makes LLMs even more human-like. 🧵👇
@karpathy It all comes down to what the neural network (biological or artificial) is tuned for. Human brains optimize for our daily professional focus. We get hyper-specialized in our specific daily environments.
@karpathy Or take parenting: someone can confidently lead corporate strategy, only to come home and be completely overwhelmed by the structural engineering of a baby’s diaper and the buttons on baby clothes. 👶🛠️
@karpathy A simple example of human jaggedness: A tech professional can architect highly complex data systems but completely fail to realize laundry must be sorted by color and temperature. 👕