We reproduced DeepSeek R1-Zero in the CountDown game, and it just works
Through RL, the 3B base LM develops self-verification and search abilities all on its own
You can experience the Ahah moment yourself for < $30
Code: https://t.co/UcGKN2SVGj
Here's what we learned 🧵
Our science team has started working on fully reproducing and open-sourcing R1 including training data, training scripts,...
Full power of open source AI so that everyone all over the world can take advantage of AI progress! Will help debunk some myths I’m sure too.
Thanks @deepseek_ai!
Pragmatically, we can say that AGI is reached when it's no longer easy to come up with problems that regular people can solve (with no prior training) and that are infeasible for AI models. Right now it's still easy to come up with such problems, so we don't have AGI.
AI is only as good as the training data used
great example from Rafael Cosentino
AI generated image of "salmon swimming down a river"
human oversight and insight is needed to understand this is not right; the AI doesn't know this is not right https://t.co/CYDbqRlFwt
Introducing Lexica – a search engine for AI-generated images and prompts.
Every image has a prompt and seed, so you can copy and remix anything for yourself.
Hopefully this makes Stable Diffusion prompting a bit less of a dark art and more of a science!
https://t.co/0YdmzHqqY0