Hundreds of billions of dollars later, the entire code needed to reproduce R1 is now available on Github. Code here: https://lnkd.in/gD8QrnGQ ↓ Are you an AI developer? Check out https://AlphaSignal.ai to get a daily summary of breakthrough models, repos and papers in AI. Read by 200,000+ devs.
What do you think about Deep Seek Eric Vyacheslav? They aren't claiming to have developed all of the code themselves, but they are claiming they're results are on older model hardware with just changes to the architecture. People in the financial world are throwing flags on it. With your knowledge in AI, from a glance, is what they did feasible and doable on a shoestring budget? Let's be honest, AI development was supposed to be left Opensource from the beginning, that way new developments and things like this could happen.
Very interesting, Eric Vyacheslav. Just confirms that Gen AI models are becoming commodities. But the models are not the product, so let’s see how this affects the actual AI market.
Tech & Real Estate Entrepreneur
9moI had access to that code a week ago, and quite frankly, this is not the code used to train the R1 model. In this script, they are using reinforcement learning with rewards (GRPO) to fine-tune a language model on the GSM8K dataset.