Big week for DeepSeek: A Chinese artificial intelligence company released its model, R1, about a week and a half ago, sending American chipmaker Nvidia's stocks plummeting. What's going on here? And why does R1 matter?
Well, the model is profoundly sophisticated. It "can mimic the way humans reason," says Bloomberg. Developed in 2023 by a quant hedge fund, the app distinguishes itself in a few major ways—"articulating its reasoning before delivering a response to a prompt" (per Bloomberg) as well as being open-source (though "open-weight" might be a better way of putting it, per people who know more about the technical side than I could ever hope to). It's also much more efficient compared with other models, using way less compute—the hardware required to train and run these programs—and thus requiring less money.
DeepSeek "claims its R1 release offers performance on par with the latest iteration of ChatGPT," reports Bloomberg. "It is offering licenses for individuals interested in developing chatbots using the technology to build on it, at a price well below what OpenAI charges for similar access." (Quick explainer video here, from my friend Melody Kim.)
