Deepseek R1
In recent days the US stock market has been heavily disrupted by the news that perhaps Silicon Valley’s advantage in the race to AGI is not as secure as it seemed. In a fascinating break from the standard LLM development narrative, a Chinese hedge fund diverted a portion of the GPU cluster it had built to run financial algorithms to the training of a reasoning model in the vein of OpenAI’s o-series. The controversy comes is significant : I downloaded a 14 billion parameter version of Deepseek onto my M3…read more