DeepSeek Fever - my personal take
A few days ago, NVIDIA experienced one of its most significant losses in capital markets to date. A major factor contributing to this downturn is the rise of the Chinese AI model "DeepSeek." This article evaluates DeepSeek from a user’s perspective, using OpenAI’s ChatGPT as a benchmark. Despite its significantly lower development cost compared to models like ChatGPT, DeepSeek delivers similarly competitive performance in areas such as text and code generation. It stands out particularly in the speed of its responses. Although still in its early stages, the model shows remarkable promise, offering a glimpse into the future of AI. The rapid pace at which DeepSeek is advancing in the AI field raises intriguing questions about how quickly the technology is evolving.
This makes me think about how Deepseek will not only affect the future of AI and Software Engineering, but also fields that do not relate to tech at all. With Deepseek taking only taking ~$5 million to develop versus OpenAI's ~$1 billion, how will Western companies compete against the race for LLM's?
Indeed it was a few days of a very interesting situation regarding DeepSeek. What surprised most of the AI investors are the fact that DeepSeek was able to produce a good model with hundred times lower the cost. In addition, there exists the ban on AI chip export. So DeepSeek essentially bootstrapped their way to their version of a ChatGPT. Funny thing is people bought up Nvidia's stock today by 6% after the 13% drop yesterday. Perhaps as the situation develops, market realized the selloff was overdone.