Summary:
The launch of Chinese AI application DeepSeek in the U.S. has raised national security concerns among officials, lawmakers, and cybersecurity experts. The app quickly became the most downloaded on Apple’s store, disrupting Wall Street and causing a record 17% drop in Nvidia’s stock. The White House announced an investigation into the potential risks, with some lawmakers calling for stricter export controls to prevent China from leveraging U.S. technology.
Beyond economic impact, experts warn DeepSeek may pose significant data security risks, as Chinese law allows government access to company-held data. Unlike TikTok, which stores U.S. data on Oracle servers, DeepSeek operates directly from China, collecting personal user information. The app also exhibits censorship, blocking content on politically sensitive topics like Tiananmen Square. Some analysts argue that, as an open-source model, DeepSeek may not be as concerning as TikTok, but critics worry its widespread adoption could advance China’s influence through curated information control.
Distilling OpenAI and Llama models probably also helped quite a bit
Although I must admit, that the architectural changes are pretty cool
but I have to add, that I’ve just started reading into the topic a few weeks ago and don’t really have any real practical experience, besides checking out some huggingface docs I got linked yesterday and stupid me hasn’t thought about looking there…
So everything I say is probably bullshit o:-)
Sure it made the training process faster, but this still takes a fraction of the energy to generate a single output compared to other LLMs like ChatGPT or Llama. Plus it’s open source. You can’t discredit a technological advancement for building upon previous advancement, especially when doing so with transparency.
As I said, the architectural changes are quite cool
As far as I’ve understood it mostly comes down to splitting it up into multiple expert systems, so you don’t need to activate the complete system with every request
But I’ve only scratched the surface…
Also, open source… The weights are made publicly available.
None of the training data or systems
Edit: regarding “open source”:
Also Meta’s Llama is on huggingface, just like deepseek. I still wouldn’t talk about transparency here