Was Deep Seek Over hyped?
There was a lot of hype about China’s artificial intelligence app, Deep Seek, being the most downloaded AI app, overthrowing its competition, Open AI and Meta. But, it now seems like the U.S. AI initiatives are still forging ahead at full speed with their investments into AI.
Meme/hype? Many analysts are now urging caution and skepticism regarding the breakthroughs that Deep Seek’s team has claimed with some analysts saying that Deep Seek is really an optimized, upgraded version of ChatGPT4o - some dispute this characterization.
On February 10th, 2025 - Google AI Lab's CEO Said: "DeepSeek showed no actual new scientific advance."
Initial hype or not, the question remains, will companies spend billions creating new models if anyone can come in and create a better, cheaper AI model for a fraction of the price? How will OpenAI, etc. now see the returns on their giant investments?
Nonetheless, the big thing of course is that Deep Seek is open source and it has prided itself on having a significantly quicker rate in their AI chips, which are produced for a cheaper cost than US AI apps like ChatGpt and Open AI use (they’re using larger and more expensive Nvidia chips) . In fact, in its first stages of development, the company used A100 graphics processor chips from the U.S. company, Nvidia, who’s now placed restriction exports to China on their graphics processor units.
Deep Seek’s initial development focused on language learning models (which are artificially based algorithms that process and understand human languages). They're trained on massive amounts of data and have self-supervised learning. But since then, it’s expanded its usage across various fields like mathematics, science, and qualitative reasoning. In fact, it rivals Open AI’s ChatGPT by more than just its advancements.
Deep Seek’s reasoning model, R1 maps out its entire thought process and the jump to an answer, giving a more in-depth reply and humanistic perspective instead of jumping to auto-response and fact checking without any background. And, Deep Seek surpasses Open AI’s reasoning model o1, beating OpenAi’s math abilities and general knowledge..
In the end - it goes back to Deep Seek being an open source app, available for anyone to view and modify specific components. From MIT’s James O’Donnell, “If researchers subscribed to certain open-source principles and showed their work, they argued, the global race to develop superintelligent AI could be treated like a scientific effort for public good, and the power of any one actor would be checked by other participants. .
So far, developers at Deep Seek have refused to provide the data it’s used for training. How much of Deep Seek’s advancements were over-hyped? Stay Tune
Sources:
AP News: What is DeepSeek, the Chinese AI company upending the stock market? | AP News
Reuters: US looking into whether DeepSeek used restricted AI chips, source says | Reuters
CBS: DeepSeek AI raises national security concerns, U.S. officials say - CBS News
Bloomberg: What Is China’s DeepSeek and Why Is It Freaking Out the AI World?
