fbpx

DeepSeek Has Deep Implications for the Future of Tech

(and the possibilities of disruption of US hegemony)
DeepSeek Has Deep Implications for the Future of Tech
Image Generated with Ai

The last few days of January have been a series of sensational headlines. They started with the US President making a very public announcement of a USD 500 investment in AI infrastructure in the US… Just four days later, a tiny, virtually unknown Chinese AI startup, DeepSeek, was responsible for wiping USD 1.5T off US tech stocks. Nvidia, the designer of the world’s fastest chips, was the biggest single loser. 

So what happened, and what does it mean?

So, for several years, there has been a global arms race in AI development, and one of the interesting things in that race was the US banning the sale of Nvidia’s best chips to China to give an advantage, even dominance, to US companies, and slow China down. Most of this war was not evident, and one of the key events that happened that changed was the public release of ChatGPT by OpenA. Then the race had really started and all the tech giants (Google, Microsoft, META) were in it, alongside hundreds of other companies like Anthropic (Claude). 

And these tech companies were all posting higher and higher stock values as the hype around AI increased.

What’s the big deal about DeepSeek?

  1. It rivals the performance, accuracy, and speed of the top US models and, in some cases, outperforms them.
  2. It is open source, meaning any developer can build on the foundation code.
  3. It is free to use.
  4. It already uses reasoning to solve complex problems.
  5. It uses a fraction of computing power (and thus energy and water) compared to its US counterparts. 
  6. Whereas OpenAI and Google have spent billions of dollars to get to where they are today, DeepSeek spent USD 6M and was only founded in July 2023.

So how did a small team of engineers, all under 30, create something to rival the best that the US has achieved, with vastly fewer resources, and in only 18 months? 

The founder, Liang Wenfeng, saw that the US was about to impose tech sanctions, particularly the high-end chips needed for high-performance AI models, and made two strategic decisions. The first was to buy 1,000s of Nvidia chips and get them to China before the blocks, and the second was to take on the challenge of building something superior to what the US companies had built at a fraction of the cost—both of these brilliant decisions paid off. And instead of trying to replicate other models he gathered a team of engineers and researchers (all from Chinese universities) who believed that they could do it, which they did.

What are the key takeaways from the most turbulent week in AI history?

  • The US is not going to be able to control all AI innovation and business.
  • AI can be rethought and reimagined, architecturally and conceptually, to something better.
  • Restricting access to the core tech forced the Chinese to really innovate.
  • Small teams of high-quality engineers and researchers can beat the biggest teams in the US with a fraction of the resources. 
  • Chinese universities are producing exceptionally talented students and research to rival the best of Harvard, Stamford, and MIT.
  • China can ship high-quality product, fast.
  • The vastly more efficient DeepSeek model will mean reduced energy and cooling needs, which could be a game changer for the overall cost to the environment and to its users.

By their very nature, Black Swan events can’t be predicted, and this one really came out of the blue. While it was clear to many analysts that there was an AI bubble already in place and that there would be some kind of adjustment coming, nobody could have foreseen how swift and decisive it would be, nor that it would come from a small Chinese startup with a new AI model built from scratch on a shoestring budget.

The views expressed in this article belong to the author and do not necessarily reflect the editorial policy of WAYA.

If you see something out of place or would like to contribute to this story, check out our Ethics and Policy section.