Cheaper, more accessible AI program catapults Hangzhou startup to stardom
Unlike the closed systems of major artificial intelligence companies in the United States, the Chinese startup sensation DeepSeek has embraced an opensource approach.
That path is seen by some as fostering collaboration and experimentation that could lead to faster breakthroughs. Its web-based chat interface and mobile app are free to use.
The strategy contrasts sharply with companies like OpenAI, which keep their models proprietary and charge monthly fees for access.
READ MORE: Experts: DeepSeek sets benchmark for Chinese firms
DeepSeek's approach has garnered strong support, with opensource advocates saying that the US could advance by embracing the cheaper, more accessible strategy.
The company has demonstrated that AI breakthroughs can be achieved through innovation rather than sheer computing power, challenging the established paradigms of AI development and sending shock waves through Silicon Valley, Wall Street and Washington.
DeepSeek, based in Hangzhou, Zhejiang province, has captured worldwide attention over the past two weeks with the release of two large language models that rival the performance of the dominant tools developed by US tech giants- but at a fraction of the cost and with far less computing power.
Jan 20 saw the release of Deep-Seek-R1, an open-source "reasoning" model that employs a "chain of thought" approach similar to OpenAI's most advanced large language model o1, which was unveiled late last year.
Meta's chief AI scientist Yann LeCun emphasized in a recent Threads post that open-source models are now surpassing proprietary ones.
Bill Gurley, a veteran tech investor, challenged the adversarial narrative surrounding DeepSeek, saying: "Why are they 'enemies'?No one that works at DeepSeek is an enemy of mine. The fact that they love open-source makes me think I would gel with them quite well."
Rather than viewing DeepSeek as an adversary, US companies should see its breakthrough as an opportunity to pursue innovation in a different direction, experts say.
"If there's a lesson from DeepSeek's triumph, it's this: Be wary when the route to progress is simply spending more money. This path fosters no innovation, and your poorer competitors will be forced to get creative, work within their constraints, and eventually… they'll win. Spending is not innovating," data scientist Drew Breunig told Defense One, a news site focused on national security issues.
Within days of its release, DeepSeek's AI assistant mobile app quickly rose to the top of Apple's App Store charts, outranking OpenAI's ChatGPT mobile app.
The prevailing notion in AI development has been that leading-edge large language models require significant technical and financial resources. Leading AI developers such as OpenAI have attracted billions of dollars in investment by arguing that more data, larger models and more computing power would lead to more advanced products.
However, DeepSeek has challenged this assumption. The company achieved remarkable results with an investment of less than $6 million and approximately 2,000 less sophisticated chips.
ALSO READ: DeepSeek is likely to catalyze big shifts in global AI ecosystem
For comparison, Meta's latest Llama 3.1 model required 16,000 of Nvidia's advanced H100 chips for training — chips that are currently prohibited for use by Chinese companies under US export controls.
Facing the constraint of high-performing chips, DeepSeek developed creative methods to train its models on less advanced chips through optimizations in model architecture, training procedures and hardware management.
Kat Duffy, senior fellow for digital and cyberspace policy at the Council on Foreign Relations, said that these innovative methods highlight a downside to the US'approach to AI.
"The focus in the American innovation environment on developing artificial general intelligence and building larger and larger models is not aligned with the needs of most countries around the world," she said in an article published by her organization. "The more the United States pushes Chinese developers to build within a highly constrained environment, the more it risks positioning China as the global leader in developing cost-effective, energy-saving approaches to AI,"Duffy said.
