By Harry Mayes, Research Technician, Neuroscience
In January, Chinese AI company DeepSeek announced the latest model of their language AI, R1, which claims to rival the capabilities of the latest rendition of Chat-GPT whilst costing far less. The news caused AI stocks to crash and raised concerns regarding the direction of Western tech investment and the rate of productivity compared to our Eastern counterparts.
The R1 model is on par to the o1 model of Chat-GPT, that was released last year. The difference is that the cost of training for the R1 model is estimated to be just $6 million, compared to the training cost of GPT-4, which the Open-AI founder alluded to costing 'over $100 million'. DeepSeek's founder, Liang Wengfeng is said to have paired NVIDIA chips with cheaper chips to enhance AI computing processes to be more efficient. This comes amid restriction on exports to china of AI chips by the (until recently) market leader, NVIDIA. These restrictions incited DeepSeek to stockpile these chips and birthed the innovation of utilising the two chips in combination.
These cheaper costs are thought to be due to using less memory than other AI's, reducing costs per task. Flaunting cheap prices and equal performance to the market-leading Chat-GPT, it comes as no surprise that DeepSeek became the most downloaded free app on the iOS App Store upon release.
Perhaps the most important lesson from the rise of DeepSeek is the way Western nations approach economic growth needs to be drastically reformed. DeepSeek's accomplishment has proven that large budgets and high-end technology are not the only way to advance. Considering the self-perpetuating nature of AI, what is considered as contributing towards growth may be under heavy questioning; gone are the days when sheer capital investment alone drove progression. Growth strategies like workforce expansion or enhancing raw computational power could soon prove insufficient. With models like DeepSeek continuously learning and redefining themselves (even perfecting their own architectures) the best bet moving forward may lie in innovation in training methodologies, data efficiency, and computational breakthroughs. This will enhance factors like the rate and capacity of learning in future AI models. This is especially relevant as AI is increasingly beginning to penetrate every facet of life and work across many industries. The success of DeepSeek has shown that growth, particularly within AI, can be achieved with innovation and limited recourses, enough to plunge NVIDIA's stock value by 17% and haemorrhage their market cap $600 billion.
Overnight the release caused NVIDIA to plummet from the most valuable company in the world to third place. To ensure that Western nations keep up, we must come to understand that throwing money at projects is not a sufficient solution; we need to make selective and smart investment. Widespread education on AI and programming can help ensure we are equipped with innovators and pioneers that help revolutionise the technology in this sector. This is more essential than simply having the best tech, as DeepSeek has clearly shown.
When asked why DeepSeek shocked so many in Silicon Valley. Liang said
"Their surprise stems from seeing a Chinese company join as an innovator, not just a follower – which is what most Chinese firms are accustomed to."
The West need to rethink how they plan to ensure growth with maintaining a market-leading status in this sector, or face becoming the 'followers'. With Australia and Italy having already banned DeepSeek over national security and privacy concerns, it is of vital importance to the West that they increase the pace of their innovation and remain as market leaders.
The news of DeepSeek's rise came as a shock to the sector, especially considering the heavy investment from the US under the new Trump administration. The President himself called it a 'wake-up call' in regard to Western companies and levels of productivity.
China, who have long aimed to develop technology that does not rely on the West, has had its goals affirmed and encouraged by the rise of DeepSeek. Chinese media have celebrated the rise of DeepSeek as a victory over the West, claiming it is 'overturning' the stock market.
Among the mentioned concerns, the divide between superpowers over AI heightens concerns of 'tech isolationism'. This refers to an attitude of 'technological sovereignty', with nations reducing their dependence on foreign innovations to stay 'cutting-edge'. While isolationism might improve national security, they run the risk of limiting the scope for global cooperation, threatening productivity. The unique internet infrastructure that Russia has developed, banning Western websites and media, is a case in point. Their distancing has reduced innovation and cooperation, reduced growth and heightened geopolitical tensions.
With a tariff war well underway and conflict the closest it has been to the West since 1945, it is essential to maintain international cooperation and relations. The rise of DeepSeek has highlighted this need and held a mirror to Silicon Valley and the leaders of the West that our means of growth need drastically re-evaluating if we are to keep up with international competition. And if credibility is given to concerns over national security and privacy, it is more important than ever that Western productivity is able to support our position as leaders in technological innovation.
Featured image courtesy of James Cameron/ Unsplash
