DeepSeek: Industry disruptor or trojan horse?

DeepSeek's meteoric rise and the global market turmoil that has ensued marks a huge milestone in AI development; it challenges industry norms, it disrupts geopolitical dynamics, and it leaves US tech leaders scratching their heads (while simultaneously checking their stock portfolios, I would assume).

At the time of writing, DeepSeek’s AI model, built with far fewer resources than its western counterparts, has wiped almost $600bn off NVIDIA’s market value. The single largest drop in US stock market history.

DeepSeek’s model is reportedly 20 to 50 times cheaper to use than OpenAI's model. OpenAI told The Financial Times that it has evidence DeepSeek has used “distillation” methods to build their chatbot. Essentially, the accusation is that DeepSeek has trained their model on OpenAI’s model.

Open source AI and democratisation?

In a nutshell, by leveraging open-source frameworks and optimising low-end hardware, DeepSeek challenges the idea that AI progress requires vast capital and elite infrastructure. Their remarkable success in doing so suggests the future of AI may well be determined not by those with the most power and money, but by those who can do more with less. Sounds good, right?

DeepSeek’s rapid ascent also forces tech firms to reconsider their approach. Western AI tech leaders have mostly focused on proprietary models (i.e. AI models that are developed and maintained by a single company) and closed ecosystems, yet DeepSeek’s reliance on open-source collaboration offers a powerful communal alternative.

If this model proves sustainable, it could trigger a shift towards leaner, more efficient AI development. In which case, tech giants may need to embrace more open innovation models or risk being outmanoeuvred by agile, cost-efficient challengers like DeepSeek.

However, open-source AI isn’t necessarily all sunshine and rainbows. While it promotes collaboration, it also means breakthroughs can be exploited by state actors, with their own (potentially nefarious) motives. Perhaps DeepSeek’s open approach is temporary, a means to accelerate development before restricting access. In which case– not so democratic.

A wake-up call for the US

The success of DeepSeek crucially exposes potential flaws in the US’ AI strategy. Washington has relied on restricting China’s access to high-end semiconductors, assuming this would hinder AI progress. Yet DeepSeek’s ability to achieve strong performance without cutting-edge chips suggests that software ingenuity can overcome hardware barriers.

This begs many questions, mine include: Are US export controls working? Can the West maintain its AI leadership when entry barriers are falling? Is the ultimate goal of AI really achievable without the use of high-end chips and huge investment, or can DeepSeek replicate its current success to achieve industry domination?

One thing is for certain though, the US needs to rethink its focus, to balance hardware dominance with greater emphasis on software agility.

Hype v reality

What DeepSeek has achieved this week is very impressive, but like with all riotous newcomers, the hype warrants some scrutiny. Experts warn that its long-term sustainability is unproven, particularly in scaling, security and ethics. The political backdrop also raises genuine concerns about government influence over AI development and data security.

Whether or not DeepSeek is truly a symbol of China’s new AI dominance will depend on if it can maintain its momentum and deliver on its promise. Ultimately, it takes time to transition from an emerging player to an industry powerhouse.

Open wide and shut

Obviously, DeepSeek aligns with China’s broader ambition to lead in AI. By proving that high-performance AI can be achieved without Western chips, China gleefully reduces its reliance on US technology and much like the Instagram of a person whose career and fitness trajectory levels up after divorce, China is showing off here.

This also broadly fits right into Beijing’s strategy of developing a self-sufficient AI ecosystem, including domestic chip manufacturing.

However, DeepSeek’s “openness” may be a short-term tactic to attract developers and refine its models before restricting availability. If true, this could be a strategic move to undermine Western AI dominance.

By taking the lead in open-source AI, and expecting to see it ‘clone’ more AI applications from Western companies in due course, China has an opportunity here to set global standards, attract developers, and gradually exert control over the field. If DeepSeek follows this path, it truly could reshape the balance of power in AI.

A gamechanger

DeepSeek’s rise presents a plethora of opportunities and risks in equal measure. It proves that AI breakthroughs can emerge from resourcefulness and collaboration, not just from massive investment. It also forces Western firms and governments to rethink their investment heavy AI strategies. However, the race to Artificial General Intelligence, is still focusing AI leader minds, and the jury is out on whether this can be done without high-end infrastructure and significant investment.

DeepSeek also raises huge geopolitical and ethical concerns. Is it a sign of genuine democratisation, or part of a long term play for China’s technological dominance?

The coming months and years will determine whether DeepSeek is an innovative industry disruptor or a strategic Trojan horse. One thing is clear: the AI race is entering a new, exciting and turbulent phase.

‍