Last week, it took only a day for the frantic AI fever dream of the past two years to break – dramatically, and from a direction few would have guessed. A tiny year-old Chinese AI start-up, DeepSeek, showed powerful AIs could be developed, built and run at a fraction of current costs.
Ever since US company OpenAI’s conversational AI, ChatGPT, launched in late 2022, interest in the complex and costly AI realm has surged. Hardly a week passes without artificial intelligence grabbing headlines and prompting vigorous debate. Hype and hyperbole abound. One day AI might be touted as the single most transformative technology of this century; the next, denigrated as a biased, ethically-questionable, existential risk to humankind.
The truth probably lies between. AI undoubtedly will revolutionise whole sectors of work, research, and daily life, in ways we still cannot foresee. And it does pose serious, potentially dangerous risks, especially if poorly understood, overhyped, and under-regulated. These challenges must be carefully and ethically navigated.
Until now, because AI has been astronomically expensive to develop, it has remained the exclusive territory of a small number of the world’s most powerful, well-funded companies. Hundreds of billions of dollars have gone into developing the most advanced AIs.
RM Block
But this week, DeepSeek demonstrated its models performed similarly to those of its famous competitors, but are cheaper to build, with lower infrastructure demands. DeepSeek is more efficient, requires fewer chips, and is 30 times less expensive to run. Within a day, AI company shares slumped.
Questions remain about DeepSeek’s true development costs. But the company nonetheless has shown that AIs can be built and run for a fraction of Silicon Valley’s lush, venture-funded budgets. And by freely offering its technology, DeepSeek has torpedoed the plush business models of the AI giants. AI remains the most exciting tech sector to watch, but for new and unexpected reasons.
