DeepSeek’s breakthrough is a pivotal moment for the democratization of AI

This Chinese model is far from unknown today. It has risen to the top spot on Apple’s app store and single-handedly caused the Nasdaq to slip by 1.5% on a single day, with Nvidia, the bellwether of AI stocks, crashing 17%.

That DeepSeek’s latest AI model performs almost as well as OpenAI’s or Google’s creations is not why it is being hailed as the second coming of AI. The reason is how little time and money it took. 

Also Read: What America’s technology denial and China’s AI success imply for India

Built as a ‘side project’ by High Flyer, a Chinese hedge fund founded by Liang Wenfeng, it took just two months and $5.6 million, as opposed to the hundreds of millions spent by OpenAI and others. As claimed, it used last-generation H800 GPU chips, having been denied Nvidia’s latest by US sanctions; it needed just 2,000 of them, compared to the 100,000 or so that big US models required. The shock and awe does not end there. As it is a small-sized model, it can even run on a high-end gaming computer on the edge, rather than on the machines of a gigantic electricity-guzzling data centre. 

Most importantly, the model is entirely open-sourced, with all its technical details published in a white paper, which makes it easy to work on for any company anywhere in the world.

How did DeepSeek’s team perform this miracle? As Perplexity’s founder Aravind Srinivas succinctly put it: “Necessity is the mother of invention." Its creators did not have fancy GPUs and big-ticket funding, so they figured out entirely new techniques to build an AI model. 

An innovative approach they used is called distillation; they built it on top of ChatGPT and others, instead of training it on raw data. They optimized memory usage to 75% instead of overloading it for unnecessary precision; they processed phrases rather than words, doubling speed, and taught it to activate only the required parts of the model to answer a question, not all of it.

What the DeepSeek shakeup means

So, what does this mean for AI and its stakeholders? DeepSeek is another ChatGPT moment for AI. It will change the whole AI narrative. Companies and investors will realize that we do not need billions of dollars and the latest GPUs to create better AI. It is a huge boost to open source, as it has started matching closed models in performance. Big Tech players like Meta and Microsoft will recalibrate their offerings to include these newer models, particularly to power their consumer and enterprise products at lower cost.

Also Read: Silicon Valley’s blind spots have been exposed by China’s DeepSeek

Nvidia has suffered a huge blow, but it will come back strongly, selling a thousand GPUs each to tens of thousands of customers, rather than hundreds of thousands of them to a handful of them. It will spur innovation at OpenAI and Anthropic, as they seek to differentiate their models and the thousand small-model flowers expected to bloom around them. They will go deeper into thinking, reasoning and the pursuit of artificial general intelligence (AGI).

For startups and innovators across the world, this is manna from heaven, as they will pay cents for the dollars they were paying OpenAI and Anthropic. It is great for India too, as it is now clearly feasible for the country to create its own LLMs, a task it was grappling with. Indian AI startups should grasp this opportunity to create AI applications cheaply for themselves and the world.

To be sure, there are definite problems. It is too early to state whether DeepSeek is too good to be true and all its claims are valid. Chinese state controls limit what it can say or not about Taiwan or Tiananmen Square, for instance. But its open-source nature can help companies mitigate both problems. In all, this is a big moment for AI democratization. AI prowess was getting centralized among a few trillion-dollar companies.

Also Read: Rahul Jacob: DeepSeek’s big AI shake-up holds policy lessons for India

Satya Nadella, the cerebral CEO of Microsoft, put it well in an X post (bit.ly/3EnJd6J). He spoke of the Jevons Paradox and how the DeepSeek moment will be a boon for AI. 

This paradox states that increased efficiency, counter-intuitively, leads to higher resource consumption. The invention of the super-efficient steam engine exponentially increased the demand for coal. The same will happen in AI, with super-efficient DeepSeeks making AI more efficient and accessible. This will see its use skyrocket, turning it into a commodity we can’t get enough of.

The author is a founder of AI&Beyond and the author of  ‘The Tech Whisperer’.