Thursday, January 30, 2025

Why China’s young super geeks could herald the West’s ‘darkest hour’: Their AI is cheaper, quicker and better… no wonder the world fears a ‘Sputnik’ moment

In October 1957, the Soviet Union launched the world’s first artificial satellite, Sputnik-1, causing shock and alarm across the West.

It had been generally assumed that the US was far more technologically advanced than the Russians, and the rude awakening provided by the ‘Sputnik crisis’ sparked the famous Space Race between the two superpowers.

Now, a leading Silicon Valley tech investor has warned of a new ‘Sputnik moment’. 

That’s how billionaire Marc Andreessen this week described China‘s latest breakthrough in the battle to develop Artificial Intelligence – currently the defining issue for the tech world.

The release of a cheap but remarkably powerful new generative AI platform by a Chinese tech company called DeepSeek has turned out to be just as unsettling for American self-confidence as the appearance of a Soviet spacecraft orbiting the Earth.

Andreessen, a major donor to Donald Trump, described the software – a ‘chatbot’ that allows people to have conversations with it – as ‘one of the most amazing and impressive breakthroughs’ he had ever witnessed. And plenty of people agree with him.

It takes a lot to wipe the smug grin off the faces of Silicon Valley’s multi-billionaire bosses, but DeepSeek has managed to seriously rattle many of them as its new AI model – and the astonishing claims made about it – sparked a bloodbath on Wall Street on Monday, taking a staggering $1trillion (£800 billion) off the value of the shares of major US tech companies.

One of them, microchip manufacturer Nvidia – which only last year was crowned as the world’s most valuable company because of its key role in supplying the chips supposedly needed for AI – plunged nearly 17 per cent in value (the equivalent of $589 billion or £490 billion) in the biggest one-day drop for a single company in US history.

Liang Wenfeng, the founder of Chinese AI startup DeepSeek (pictured right) when he was younger

Wenfeng, who by Chinese media accounts has just turned 40, launched his app in the United States on the same day as Donald Trump 's inauguration

DeepSeek's AI Assistant is said to perform on-par with ChatGPT at a fraction of the price

Why the panic? China has shattered American certainty about which country will lead the world in developing AI.

DeepSeek – run by Liang Wenfeng, a nerdy 39-year-old Chinese billionaire hedge funder with a terrible haircut – has unveiled an AI system which, with certain caveats, matches the achievements so far of America’s most advanced chatbots such as the latest version of ChatGPT. 

And, worryingly for Silicon Valley and Wall Street, DeepSeek says it has demonstrated that low-cost AI can be created without using expensive US technology.

Demand from those wanting to try out DeepSeek’s chatbot (which can be accessed via an iPhone app, or simply on the internet) has been immense, although yesterday it proved difficult to register for it, with the company’s website curiously blaming ‘large-scale malicious attacks’ on its services.

President Trump warned that DeepSeek should be a ‘wake-up call’ for American industries.

It’s particularly galling for him as he has been talking up Silicon Valley’s AI brilliance, recently announcing a $500 billion initiative called Stargate to develop the technology.

Sam Altman, boss of OpenAI, the leading US artificial intelligence developer, has tried to see the bright side: he called the Chinese rival ‘impressive’ and claimed it was ‘invigorating to have a new competitor’.

This was quite a volte-face for a man who only last summer characterised AI development as a ‘race’ between democracy and authoritarianism.

President Donald Trump has branded Chinese AI startup DeepSeek as a 'wake-up call' for US tech titans after fears of upheaval in the AI gold rush rocked Wall Street

AI is already entrenched in our daily lives. It’s employed in everything from mobile phones and Amazon’s ‘virtual home assistant’ Alexa to delivery drones and driverless cars.

Sceptics have warned that AI has been ridiculously over-hyped, but others say it could revolutionise all manner of services, including health care and defence, albeit at the potential cost of millions of people’s jobs.

DeepSeek is a so-called large language model, which harnesses the vast trove of information on the internet to create super-intelligent computers that can outperform humans.

DeepSeek’s chatbot has been hailed as a giant leap forward because it is said to achieve the same results as the equivalent AI models in Silicon Valley, but more cheaply, quickly and efficiently.

It also consumes less energy (a major factor given that AI requires enormous computing power), and requires fewer and less advanced chips than those powering the AI being developed in the US. 

While America’s OpenAI says it spent $1billion (£800 million) ‘training’ ChatGPT, for example, DeepSeek reportedly spent less than $6 million (£4.8 million) on its last chatbot. 

And although some advanced US models charge users, DeepSeek’s chatbot is ‘open source’, meaning anyone around the world can use it for free. Little wonder there is such demand.

China has said it aims to become the global leader in AI by 2030, but many in the US believed Washington had put an end to that notion when the Biden administration in 2022 started putting in place controls on exports of high-powered chips to China.

The new Chinese chatbot launched last week and quickly overtook its rival ChatGPT to become the downloaded free app on Apple's US App Store

However, it appears DeepSeek not only got around those controls by finding a loophole in US regulations that allowed them to keep purchasing advanced chips before Washington finally closed the door. 

It also adopted a policy of hiring super-bright young computer engineers.

DeepSeek’s boss, Mr Liang, who made his fortune using AI to spot good investments on stock markets, believes that inexperienced workers are more likely to ‘think outside the box’. 

So it proved as those workers developed a way of streamlining the traditional, very time-consuming method of building an AI chatbot by processing every last gobbet of information on the internet.

Instead, the DeepSeek system relies on knowing where to look for answers online – a process that may take a little longer but is far more efficient.

To many, especially those who instinctively distrust something that comes out of China and appears designed to damage business rivals in the US, the stunning success of DeepSeek sounds a little too good to be true.

Some suspect that the Chinese government provided financial backing to DeepSeek, founded in 2023, although experts who’ve looked at the company’s data say that it appears the business genuinely found a way of cutting costs.

It’s also been claimed that the company hasn’t been entirely honest about the microchips it used, because it managed to stockpile tens of thousands of the advanced American Nvidia chips before those restrictions came into force.

DeepSeek's arrival destabilized the AI market and caused the US stock market to hemorrhage $1trillion on Monday

OpenAI CEO Sam Altman wrote that R1, one of several models DeepSeek released in recent weeks, 'is an impressive model, particularly around what they're able to deliver for the price'

However, perhaps the most powerful reason to worry about DeepSeek’s AI is that it may provide China with a hugely productive means of spying on the West.

Under a 2017 Beijing law, all Chinese organisations are required to ‘cooperate with national intelligence efforts’ – a fact that hardened Washington’s opposition to the Chinese-owned video-sharing giant TikTok (although the company has denied sending US user data to the Chinese government).

The popular service has been under a court-ordered ban in the US since earlier this month, due to government concerns it is hoovering up users’ private information and passing it to the communist regime, significantly weakening US national security.

DeepSeek’s chatbot is itself collecting a vast amount of information and sending it back to China, where it is headquartered in the city of Hangzhou, one of the country’s tech hubs. 

This much the company freely admits: according to its English-language DeepSeek privacy policy, ‘We store the information we collect in secure servers located in the People’s Republic of China.’

Meanwhile, even the information that DeepSeek sends back in the other direction has a pro-Beijing bias. Users have already discovered that its chatbot declines to answer politically sensitive questions such as those about the 1989 Tiananmen Square massacre or President Xi Jinping, while at the same time putting a pro-Beijing gloss on other controversies like the mistreatment of the Uyghur ethnic minority.

In China, DeepSeek has been hailed as a national triumph that could change the country’s destiny – and which has plunged Silicon Valley into its ‘darkest hour’. 

Yet the new AI platform may not be all it seems, and – like the Soviets being beaten to the Moon in 1969 – China may not win the AI race.

But like Sputnik, DeepSeek has delivered a short, sharp shock to the notion of American ‘exceptionalism’ that its new president and the tech titans who have embraced him love to trumpet.

This post was originally published on this site

RELATED ARTICLES
Advertisements

Most Popular

Recent Comments