By Ange Lavoipierre, ABC
Some (though not all) of DeepSeek's code and technical explanation is open source, meaning it can be viewed, downloaded and used by anyone. Photo: Getty Images via AFP
Analysis - If you buy the hype about China's new chatbot DeepSeek, building a top AI model might cost you a similar amount of time and money as your dream home, replete with a tennis court, swimming pool and water views - roughly two months hunting and $US5.5 million.
Depending on the location, inventing a new AI might even be cheaper and faster than breaking into the property market.
There are reasons to believe that DeepSeek's creator, a Chinese company by the same name, has exaggerated its thriftiness - more on that shortly - but ignoring the global freak-out it's triggered would be a mistake.
DeepSeek rocketed to number one in app stores on the weekend and wiped more than half a trillion US dollars off the world's largest company and leading AI chip supplier, Nvidia, in Monday trading.
It was a public routing of the world's hitherto unshakable confidence in Silicon Valley's best labs, which have always insisted that the only path to a shining AI future is paved with as many expensive computer chips as possible.
DeepSeek now appears to have debunked one of the tech world's holiest scriptures, achieving similar success with far fewer and far older chips.
"It marks the end of US dominance of the AI race," says Toby Walsh, chief scientist at the AI Institute of the University of New South Wales.
"It's like running the four-minute mile," he says.
"When Roger Bannister first ran the four-minute mile, it demonstrated it was physically possible."
And unlike Bannister, DeepSeek has handily published most of an instruction manual on the internet.
Some (though not all) of DeepSeek's code and technical explanation is open source, meaning it can be viewed, downloaded and used by anyone.
"It's not going to stop just with this Chinese firm DeepSeek, that has managed to match the best of the West," Professor Walsh says.
How over-hyped is the DeepSeek story really?
As for exactly what DeepSeek has done, some of the claims are shaky or yet to be thoroughly tested.
Many experts are deeply sceptical of the touted $US5.5 million ($8.8 million) price tag and two-month timeline.
"I think it needs to be questioned," says Jenny Wong Leung, senior analyst at the Australian Strategic Policy Institute (ASPI).
The AI model in question, DeepSeek-R1, isn't the company's first effort.
"The $5.5 million in the last two months are specifically attached to the latest version, and builds on the shoulders of previous investment dating back to 2021," she says.
The US government introduced export restrictions for Chinese companies on the best AI chips, known as GPUs, in 2022, but she points out that DeepSeek's founder had reportedly been stockpiling them prior to the ban.
In other words, beware of the DeepSeek spin.
"This is likely a framing to capture maximum publicity," says George Buchanan from RMIT.
"[It's] very likely there is strong support from the Chinese government to emphasise Chinese innovation."
So DeepSeek's engineers didn't knock their chatbot prodigy together during a lunch break, but that doesn't diminish the feat for Professor Walsh.
"It might have been $12 million or $24 million, but whatever it was, just matching [OpenAI's latest model] is impressive," he says.
When it comes to the AI race, he argues DeepSeek's model "puts China neck-and-neck with the US".
It's an unwelcome shock for America's AI leaders, and a state of affairs that analysts believe will provoke a response.
Even prior to DeepSeek's breakthrough, US President Donald Trump was cutting the brakes on development, repealing a Biden executive order on AI safety in his first week. Photo: AFP
The AI bubble is a long way from over
The damage to tech stocks was significant, but don't expect the AI bubble to burst any time soon, says ASPI's Ms Wong-Leung.
"Whether justified or not, the underlying belief in the power of AI to transform our economies and societies is still strong," she says.
If anything, experts believe the US government is likely to redouble its efforts.
"In geopolitical terms, the US is likely to want to keep ahead of China," says Jeannie Paterson from the University of Melbourne.
"I don't think it will quell AI investment," she says, although she warns a race will come with risks of its own.
"What we should be talking more about is whether this is the end of responsible and safe AI."
Even prior to DeepSeek's breakthrough, US President Donald Trump was cutting the brakes on development, repealing a Biden executive order on AI safety in his first week.
"My concern is that a possible AI race means the guardrails on the US-developed AI models [will] drop off," Professor Paterson says.
In short, brace yourself for AI development to move even faster - and sometimes at the hands of unfamiliar players.
"Just as with the early days of the web, we should expect early leaders to be overtaken by new entrants," Professor Buchanan says.
"This is a warning shot to all the current AI innovators, who have been able to progress with few new challengers.
"It isn't just the US tech giants that are going to be dominating the AI race".
- ABC
