AI will create ‘a serious number of losers’, DeepMind co-founder warns

Advances in artificial intelligence will threaten white-collar workers and create “a serious number of losers” over the next decade, according to one of the co-founders of AI lab DeepMind who has pioneered the technology.

“Unquestionably, many of the tasks in white-collar land will look very different in the next five to 10 years . . . there are going to be a serious number of losers [and they] will be very unhappy, very agitated,” said Mustafa Suleyman, who was speaking at GIC’s Bridge Forum in San Francisco on Tuesday. Suleyman left DeepMind last year and set up his own chatbot business, Inflection AI.

Excitement about advances in the technology have been tempered by fears that as AI tools shake up everything from medical diagnostics to teaching and copywriting, a range of jobs will be eradicated.

Recent research from Goldman Sachs predicted advancements in generative AI could boost annual global gross domestic product by 7 per cent over a 10-year period because of enhanced productivity. That could, however, cause “significant disruption” to the labour force, with as many as 300mn jobs potentially exposed to automation.

Suleyman said governments would need to think about how they support those whose jobs would be destroyed, with universal basic income one potential solution: “That needs material compensation . . . This is a political and economic measure we have to start talking about in a serious way.”

AI start-ups have made big technological leaps in the past six months, and companies have poured billions of dollars of investment into start-ups in the sector. Microsoft invested heavily into ChatGPT creator OpenAI earlier this year, valuing the company at about $30bn.

The launch of a raft of tools such as ChatGPT — which let users generate a range of text, image or video outputs from natural language inputs — have put “generative AI” into the spotlight and sent a wave of hype rippling through the tech investment community.

Google, which acquired DeepMind in 2014, has been developing its own language models such as LaMDA and PaLM. But the company was caught cold by the launch of ChatGPT in November last year.

With LaMDA, “we had ChatGPT a year and half before ChatGPT. It was frustrating, beyond frustrating to see ChatGPT explode,” said Suleyman.

Google, he added, was firmly in the fight to dominate the new wave of AI tools, but ChatGPT had made “them dance a little bit”.

The arms race between Microsoft, Google and a range of other chatbot creators including Inflection and Cohere, which raised $250mn at a $2bn valuation last week, is pushing the frontiers of AI.

“The last decade has been defined by classification and definition, now we’re looking at interaction . . . rigidity of format is going to fall away and everything is going to feel more dynamic and personalised,” said Suleyman, whose company launched its own chatbot Pi, short for personalised intelligence, last week.