This article is more than 1 year old

Investment bank forecasts LLMs could put 300 million jobs at risk

Meanwhile technology-driven job creation is possible, Goldman Sachs says

Research from global investment bank Goldman Sachs claims that as many as 300 million jobs could be at risk from automation powered by generative AI.

The tech industry has seen burgeoning releases based on large language models, from the likes of Google and OpenAI, in partnership with Microsoft. Impressively aping human language and even producing working computer code, LLMs are at the centre of a media frenzy right now.

Impressive though they often are, any notion that they think or understand what they are saying is false, given their reliance on a statistical analysis of a large corpus of text.

Nonetheless, Goldman Sachs sees a significant disruption to the global jobs market, with around 300 million full-time equivalent workers in large global economies at risk, according to economists Joseph Briggs and Devesh Kodnani, who wrote the paper.

On the upside, investment in LLMs in the workplace is set to create a much-needed fillip for productivity. Over 10 years, generative AI could produce a $7 trillion boost to global GDP.

To get the benefits though, organizations need to invest. The Goldman report said that, assuming it followed similar trends to software investment in the 1990s, US spending on workplace LLMs and the surrounding tools could hit 1 percent of US GDP by 2030, amounting to about $230 billion, based on current figures.

The research assumes that about two-thirds of jobs in the US and Europe are set for some degree of AI automation. Those with less than half of their workload automated (around 63 percent in the US) would gain time to do something more useful than humdrum repetitive tasks.

Lawyers and those in administrative jobs are thought to be among the roles most at risk.

Meanwhile, many outdoor and physical jobs would be unaffected, at least by LLMs. That's another 30 percent in the US. The remaining 7 percent, though, would be at risk of being replaced.

The study does note, though, that displaced workers may yet find new jobs themselves created by technical change. Citing a paper by economist David Autor, they reason that six in 10 workers are today in occupations that did not exist in 1940.

Among candidates for new jobs could be police or security workers employed to catch criminals who use LLMs for various nefarious activities. Pan-European crime agency Europol said AI language models could fuel fraud, cybercrime, and terrorism. ®

More about


Send us news

Other stories you might like