New technologies create excitement. The invention of the railroad led to the “Railway Mania” in 1840s UK, which saw investors pile into railway stocks. In the 1920s the radio similarly captured the imagination. And more recently, euphoria over internet adoption saw the Nasdaq rise fivefold between 1995 and 2000. The hysteria stems in part from high expectations over how far and how fast innovations can boost human wellbeing and productivity, in addition to the “fear of missing out”. But in each case, the initial bubble burst as reality caught up with expectations.
The rise of generative AI — particularly large language models, like OpenAI’s ChatGPT — has sparked a similar frenzy. Since ChatGPT’s launch last November, glowing reports on the potential economic impact of the technology — which can automate tasks from writing essays to generating code — have been coming thick and fast. Goldman Sachs estimated that it could drive gains in productivity that could raise global GDP by 7 per cent over a ten-year period.
Estimating the benefits is largely a guessing game. Most experts agree it will take time to pay off. Indeed, the impact of prior technologies has often been conceptualised using the “J-curve effect”, where productivity may initially fall as it is adopted, before rising steeply. Will generative AI follow a similar path?
Electricity, railways, and computers all took decades before they generated productivity booms. In comparison, AI technology is far less capital-intensive and does not require the widescale development of new infrastructure. ChatGPT can be accessed at the click of a mouse; over 100mn people have already done so. Generative AI systems will, however, need vast computing power, which is not cheap. Companies will need to retrain staff and adjust their business models. This will take time, though user-friendliness will make adoption easier than prior technologies.
Other factors could offset that edge — such as regulation. Given its power, AI gurus have already called for a moratorium on the further development of leading edge models. Unlike historic inventions that have replaced human brawn, generative AI can do cognitive tasks such as writing, analysis, and design. It can augment these activities alongside humans, but policy and law also need to evolve to govern it and navigate the social and workforce consequences. The uncertainty, disruption and compliance involved will be speed bumps along any J-curve.
Generative AI could also directly sap productivity. For efficiency gains to benefit an economy, freed-up time needs to be used productively. But the technology could create new problems. It can be used to impersonate, manipulate data and help students pass off homework as their own. Tackling that could be cumbersome. Ironically, it also allows productivity killers such as junk mail and online distractions to be, well, more productive. By improving fraud detection, it could in time help clear its own mess.
Just how high generative AI can rise up the curve, once — and indeed if — it clears the dip, depends on its usefulness. It may boost productivity in knowledge-based jobs, speeding-up doctors’ diagnoses and legal contract-writing, but some service sectors may be less affected. By accelerating the research process itself, it could drive tech progress and iterative productivity gains. Complementary changes are important too. Railways eventually raised efficiency, but that is because industry and trade were also booming. If governments adopt AI, to slash form-filling for example, that would reduce other brakes on productivity.
There is no doubting the potential of generative AI. Its ability to boost cognitive activities, which are harder to value, means we may be unable to accurately measure its impact. But as prior technologies demonstrate, productivity gains are not guaranteed until the technology can be harnessed effectively. We should keep our feet on the ground.
Get Best News and Web Services here
