Private sector investment in artificial intelligence (AI) has grown exponentially in recent years amid the emergence of breakthrough technologies like ChatGPT.
According to Goldman Sachs Economics Research, the global value of private investment in AI grew approximately 130 per cent in just three years, from US$48 billion (AU$76 billion) in 2020 to US$110 billion (AU$174 billion) in 2023.
This trend is expected to continue, with AI investment tipped to hit US$200 billion (AU$316.8 billion) by 2025.
The growth in demand for AI investments has been underpinned by expectations of a marked tech-driven improvement in economic productivity.
However, a new analysis from Oxford Economics’ lead economist, Adam Slater, suggests the touted boost to productivity may be overstated.
“Our historical evidence strikes a note of caution about extrapolating dramatic rises in economy-wide productivity growth from the adoption of AI,” Mr Slater said.
The productivity benefits of general-purpose technology (GPT) like AI, the group added, may “take decades to bear fruit”.
Citing principles outlined by economics professor Nicholas Crafts in a paper published in 2002, Mr Slater said initial impacts of GPT technologies on growth “can be small or even negative”.
“It takes time to transform a capital stock given that annual investment is only a small fraction of the total,” Mr Slater observed.
“New technologies may require heavy investment in new equipment to yield benefits, plus significant complementary investments (including in human capital, e.g. training) too. Some firms may be reluctant to engage in costly new investment if it’s not rapidly profitable.
“…AI is unlikely to be very different, with substantial investments necessary in specialised hardware, software, and training to reap the benefits. These are costs that some firms may not wish to bear initially, and which may be prohibitive for others (including in poorer economies).”
Moreover, Mr Slater warned the benefits of AI adoption may not endure upon integration.
“It’s one thing to yield increases in productivity levels for a given industry or even a whole economy, but another to permanently raise the rate of productivity growth,” he observed.
“In the case of AI, a permanent boost probably rests on whether the use of AI leads to sustained improvements in the rate of innovation.”
Mr Slater went on to claim AI technology itself may ultimately fail to meet broader societal expectations.
“Supersonic jets failed as a commercial proposition and nuclear power was never the ultra-cheap transformative technology that was hoped for (fusion remaining distant),” he said.
“Some of the remarkable recent improvements in areas like mobile phones have also done little yet to boost broader productivity.
“Maynard Keynes in the 1930s speculated that the rate of technical progress would leave us all leading lives of ease by now, with all our material wants satiated. This hasn’t proved to be the case.
The Oxford Economics analysis comes amid mounting concern over waning economic productivity since the outbreak of the COVID-19 pandemic.
In Australia, former Reserve Bank governor Philip Lowe flagged these risks during his final address at the helm of the central bank.
“Unfortunately, the recent productivity record isn’t encouraging,” he said.
“…The problem is not a lack of ideas. Instead, it is in building the consensus within society to implement some of these ideas.”
Former governor Lowe said the productivity issue is ultimately a “political problem”.
“If we can’t build a consensus for changes, the economy will drift and there is a material risk that our living standards will stagnate,” he warned.