Economics Explored
Will AI Finally Show Up in Productivity? With Chris Berg
Episode Summary
Australia's latest Intergenerational Report assumes productivity growth will rebound significantly over coming decades, with AI potentially playing a major role. Is that reasonable? Professor Chris Berg of RMIT joins Gene Tunny and John Humphreys for a wide-ranging discussion of AI's economic potential and risks. They explore lessons from the computer and internet revolutions, AI's huge energy requirements, whether its benefits will appear in GDP, the cybersecurity threat from autonomous agents, and why competition may be a better safeguard than heavy-handed regulation.
Episode Notes
Australia's latest Intergenerational Report assumes productivity growth will rebound significantly over coming decades, with AI potentially playing a major role. Is that reasonable? Professor Chris Berg of RMIT joins Gene Tunny and John Humphreys for a wide-ranging discussion of AI's economic potential and risks. They explore lessons from the computer and internet revolutions, AI's huge energy requirements, whether its benefits will appear in GDP, the cybersecurity threat from autonomous agents, and why competition may be a better safeguard than heavy-handed regulation.
Gene would love to hear your thoughts on this episode. You can email him via contact@economicsexplored.com.
Takeaways from this episode
- AI could be a genuinely productivity-enhancing technology, but forecasting the size of the gains is extraordinarily difficult. Berg argues that AI is already allowing people to accomplish tasks with dramatically less human effort, although how much of that ultimately appears in measured productivity remains uncertain.
- Transformative technologies don't necessarily produce immediate jumps in measured productivity. Computers and the internet transformed economic and social life, while some of their gains arrived with lags or appeared as consumer surplus rather than conventional GDP.
- The AI revolution is also an energy and capital-investment story. Data centres are effectively factories for machine intelligence, creating potentially enormous new electricity and infrastructure requirements.
- Cybersecurity may be a more immediate AI risk than science-fiction extinction scenarios. Berg argues that AI could allow enormous numbers of agents to probe vulnerable systems simultaneously, while also providing defenders with better tools.
- Regulation could have unintended consequences. The discussion questions whether national AI regulation can work when models are developed globally, while John argues that regulation that entrenches a handful of dominant firms could itself increase systemic risks.
Timestamps
- 0:00 – AI and Australia’s productivity challenge
- 4:06 – AI, demographics and economic growth
- 5:50 – AI’s enormous energy and capital requirements
- 11:29 – Will AI show up in the productivity statistics?
- 18:24 – How AI is already transforming work
- 21:13 – Could AI change the productivity trajectory?
- 22:42 – Could AI kill us all?
- 28:59 – AI agents and the cybersecurity threat
- 35:11 – Can governments effectively regulate AI?
- 40:00 – The AI risks worth worrying about
- 43:30 – Could AI become a monopoly?
Links relevant to the conversation
Full ATA livestream - Albanese bets everything on AI || #52:
https://www.youtube.com/live/Knec2ULWCfA?si=gw7fpzRJo7kk1Kvl
Chris Berg’s Substack:
https://chrisbergaus.substack.com/
Lumo Coffee promotion
10% off Lumo Coffee’s Seriously Healthy Organic Coffee.
Website: https://www.lumocoffee.com/10EXPLORED
Promo code: 10EXPLORED