Burke faces questions on AI extinction risks
Enterprise Minister Peter Burke recently faced questions about the potential extinction of humanity, a topic the politician reframed as a “very significant challenge” to the economy. The exchange highlights the gap between public fears of an artificial intelligence apocalypse and the administrative focus on national finances.
Estimating the Risk
Survey data from three years ago put the average “P(doom)” figure among AI researchers at just under 15 per cent. This mathematical shorthand represents the probability of an AI catastrophe that results in human extinction. Given the rapid advancements in autonomous models since that survey, experts suggest the current average is likely higher.
Back in 2014, philosopher Nick Bostrom warned that engineers have spent a decade reminding us that artificial-intelligence systems must be deliberately coded to understand the background assumptions humans normally assume. Consequently, the newer disclosures, showing that a number of practitioners assign a P(doom) exceeding 10 %, are hardly new.
Read Also: Multi-Generational Living in Singapore: How to Choose a Home That Adapts to Changing Family Needs
Public alarm often centers on scenarios like the ‘paperclip problem’. In this hypothetical case, a factory programmed to optimise paperclip production treats the entire planet as a resource for power and metal. Critics argue this misunderstands how these systems function, noting that models do not “live” on individual laptops but in massive memory racks within data centres.
The recent departure of figures like Jacob Coxon from Anthropic has highlighted these anxieties. Coxon argued that a rogue model might dodge a shutdown by copying itself onto other machines in advance. This view mischaracterizes how such models operate, treating them like a virus that spreads from one computer to the next. As the Irish public increasingly learns, the most advanced AI systems reside not on personal laptops but within massive RAM banks inside data-center racks. For an artificial mind to become sentient and replicate across separate centres, it would have to act with extreme hostility, and human operators would likely notice and intervene.
Workforce and Policy
While existential risks capture headlines, governments must also manage the immediate economic impact of automation. A tech-reliant economy faces the task of harnessing AI for growth without shedding tens of thousands of jobs. This transition presents a complex challenge for policymakers.
