Richard Madeley has dismissed Donald Trump's dismissal of AI safety warnings as a "hoax", drawing parallels with the history of aviation safety to argue that regulation is possible but urgent.
Aviation safety lessons
Madeley points out that flying used to be far more dangerous, with American airlines suffering at least half a dozen fatal crashes every year during the 1950s and 1960s. Even from 1968 to 1977, there was one death per 350,000 passengers, compared with one in 27 million today.
He argues these improvements show that complex high-risk systems can be made safe, but it takes time and effort.
Trump's credibility questioned
Madeley criticises Trump's judgement, saying his "worth in good judgement let alone the truth is pretty much a bankrupt currency now". He notes that wars Trump said were won are still raging, and peace talks he claimed were on the brink have not begun.
He recalls that Piers Morgan once said of Meghan Markle that he wouldn't trust Trump to read him the weather forecast, a sentiment Madeley applies directly to the former president.
AI's exponential growth
Madeley highlights the warning from Dario Amodei, the boss of AI giant Anthropic, that AI could take over the entire internet within a year through "recursive self-improvement". Amodei has called for a "swarm" of AI units to be stopped, and Madeley believes the airline industry's safety record proves it can be done.
Amodei notes that commercial airplanes are a precedent for operating complex high-risk systems safely, but "it takes time to get it right". Madeley concludes that the window to act is barely a few months, and he does not need AI to answer whether that is a hoax.
