Fearing AI, climate or nuclear Armageddon? Why we might just live – John McLellan

Even by the standards of modern climate change catastrophism, the assessment by leading Anthropic safety researcher Evan Hubinger, that artificial intelligence is developing at such a rate that there is more than a ten per cent chance it will kill us all within a decade, certainly takes the cyanide pill.

“We really do earnestly believe AI could kill all humans!” said Mr Hubinger on Twitter/X this week, responding to another ex-Anthropic researcher, Jacob Coxon, who posted that “the people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible.”

READ MORE: Why the most extreme data centre opponents are actually making developers' jobs easier

Genie is out of the lamp

As if that wasn’t bad enough, Mr Hubinger added that although Anthropic was trying its best, “we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to”. In which case, Evan old bean, you and your sibling proprietors Dario and Daniela Amodei better, as the techies might say, extract the bloody digit pronto.

On the bright side, it could mean I don’t need to worry about equity release or whether I’m drinking too much, and I suppose there won’t be much point in reviving social distancing, but with more than half the world now controlled by authoritarian regimes, most with scant regard for human rights, this time it might just be serious and, no matter what Anthropic does, the genie is well and truly out the lamp.

Given how difficult it was to get to the definitive truth about whether the Covid pandemic originated in a Wuhan laboratory or a wet market, it will be impossible to track what Chinese developers are doing with AI. But we know it will be fast and unlikely to be scrutinised by any regulatory system devised by the United Nations or whoever, in the same way that, for all the climate change action taken by liberal democracies to limit carbon emissions, China is still building coal-fired power stations.

READ MORE: SNP once again show how they're constantly meddling while achieving little

High price of cutting carbon emissions

But it is not impossible for countries with very different world views to reach agreement on stopping a mutual threat, as happened in 1987 when the Montreal Protocol was signed off by the member states of the United Nations Environment Programme to prevent the destruction of the ozone layer. After the hottest August on record, today’s petrified 20-somethings might be surprised to learn that, half a century ago, we were supposed to worry we were all going to freeze to death.

The difference then was that the cause of the danger was clear and alternatives identified and accessible with little impact on national economies. Radical action to slash carbon emissions has proved much more difficult because, as Oxford Professor Dieter Helm eloquently told a Scottish Parliament committee last week, the choices facing governments in a world still dominated by carbon consumption are too painful.

“Renewables are contributing to the extra demand for energy as part of the doubling of the world’s energy demand. That is not the same as everyone moving to net zero,” he said. “We need to reduce carbon consumption, not territorial carbon emissions… but you cannot escape the fact that the price is really high, which has damaged the Scottish economy and damaged affordability for consumers who are poorer and less able to pay.”

Part of the answer, said Prof Helm, was to embrace nuclear energy, something the Scottish Government repeatedly rejects, so without agreement, and China’s share of global productivity rising partly on the back of coal, ‘net zero’ is the chance of a meaningful reduction of global carbon emissions.

READ MORE: Sturgeon is right on one thing, there is no vision in SNP

Growing catastrophism

But then why worry if AI is going to destroy us all anyway? That would be one way to cut emissions. As they say, every cloud, but even by Mr Hubinger’s gloomy prediction, there is still about a 90 per cent chance we will not be eradicated.

And with computing power so powerful it can wipe out humanity, it must surely be the case it can also work out how that can be avoided, and even a rogue state will understand that self-destruction would be somewhat self-defeating. Notably, so much of the Kremlinology about Vladimir Putin’s intentions and motivations, a constant theme has been how he prioritises his own health and safety.

The probability calculations about extra-terrestrial life on other planets factor in the chances of an advanced civilisation with the ability for inter-galactic travel also destroying itself, previously based on the likelihood of nuclear war. We have had the capacity to obliterate ourselves ever since Little Boy exploded over Hiroshima in 1945, but we are still here.

As I wrote here recently, catastrophism is now feeding a protest movement calling for a complete halt to hyperscale data centres and Mr Hubinger has just given them all the ammunition they could have desired. But protest is a luxury denied to people in China, so the AI revolution and carbon consumption will carry on apace, whether we are on board or not.

Are we all going to die? Well yes, but not all at once and I’d put my money on the 90 per cent, not the ten. As Peter Cook said to his followers when the Earth didn’t end, “it’s not quite conflagration we were expecting… same time tomorrow?”