hanks to Oxford University’s Future of Humanity Institute and the Global Challenges Foundation, we now know the 12 ways in which humanity is most likely to go up in smoke. Their proposals for mitigating the threats (the usual suspects: asteroids, pandemics, and so on) make for interesting reading. The team makes ten recommendations, such as improving early-warning systems, increasing focus on the more extreme scenarios and looking at the possibility of establishing a global risk organisation. But it may have missed a trick.
One threat on the list is artificial intelligence (AI): robot minds capable of rising up and killing us all. Yet the discussion makes clear that politicians are also a big threat. Perhaps we could engineer AI to rise up and implement sensible decisions that will save, rather than threaten, humankind.
The AI threat is a popular notion. Stephen Hawking has pronounced on a few occasions that we should fear AI and be extremely careful about the kind of intelligence we create. The robots, he says, may well turn against us. It is a measure of Hawking’s cultural cachet that he can get away with this kind of speculation. There is very little evidence – perhaps none, outside science fiction – to support his claim.
AI isn’t very good at anything yet, let alone taking over the world. A brief conversation with Siri, the iPhone’s “knowledge navigator”, will allay all fears of a robot uprising.
And although the capabilities of Google’s self-driving cars are impressive, let’s remember how good we are at driving. Every day, human beings make millions of journeys that rely on complex decision-making algorithms operating at lightning speed in the brain.
Not only are our brains agile enough to do this, but they have ensured that the computers and AI that we invent are placed in control of cars only within an extremely tight regulatory framework. So we’re not stupid, after all.
Yet our minds are also naive and easily panicked. That is why we can be persuaded, with very little evidence, that a silicon-based creation of our making could become an existential threat.
This hair-trigger facility for suspicion and peril-spotting is part of our evolutionary heritage. It is the mental equipment that enabled us to survive in environments full of predators. But it is also what makes lasting international agreements so hard to reach, creating threats that are far more dangerous than AI.
We can focus on Ukraine, to take a topical example, but zoom out and we’ll see that the overall threat from politicians is huge. That’s why the Bulletin of the Atomic Scientists has adjusted its “Doomsday Clock”, a measure of the imminence of our demise as a species. In January, the time on it was moved forward: we are now at “three minutes to midnight”. The Bulletin warns that, with the governments of the US and Russia racing to modernise their nuclear arsenals, “International leaders are failing to perform their most important duty – ensuring and preserving the health and vitality of human civilisation.”
And it’s not just nuclear Armageddon. There are slow deaths on the horizon, too. Rising levels of carbon dioxide in the atmosphere were identified as a possible threat to humanity decades ago, yet little of consequence has been done about this.
It doesn’t seem to be within our capabilities to find a lasting solution to these kinds of problems. Perhaps we should encourage AI researchers to forget self-driving cars and focus on self-driving nations: a political intelligence that can steer us through dangerous times. When you consider the situation we have now, would robot overlords really be so scary?