BARRY COLLINS
Artificial intelligence could come up with a creative solution to global warming – but we might not like it
Artificial intelligence could come up with a creative solution to global warming – but we might not like it.
I’m still being outsmarted by the AI, even on the lower difficulty levels – if Deity is the game’s hardest level, I’m on Shane Richie
We can’t afford to wait for the computers to overtake our broad intelligence before we decide to do something about it – that’s too late
Iam in deep, deep trouble. I have roughly 15 minutes of disposable downtime a month and I’ve been sucked in by Civilization VI. I’m now weighing up what to sacrifice: sleeping or eating.
The Civilization (or Civ for short) games, for those unfamiliar with this long-running series, are for desktop despots everywhere. You get a fledgling city on a map and you have to expand your empire, turn by turn, as the game crawls through the ages. There is thunderously more to it than that, but if I tried to explain the whole game this issue would be the biggest in our 25-year history.
After hell-knows how many hours of tutorials, YouTube explainers and reading up online, I’m still being comprehensively outsmarted by the AI, even on the lower difficulty levels – if Deity is the game’s hardest level, I’m on Shane Richie.
In one game, for example, I saw Hong Kong sitting there as an isolated city state, with tons of resources close by and a prime location. No empires seemed interested in it so I made my move: I lined up the chariots, sharpened my warriors’ spears and prepared for the easiest romp to power since the Conservative leader election. Then Gandhi piped up.
History scholars may be dimly aware that old Mahatma’s none too keen on violence and he tells me to back off. I tell him to do one, making a mental note to send him a vat of wine from my reserves to patch things up later. How was I to know Gandhi doesn’t drink? Two turns later, Gandhi’s lined up his entire
army across the border to Hong Kong and now I have two choices: start a massive, unwinnable war or beat an embarrassing retreat. I could practically hear my warriors tutting in disgust as they trudged back home.
Why am I recounting my tale of Civ ineptitude? Because it made me feel like a piece of cheese to be so easily outsmarted by the AI. But then the book I’m reading, Max Tegmark’s
Life 3.0, suggests I’d better get used to it, as we’re not far from a time when AI is going to outsmart us all.
Tegmark, an MIT physicist, is one of our 25 Most Influential People In Tech ( see p30) – and was nominated by PC Pro’s own brain-the-size-of-aplanet, Dick Pountain. His book pulls together many of the world’s leading AI experts to answer what I think could become the defining question of the 21st century: how do we prepare for a world where the computers are smarter than we are?
Right now, we’re still just about in the era where AI is “merely” capable of outsmarting us at very specific tasks, most notably games of chess, Jeopardy or Go. But deep learning is edging AI closer to artificial general intelligence (AGI), where computers can learn any task that we’re capable of. Computers are already teaching themselves languages or how to play computer games without any programming, not even the norms of grammar nor the rules of the game. AGI could be as little as decades away.
That’s why, Tegmark argues, we need to be thinking ahead. We can’t afford to wait for the computers to overtake our broad intelligence before we decide to do something about it – that’s too late. We need to decide now how much latitude we’re going to give AGI to drive our cars, run our smart homes or even start our wars. Because we may not like the answers it comes up with if left to its own devices.
The first chapter of Life 3.0 is a fictional account of what might happen if the AGI is left largely to its own devices. I won’t spoil it for you, but it’s reasonably benign for us. But what if AGI were left to reach the (not unreasonable) conclusion that the only surefire way to save the planet from global warming was to make sure there were a lot fewer of us?
To be clear, I’m not talking about self-driving tanks powering through Chelmsford of a Saturday afternoon, wiping us out Terminator-style. But computers control our lives in all manner of ways. Could an AI crash the banking system, disrupt the food supply or manipulate social media in such a way that might spark civil war? It’s certainly conceivable.
The underlying message of Life 3.0 is don’t underestimate AI. Tegmark recounts the tale of how the Alpha Go computer came up with a move that broke all the conventions of Go to beat the world’s best human player. Computers are capable of creative thinking, too. For proof, check out the subreddit that’s written by AI: reddit.com/r/SubSimulatorGPT2.
Perhaps AGI will come up with a creative solution to global warming that none of us have thought of yet. Perhaps, like Civ’s AI Gandhi, it will find a peaceful solution. The question is: are we willing to take that chance? barry@bigtechquestion.com