Nearly 50 years ago, when computers were just being born, the world’s greatest science fiction writer, the late Arthur C Clarke, created a fictitious computer called HAL (Heuristically Programmed Algorithm). HAL was sentient and could nearly behave like a human. It was the main brain of a spaceship and developed a fault. No one was able to control it after that. The book was called “2001: A Space Odyssey”. It popularised the idea of artificial intelligence. Clarke even foresaw self-replicating computers that destroy Jupiter.

Now, after half a century, we are at the point where what Stephen Hawking feared could come true, namely, self-replicating, sentient computers. This is what makes AI different from all new technologies of the past. But being either just sentient or just self-replicating is not sufficient. Both conditions must be met before AI becomes a danger. The worry is that when that happens, it will spell doom for the human race. So how should humans deal with the problem? While we know what AI is, we don’t know its future trajectory. We also know that all the beneficial things which humans invent can also harm them. And as always happens when a lack of knowledge breeds insecurities, science is getting mixed up with semi-superstition. It’s this that the world must tackle without taking what is now called a pause in the development of AI. There are a few things that simply must be done.

First, a globally representative agency like the UN must step forward to initiate the next step. This is to convene a Commission comprising the best (and only) scientists of the world and absolutely no one else. The Commission should work out its own terms of reference to evaluate the risks, namely, sentient and self-replication. Second, since there is considerable private intellectual property involved, a way must be found — like the US FDA — to validate the separate outcomes of AI research. The body must have global authority meaning it must be created by treaty like the other agencies like the WTO, WHO etc. Third, in parallel, individual governments must do the same within their jurisdictions. Their main focus, however, should be regulation. But this requires an understanding of what is to be regulated. So some coaching and training of would-be regulators is a necessary prerequisite. At all costs, career bureaucrats should not be involved in this exercise. Again this should be left to scientists and engineers. Fourth, since the US and China will seek to dominate the outcomes, it must be recognised that while the US bats for private interests, China bats for the interests of the CCP. These are diametrically opposed. Controlling their urges will form a huge challenge.

Last but not least there is India which simply doesn’t apply itself when it should. There are strong reasons to believe that the government hasn’t got around to taking a good hard look at AI. This perception must be corrected without delay and not when, as often happens today, after someone gets an order from the Supreme Court on a writ of mandamus.

comment COMMENT NOW