麻豆传媒

IT WAS supposed to be a meeting of minds 鈥� scientific, cultural and artificial. A recent closed-door colloquium held at a top British university brought together two dozen leading AI researchers and a group of illustrious humanities scholars to address philosophical questions raised by AI.

In the event, however, the two contingents struggled to find common ground. The engineers described their work in precise, dense terms, mostly skipping over its implications. The scholars, too, critiqued the descriptions rather than tackling the implications; some seemed affronted by the very idea of thinking machines.

That鈥檚 a pity, because we really need fresh thinking on AI. Public discussion is dominated by scary stories about the elimination of human workers 鈥� or of humans, full stop. Meanwhile, machine learning applications are quietly becoming integral to our daily lives, with little consideration of the consequences (see 鈥�How scaremongering steps us asking the right questions about AI鈥�).

Those consequences are myriad, poorly understood, and evolving rapidly. AIs are now involved in many decisions that significantly affect our lives. But there is also growing evidence that they aren鈥檛 necessarily ready to shoulder all the responsibilities they are being given.

Machine learning alone may be fine when it comes to optimising, say, your power usage. But in less clear-cut tasks, such as healthcare, the combination of human and machine can be more effective. Then there鈥檚 the problem of 鈥渁lgorithmic bias鈥� in AI decision-making (see 鈥�Biased policing is made worse by errors in pre-crime algorithms鈥�).

Artificial intelligence really has made great strides in recent years, so much so that initial awe at these machines鈥� uncanny abilities is giving way to hucksterism. But for all their smarts, the machines can鈥檛 tell us how we could best deploy them. For that, we need people willing to engage in a genuine meeting of minds.

This article appeared in print under the headline 鈥淢ore smarts needed鈥�