DONE wisely, artificial intelligence 鈥渃an be the best thing ever for humanity鈥�, says the fundamental physicist turned AI researcher Max Tegmark in our interview this week (see 鈥淚f we do it wisely, AI can be the best thing ever for humanity鈥�). We subscribe wholeheartedly to his assessment. Seldom has there been a technology with such an obvious power to improve our lot 鈥� or one with such obvious dangers.
The risks are potentially existential. Our recently published , contains a classic article from strategy researcher Kenneth Payne on AI warbots which provides a case in point. Yet they are also insidious: in amplifying biases, for instance against women and minority groups, that exist in the data AI feeds on; in undermining democracy and the rule of law via the automated propagation of information designed to play to prejudices; in entrenching the economic power of companies that control this technology, at the expense of those whose jobs it takes.
Advertisement
The answers to these issues are simply stated: more research, more regulation and more agreement on where AI鈥檚 boundaries should lie. The work on ways to balance AI accuracy and fairness we report on in this issue (see 鈥�Teaching an AI to be less biased doesn鈥檛 have to make it less accurate鈥�) is a good example of how basic research can help. There are encouraging signs that big tech is finally, albeit reluctantly, waking up to the idea that with the great power AI grants comes great responsibility.
鈥淭he coronavirus crisis found many governments asleep at the wheel. We can鈥檛 afford the same mistakes with AI鈥�
Yet the scale of the challenges involved means that, ultimately, governments and international organisations must step up to the plate. The Council of Europe, which exists to uphold democracy, human rights and the rule of law across the EU, the UK, Russia and other European countries, has already .
More action and urgency are needed, however unfashionable global cooperation might be right now. The coronavirus crisis found many governments asleep at the wheel, faced with a pandemic that researchers had warned for years was a matter of if not when. We can鈥檛 afford the same mistakes with AI. It could well be the best thing for humanity 鈥� but we must think carefully about what we want it to be now.
