Amongst my colleagues and friends there is a lingering fear about AI going rogue and taking over, a fear possibly fuelled by popular and even classical culture (both Terminator and Frankenstein envision a human creation going rogue). Though not totally unthinkable this is a scenario from which we are far removed. For me one book on this subject stands out, Nick Bostrom’s Super Intelligence.
The opening of Super Intelligence contains ‘The unfinished fable of the sparrows’, symbolising our wish to ‘train’ an entity which might outgrow us. In short, the fable tells the story of a flock of sparrows which decide to breed and train an owl as one of their own to gain its protection. The ending is formed in such a way that it perfectly connects to the core of the book itself. However, it is aimed at general intelligence, a topic which is currently out of our reach.
Our current state of play related to AI is wholly aimed at narrow AI, a broad selection of techniques which can do a single task extremely good. Looking at the owl we could take parts of the whole entity and place them in this context, making the fable relevant to our current situation. The claws, eyes and wings all provide us with opportunities and dangers. Its claws and eyes can help us solve challenging problems, but there is a danger we could misuse them to gain an advantage over outsiders. Wings might help us soar, but as Icarus tells us we should not haste and heed the warnings of others.
Modern day AI algorithms provide us with an enormous amount of chances, but we should also heed the dangers. We not only face the direct impact of AI implementation and possible errors it might make; the societal impact is also likely to be huge. Think about self-driving cars and the role of truck and taxi drivers or automatic x-ray image diagnosis. Both promising techniques which are positive in nature but the impact on employment is likely to be huge. On the negative side of the spectrum we have the dangers of mass surveillance or racist algorithms, problems we and AI are already struggling with today.
In short, I still believe AI will give us a lot of opportunities but we, in the broadest sense of the word, should also heed the dangers.