I’m writing to you from my couch, it’s a bleak wintery new years day in London, I’m 43 years old, wearing a big fluffy electric blue jumper, and I’m head over heels with my new robot vacuum cleaner. It's an unexpected love since I would say I’m rather indifferent to vacuuming. But the cleaning history doesn’t lie, I’ve only had it a week and I’ve sent it out to clean about 8 times, as well as 2 times to mop and 3 times to redo under the bed! To temper this new found love, I’ve also spent the holiday break listening to the Reith lectures, which this year features the topic ‘Living with Artificial Intelligence’, delivered by AI professor Stuart Russell.
I must admit I’d already started to dream about the other nondescript chores in my life I could farm out to robot helpers, or better yet, put that all together into one robot that could learn any new tasks I’d like to give it, i.e moving from specialised to generalised intelligence, then all the better right? Of course dabbling as I have in the area of Responsible Tech I knew it was highly unlikely Stuart Russell was about to let us know of all the joys that no longer having to fold up my own clothes would bring, but instead to examine the dangers that might come alongside artificial generalised intelligence.
There’s quite a lot to unpack from these lectures, but in short, the overarching question of the series is, in creating general purpose AI we create entities that are far more powerful than humans, how do we ensure that they never have power over us? And should we, as Alan Turing warned, have to expect the machines to take control? This is of course a big, rather headache inducing question, to which it seems not even Stuart Russell has an answer. Although he does have some good ideas, (see lecture 4). But don’t despair, he also reminds us that we are not there yet and I’m reminded by a quote from Kate Crawford’s Atlas of AI, that currently AI is neither artificial (referring to all the human labour that goes into it) nor intelligent.
Still it seems there is already plenty to be wary of in the area of specialised artificial intelligence. I would liken the message of the lecture on warfare to suggesting that we were all wearing big fluffy electric blue jumpers and standing only millimetres from an open flame. If you want to sustain your nightmares from this lecture you could follow up by re/watching the Black Mirror episode ‘Hated in the Nation’, in which a swarm of robotic bees use facial recognition to hunt down and kill the most hated person voted for in a daily twitter poll. A technology he claims is entirely plausible and seems to suggest perhaps you could already buy something similar on the market.
Then there is of course the question, who is in control, us or the machine? Russell argues we have already seen this control over us in action, in the algorithms that govern social media content selection. Rather than serve us up more stuff we like, he argues, they have increased their goal of user clickthrough by instead making us more predictable. Users with more extreme preferences are easier to predict. Knowing this you’d think we’d be able to turn them off, but neither the majority of the users, as they are too addicted, nor the companies that govern the algorithms, as well they simply make too much money, has yet been able to act.
Of course this wouldn’t be a lecture series on the perils of technology if it didn’t cover what will we do when the robots not only do all of the vacuuming and clothes folding, but all of the work? Should we learn to enjoy the simple art of being or do we still require some kind of purpose to make us get out of bed? I lean on the side of requiring purpose to make life worth living, but perhaps that's because, as he puts it I’m much more British than I am French or Italian (I’m in fact from New Zealand). I offer as evidence the fact that I’m writing this on my holiday and already setting goals for my 3 month sabbatical which I am starting in February. Although perhaps I just need to unlearn my goal oriented tendencies and as Jon Kabat-Zinn teaches embrace non striving? I’ll get back to you on that one after the 3 months.
There is a lot in the lectures that I haven’t covered and which will require many evenings of pondering. But if I could leave you with some final parting advice from Russell on how we might give ourselves a fighting chance of staying in control of the machines, it's to get out our ethics book. To quote him “What works is for a real conversation to happen, for the AI researchers to understand that they don’t know much about the last two and a-half thousand years of ethics research and be willing to learn about it and read it, and just see the pitfalls, because philosophers, in a way, have been debugging the moral programmes of other philosophers". So it seems that my love for my robot vacuum cleaner is ok, for now, but it might be time to turn off Netflix and learn the difference between rights-based, virtue-based and utilitarian approaches to ethics.
No comments:
Post a Comment