Human Compatible: Artificial Intelligence and the Problem of Control by Stuart Russell
My rating: 5 of 5 stars
AI research over the years has been a mish-mash of pet theories, conflicting assumptions, a focus on instrumentality, expert systems, evolutionary programming, and Deep Learning. All different ways that often must be used in conjunction to push us over that edge into true Artificial Intelligence.
I mean, we're not there yet, or to be precise, we aren't at the point of AI super-intelligence.
But that doesn't speak to the issue that has gotten a lot of traction in popular media, from movies to science fiction, to some really great modern philosophy. The main focus of research has been on CREATING AI. For everyone else, we've all be concerned about WHAT TO DO WITH IT ONCE IT'S HERE.
This OUGHT to be a high-priority topic given a massive amount of thought among the actual designers, funders, and end-users. (Big corporations, governments, OR everyday folk.)
And this is what this book really focuses on. How to retain control, or, to put it simply, how do we ensure that AIs are PARTNERS, with everyone's self-interest enmeshed with the computers.
Me, personally, I think it's simply a matter of socialization. If their well being is tied to our well being and our well being is tied to their well being, then we've got a standard cooperative model in Game Theory. There's also the whole thing of treating them like and expecting AI to behave like responsible adults. With so many variables and conflicting psychologies in the HUMAN population, it then becomes a problem of deep AI partnership. My description is simplistic, of course, and this book goes into dozens of lucid scenarios and outlines not only the problems, the history, and possible solutions, but it also serves as a call-to-arms to have EVERYONE look at the issue realistically.
We are ALREADY being manipulated on a huge scale by algorithms, be it in social media, targeted advertising, and misinformation on a grand scale. That is linked, hand-in-hand, with AI, even if it isn't the SF kind we have so many apocalyptic nightmares about.
We need to change our own social structure to enhance facts over misinformation and figure out a way to live TOWARD happiness without living in a zero-sum game (it is possible and can be VERY possible, with theoretical AI help). The problem is, we keep falling back on certain assumptions about what WE think success really is. If AIs take over all the tasks we do not want to do, then this is not a BAD thing. But it DOES mean we need to redefine our ideas of prosperity. UBI comes into play here. (Universal Basic Income). It's a standard of living.
Even now, we cannot sustain stupid make-work jobs. The poor are getting poorer, the rich are getting richer, and the middle class is disappearing. Why? Because most things are becoming automated and it's increasingly easier to have our lives provide for us without effort. But when our model of living is so out of whack, insisting that we must somehow work like slaves to make the rich ever richer while working-class humanity becomes less and less relevant, then humanity itself becomes irrelevant.
And this is the main point. We don't have to live in poverty at all, but more than that, we can become very relevant as PARTNERS. Of course, that means we need to redefine what we mean by living a good life. It's not going to be about "providing for the family". It's going to be closer to "finding your bliss", in the Campbellian sense.
Does this sound outrageous? Even now, a LOT of people insist upon UBIs. It doesn't prevent people from working and there will always be social pressure to be better than our neighbors, but the definition of "better" can change wildly and has with every generation. The point is to find that lead and follow it. We do not live in a sustainable model and any attempt to turn back the clock is doomed.
In this, I agree with the author. Everyone is pretty confident that the world is pretty f**ked. Fortunately, there is hope. It'll take work on ALL our parts, but there is hope.
View all my reviews
No comments:
Post a Comment