robotics

A letter from the real singularity

PNRJ computer science, critique of neoclassical economics, development economics, ethics, futurism, inequality, macroeconomics, market failure, public policy, science and academia , , , , , , , , , , ,

Apr 5 JDN 246136

I’ve been unable to find it, but several years ago someone famous wrote a sci-fi work entitled something like “A letter from the post-singularity” about how great life is after AI takes over everything. Today I thought I’d write a more realistic take on the path we actually seem to be on.

The year is 2073. Technically we still don’t have true AGI; as far as we can tell, AI still isn’t actually sentient and AIs aren’t people. (Some of us wonder, though. Philosophers debate it.) But that doesn’t really matter, because all white-collar work has been completely automated, and so has any blue-collar work that doesn’t require fine dexterity or unusual expertise. Plumbers and electricians are still doing all right (though they do more of their work at data centers than homes these days); sometimes I wish I’d apprenticed to be an electrician. Then again, the world can still only support so many electricians. AI managers command AI-run asteroid mines to extract ores to transport with AI-run spacecraft to Mars where AI-run factories process those ores and fabricate chips for more AIs that more AI-run spacecraft carry to Earth to be installed in AI-run data centers. And each time one gets sold, some trillionaire’s number goes up, and that’s the only thing people like him have ever cared about in their lives.

There are of course a handful of super-brilliant, super-creative, or just super-lucky individuals who manage to get rich making art or music or books or video games or whatever, but the vast majority of people who do art are still starving artists, just as they’ve always been, I guess. And the AI-generated stuff is good enough now that most of the time people will just use that instead of paying extra for the “authentic” “artisanal” stuff. (And most people can’t even tell the difference anyway.)

Harvard and Oxford still have professors, but most universities have fully automated teaching and most of their administration—and yet somehow tuition is barely any cheaper now than it was in your time, even adjusted for inflation. And if you were thinking of becoming a professor yourself? You should probably just go play prediction markets or something; you’d have better odds. The number of research papers published every year is astronomical, but they’re all written and reviewed by AI, rarely if ever even read by any human being, and so it seems like the actual progress of scientific knowledge has pretty much ground to a halt. (Seriously, how are there still string theorists? It’s been a century.) I guess corporate R&D still keeps on improving those graphics cards somehow; maybe they’ve discovered something important, but if they have, they’re keeping it to themselves. And I keep reading about amazing advancements done by AIs (especially in pure math that I’m not sure anyone understands), but none of it actually ever seems to affect anyone’s actual lives.

As for me, I live on UBI. Like 90% of people do. It’s enough to rent a cheap apartment (but own a home? Are you serious? Only millionaires own homes.), buy basically-adequate food (as long as you don’t eat out too much anywhere that’s not fast food), and pay for all the subscriptions to media services and home assistants and whatnot. What you make on UBI will only buy you the ad-supported versions, so while my fridge will order milk for me (delivered by drone in a couple of hours) and my robot maid will cook breakfast, fold the laundry and put the dishes in the dishwasher, my fridge is also constantly running ads and my maid will intersperse targeted sales pitches into its casual conversation. Sometimes I think I should just get rid of it (her?) and do my own cooking and cleaning myself, so I would never be able to sell it for half what I paid for it. If I could make some extra money, maybe I could at least upgrade to the ad-free subscription for my maid. (The Pro subscription and hardware addons to make her your girlfriend are just gross, but I’m sure they make tons of money.)

Every year, some politician makes a big deal about how the UBI trust fund is draining and will be gone in ten years or whatever; but it’s obvious that all they’d have to do to fix that would be raise the taxes on trillionaires a little bit, yet somehow that never seems to happen. But they also don’t cut UBI payments either, except sometimes to reduce our cost-of-living adjustments. I dunno; maybe they will really cut UBI payments in a few years. Or maybe we’ll get lucky and they’ll actually raise taxes for once.

At least I wasn’t dumb enough to move to Mars, where “employment is guaranteed!” but you have to pay a subscription for your oxygen.

My worst days are probably… about as bad as your worst days. Frankly they couldn’t be much worse, because sometimes I just want to die. Like a lot of people on UBI, I feel like a burden on society, like the world would be better off without me. Medicaid won’t cover neuroregulator implants, so I still take pills for my depression; they’re probably better than the pills you could get, but they’re still far from perfect.

My best days are maybe better than yours, maybe worse; it depends, I guess. If you’ve got cancer, your days are probably worse than mine, because we can pretty easily treat most cancers now. But if you’re healthy and you’ve got a steady job and a tight-knit community where you live, I’m guessing your days are better.

I have access to faster computers and faster Internet than you can probably imagine; my understanding was that back in the day you had to wait for downloads sometimes? Or even sometimes wait for webpages to load? What was that like? And your storage was measured in gigabytes, not petabytes? Humanity has never been so connected; human beings have never been so isolated. I theory I can contact anyone in the Solar System at the speed of light, but in practice my friends and I always seem to have trouble keeping in touch. (Oddly, it’s my best friend who lives on a station over Ganymede that I seem to stay in touch with best; we have to write full-length emails, because there’s no way to have a conversation on a ping of an hour and a half. It feels like being an old-timey pen pal, I guess.)

It’s not all bad. Some things are definitely better these days.

People often live to be 110 or even 120 nowadays. (So you might still be alive when I write this.) Rarely does anyone seem to make it past that, though; aging is just… really hard to beat. (The Boomers are finally almost gone, but Gen X is still gonna be with us for awhile yet.) We’ve cured a lot of the diseases that were bad in your time, but not all of them. And sometimes only people rich enough to pay for their own healthcare can afford the cures.

Language barriers are pretty much gone. If I wanna read something that was written in Japanese or Xhosa, I just have an AI translate it, and the translations are good enough now that you’d have to be really deeply-versed in the language to find any problems with it. Like, okay, maybe I’m not getting all the subtle connotations of Japanese literature, but was I ever going to actually learn kanji to read the originals? No. That kind of thing is for people with crazy obsessions.)

Our video games are definitely way better than yours. Characters with AI personalities that adapt in real time to how you behave. Procedurally-generated open worlds that can literally expand to the size of entire planets. (Actually, I vaguely remember reading you had a couple games that did something like the second one? Minecraft and Factorio​, I think they were called? Impressive that you could pull that off on a gigaflop processor.) Worlds and factions that adapt to your actions and provide realistic consequences so that no two players’ experiences of the game are exactly the same. It’s easy to lose yourself in a game like that (especially if you’ve got a VR setup), and when you’re playing in such a rich, interesting world for hundreds of hours. you can sometimes forget how bleak things are back in the real world your flesh-and-blood body lives in. (But then you get hungry or have to pee and you get forced back into reality.)

Economists keep telling us that per-capita GDP and productivity have never been higher, and that we have access to all these wonderful goods and services that previous generations could scarcely even imagine.

But if that’s true, why do I sometimes just want to die?

Love is rational

PNRJ cognitive science , , , , , , , , , , , , , , , , , , ,

JDN 2457066 PST 15:29.

Since I am writing this the weekend of Valentine’s Day (actually by the time it is published it will be Valentine’s Day) and sitting across from my boyfriend, it seems particularly appropriate that today’s topic should be love. As I am writing it is in fact Darwin Day, so it is fitting that evolution will be a major topic as well.

Usually we cognitive economists are the ones reminding neoclassical economists that human beings are not always rational. Today however I must correct a misconception in the opposite direction: Love is rational, or at least it can be, should be, and typically is.

Lately I’ve been reading The Logic of Life which actually makes much the same point, about love and many other things. I had expected it to be a dogmatic defense of economic rationality—published in 2008 no less, which would make it the scream of a dying paradigm as it carries us all down with it—but I was in fact quite pleasantly surprised. The book takes a nuanced position on rationality very similar to my own, and actually incorporates many of the insights from neuroeconomics and cognitive economics. I think Harford would basically agree with me that human beings are 90% rational (but woe betide the other 10%).

We have this romantic (Romantic?) notion in our society that love is not rational, it is “beyond” rationality somehow. “Love is blind”, they say; and this is often used as a smug reply to the notion that rationality is the proper guide to live our lives.

The argument would seem to follow: “Love is not rational, love is good, therefore rationality is not always good.”

But then… the argument would follow? What do you mean, follow? Follow logically? Follow rationally? Something is clearly wrong if we’ve constructed a rational argument intended to show that we should not live our lives by rational arguments.

And the problem of course is the premise that love is not rational. Whatever made you say that?

It’s true that love is not directly volitional, not in the way that it is volitional to move your arm upward or close your eyes or type the sentence “Jackdaws ate my big sphinx of quartz.” You don’t exactly choose to love someone, weighing the pros and cons and making a decision the way you might choose which job offer to take or which university to attend.

But then, you don’t really choose which university you like either, now do you? You choose which to attend. But your enjoyment of that university is not a voluntary act. And similarly you do in fact choose whom to date, whom to marry. And you might well consider the pros and cons of such decisions. So the difference is not as large as it might at first seem.

More importantly, to say that our lives should be rational is not the same as saying they should be volitional. You simply can’t live your life as completely volitional, no matter how hard you try. You simply don’t have the cognitive resources to maintain constant awareness of every breath, every heartbeat. Yet there is nothing irrational about breathing or heartbeats—indeed they are necessary for survival and thus a precondition of anything rational you might ever do.

Indeed, in many ways it is our subconscious that is the most intelligent part of us. It is not as flexible as our conscious mind—that is why our conscious mind is there—but the human subconscious is unmatched in its efficiency and reliability among literally all known computational systems in the known universe. Walk across a room and it will solve reverse kinematics in real time. Throw a ball and it will solve three-dimensional nonlinear differential equations as well. Look at a familiar face and it will immediately identify it among a set of hundreds of faces with near-perfect accuracy regardless of the angle, lighting conditions, or even hairstyle. To see that I am not exaggerating the immense difficulty of these tasks, look at how difficult it is to make robots that can walk on two legs or throw balls. Face recognition is so difficult that it is still an unsolved problem with an extensive body of ongoing research.

And love, of course, is the subconscious system that has been most directly optimized by natural selection. Our very survival has depended upon it for millions of years. Indeed, it’s amazing how often it does seem to fail given those tight optimization constraints; I think this is for two reasons. First, natural selection optimizes for inclusive fitness, which is not the same thing as optimizing for happiness—what’s good for your genes may not be good for you per se. Many of the ways that love hurts us seem to be based around behaviors that probably did on average spread more genes on the African savannah. Second, the task of selecting an optimal partner is so mind-bogglingly complex that even the most powerful computational system in the known universe still can only do it so well. Imagine trying to construct a formal decision model that would tell you whom you should marry—all the variables you’d need to consider, the cost of sampling each of those variables sufficiently, the proper weightings on all the different terms in the utility function. Perhaps the wonder is that love is as rational as it is.

Indeed, love is evidence-based—and when it isn’t, this is cause for concern. The evidence is most often presented in small ways over long periods of time—a glance, a kiss, a gift, a meeting canceled to stay home and comfort you. Some ways are larger—a career move postponed to keep the family together, a beautiful wedding, a new house. We aren’t formally calculating the Bayesian probability at each new piece of evidence—though our subconscious brains might be, and whatever they’re doing the results aren’t far off from that mathematical optimum.

The notion that you will never “truly know” if others love you is no more epistemically valid or interesting than the notion that you will never “truly know” if your shirt is grue instead of green or if you are a brain in a vat. Perhaps we’ve been wrong about gravity all these years, and on April 27, 2016 it will suddenly reverse direction! No, it won’t, and I’m prepared to literally bet the whole world on that (frankly I’m not sure I have a choice). To be fair, the proposition that your spouse of twenty years or your mother loves you is perhaps not that certain—but it’s pretty darn certain. Perhaps the proper comparison is the level of certainty that climate change is caused by human beings, or even less, the level of certainty that your car will not suddenly veer off the road and kill you. The latter is something that actually happens—but we all drive every day assuming it won’t. By the time you marry someone, you can and should be that certain that they love you.

Love without evidence is bad love. The sort of unrequited love that builds in secret based upon fleeing glimpses, hours of obsessive fantasy, and little or no interaction with its subject isn’t romantic—it’s creepy and psychologically unhealthy. The extreme of that sort of love is what drove John Hinckley Jr. to shoot Ronald Reagan in order to impress Jodie Foster.

I don’t mean to make you feel guilty if you have experienced such a love—most of us have at one point or another—but it disgusts me how much our society tries to elevate that sort of love as the “true love” to which we should all aspire. We encourage people—particularly teenagers—to conceal their feelings for a long time and then release them in one grand surprise gesture of affection, which is just about the opposite of what you should actually be doing. (Look at Love Actually, which is just about the opposite of what its title says.) I think a great deal of strife in our society would be eliminated if we taught our children how to build relationships gradually over time instead of constantly presenting them with absurd caricatures of love that no one can—or should—follow.

I am pleased to see that our cultural norms on that point seem to be changing. A corporation as absurdly powerful as Disney is both an influence upon and a barometer of our social norms, and the trope in the most recent Disney films (like Frozen and Maleficent) is that true love is not the fiery passion of love at first sight, but the deep bond between family members that builds over time. This is a much healthier concept of love, though I wouldn’t exclude romantic love entirely. Romantic love can be true love, but only by building over time through a similar process.

Perhaps there is another reason people are uncomfortable with the idea that love is rational; by definition, rational behaviors respond to incentives. And since we tend to conceive of incentives as a purely selfish endeavor, this would seem to imply that love is selfish, which seems somewhere between painfully cynical and outright oxymoronic.

But while love certainly does carry many benefits for its users—being in love will literally make you live longer, by quite a lot, an effect size comparable to quitting smoking or exercising twice a week—it also carries many benefits for its recipients as well. Love is in fact the primary means by which evolution has shaped us toward altruism; it is the love for our family and our tribe that makes us willing to sacrifice so much for them. Not all incentives are selfish; indeed, an incentive is really just something that motivates you to action. If you could truly convince me that a given action I took would have even a reasonable chance of ending world hunger, I would do almost anything to achieve it; I can scarcely imagine a greater incentive, even though I would be harmed and the benefits would incur to people I have never met.

Love evolved because it advanced the fitness of our genes, of course. And this bothers many people; it seems to make our altruism ultimately just a different form of selfishness I guess, selfishness for our genes instead of ourselves. But this is a genetic fallacy, isn’t it? Yes, evolution by natural selection is a violent process, full of death and cruelty and suffering (as Darwin said, red in tooth and claw); but that doesn’t mean that its outcome—namely ourselves—is so irredeemable. We are, in fact, altruistic, regardless of where that altruism came from. The fact that it advanced our genes can actually be comforting in a way, because it reminds us that the universe is nonzero-sum and benefiting others does not have to mean harming ourselves.

One question I like to ask when people suggest that some scientific fact undermines our moral status in this way is: “Well, what would you prefer?” If the causal determinism of neural synapses undermines our free will, then what should we have been made of? Magical fairy dust? If we were, fairy dust would be a real phenomenon, and it would obey laws of nature, and you’d just say that the causal determinism of magical fairy dust undermines free will all over again. If the fact that our altruistic emotions evolved by natural selection to advance our inclusive fitness makes us not truly altruistic, then where should have altruism come from? A divine creator who made us to love one another? But then we’re just following our programming! You can always make this sort of argument, which either means that live is necessarily empty of meaning, that no possible universe could ever assuage our ennui—or, what I believe, that life’s meaning does not come from such ultimate causes. It is not what you are made of or where you come from that defines what you are. We are best defined by what we do.

It seems to depend how you look at it: Romantics are made of stardust and the fabric of the cosmos, while cynics are made of the nuclear waste expelled in the planet-destroying explosions of dying balls of fire. Romantics are the cousins of all living things in one grand family, while cynics are apex predators evolved from millions of years of rape and murder. Both of these views are in some sense correct—but I think the real mistake is in thinking that they are incompatible. Human beings are both those things, and more; we are capable of both great compassion and great cruelty—and also great indifference. It is a mistake to think that only the dark sides—or for that matter only the light sides—of us are truly real.

Love is rational; love responds to incentives; love is an evolutionary adaptation. Love binds us together; love makes us better; love leads us to sacrifice for one another.

Love is, above all, what makes us not infinite identical psychopaths.