Stephen Hawking’s Reddit AMA Shows That Richard Feynman Was Right: Just As Dumb As The Next Guy
Stephen Hawking has done a Reddit AMA and much of it is both interesting and about AI which isn’t what I’m about here. However, Hawking strays over into the economic effects of AI, of the robots coming to take all our jobs, and sadly gets the answer entirely wrong. Thus proving Richard Feynman correct, that even great scientists working outside their own specialty can be just as dumb as the next guy.
The point being that the rise of the robots cannot possibly make us any poorer than we are now. And that’s in the very worst case: the worst that can possibly happen is that some other people become richer and we get to jog along much as we do now. That’s also the result that is vanishingly unlikely to actually happen. What is far more likely to happen is that we all, jointly, become vastly wealthier.
Tim, you are wrong.
Gregory Clark — A Farewell To Alms: “There was always a wage at which all these horses could have remained employed. But that wage was so low that it did not pay for their feed, and it certainly did not pay enough to breed fresh generations of horses to replace them. Horses were thus an early casualty of industrialization.”
More capital will increase total output, but it won’t necessarily increase everyone’s wages. The argument is that soon robots will be more intelligent and versatile than some people. These people will then become permanently unemployed, and die out if they are not supported by others, because they will not be able to compete with robots except at a wage which doesn’t pay for enough food to keep them alive.
Artificial intelligence will quickly reach a point at which no humans can compete. Then only those people who own robots will have an income. That’s assuming the “Friendly AI” problem is solved to keep self-improving AI under control.
Robin Hanson has a book coming out next year, “The Age of Em”, about the economic implications of only-slightly-better-than-human AI. He doesn’t believe in the self-improving AI explosion hypothesis.
Read the piece. I link to a Nordhaus paper that shows wages rising strongly as a result.
I think the problems the horses had don’t apply to people in the same way. Horses did one thing- pulled things. They did so at a high cost. They did have a single alternative use- food.
I don’t think that applies to people. At least not in the UK.
I think it’s also likely that the cost of people will remain below that of robots for a long time. And that the labour of people will simply be displaced into higher value activities, as with car manufacturing and mining and other highly mechanised industries. You could also have a look at IT as a job where increasing Automation has led to more jobs and higher wages as people’s input move up the value chain.
Horses
In other words, horses aren’t able to decide to continue as they did before.
Which can’t happen to people – unless, there was some extremely authoritarian scenario so that it did in fact apply to “the humans who weren’t benefiting from robot work”?
Intelligent Al? But that’s a long way forward from here…
Stephen Hawking is strangely robophobic, for a cyborg.
The AI bit is grotesquely puffed-up. They are nowhere near robots/computers replacing people. There is lots of hype–as with driverless cars etc. That is being pushed because the scum of the state want the control they see that giving them. But it is much further off than most imagine.
Tim, you quote Hawking as saying:
‘If machines produce everything we need, the outcome will depend on how things are distributed’.
And then you said ‘The first part of that first sentence is correct’.
But the first part of that sentence is an ‘if’ clause. How can an ‘if’ clause be correct? ‘If the moon is made of green cheese’, for eg, just isn’t the sort of thing that can have a truth value.
Well the problem is, nobody knows. Because we’ve only gone through one really fast technological obsolescensing of people (the green revolution), and if we’re perfectly honest, that didn’t seem to do wonders for the wages of the proletariat, did it?
If you owned lots of robots which catered to your every whim, and which could make the potatoes for which the newly-unemployed billions would trade all their worldly goods, what incentive would you have to give some of your robots to them, as opposed to own the world?
Clearly the real-world scenario is likely to be somewhere between that of an uber-capitalist class ruling over the dispossessed and the dream of us all being grossly rich for doing little to no work. But I wouldn’t personally assume it’s going to make everyone better off. Or what the psychological effect on everyone of having 167 hours per week of “leisure” time would be.
BiG
“Because we’ve only gone through one really fast technological obsolescensing of people (the green revolution), and if we’re perfectly honest, that didn’t seem to do wonders for the wages of the proletariat, did it?”
Really? All those people who traded a life of hand-planting, weeding and harvesting for a move to a city and a – to us, shitty, to them, much improved – job in a garment factory aren’t substantially better off? The fact that food prices, relatively’, have tumbled, hasn’t made the ‘proletariat’ better of?
Mr Ecks,
“The AI bit is grotesquely puffed-up. They are nowhere near robots/computers replacing people. There is lots of hype–as with driverless cars etc. ”
I’ve been hearing about AI since the 80s and we’re not even close. Most of what gets labelled as AI is just Bayesian logic sorting things and is specific to a problem and the improvement to it requires human intervention.
Drivers who know an area still beat sat navs because of all the accumulated data in our brains. We know to avoid a road at a particular time because it’s early day at the factory, or that the Mop Fair is on, or that we’d rather take a 2 minute longer country road than a motorway because of the risk of being on there for 6 hours if there’s an accident. We’re better than computers at face recognition which is a crude algorithm. We’re only killing off black cab drivers because sat nav is cheaper than The Knowledge, not because The Knowledge isn’t better.
I think a lot of people outside computing don’t understand that the problems we solve with computers are the ones that are economic to solve with computers. Banks leapt on computers because replacing people calculating interest was a big saving and what computers are really very good at. It’s like when Cameron said “why can’t these geniuses at Google and Microsoft sort out the problems of kids seeing porn” and the answer is that that’s a really hard problem. It’s something humans are very good at, but computers aren’t.
The science fiction writer John C. Wright addressed this in his (brilliant and strange) Golden Oecumene trilogy.
It describes a far, distant future where vast machine intelligences called Sophotechs deal with most of humanity’s scut work:
Diomedes said, “Aren’t men right to fear machines which can perform all tasks men can do, artistic, intellectual, technical, a thousand or a million times better than they can do? Men become redundant.”
Phaethon shook his head, a look of distant distaste on his features, as if he were once again confronted with a falsehood that would not die no matter how often it was denounced. In a voice of painstaking patience, he said:
“Efficiency does not harm the inefficient. Quite the opposite. That is simply not the way it works. Take me, for example. Look around: I employed partials to do the thought-box junction spotting when I built this ship. My employees were not as skilled as I was in junction spotting. It took them three hours to do the robopsychology checks and hierarchy links I could have done in one hour.
“But they were in no danger of competition from me. My time is too valuable. In that same hour it would have taken me to spot their thought-box junction, I can earn far more than their three-hour wages by writing supervision architecture thought flows.
“And it’s the same with me and the Sophotechs.
“Any midlevel Sophotech could have written in one second the architecture it takes me, even with my implants, an hour to compose. But if, in that same one second of time, that Sophotech can produce something more valuable—exploring the depth of abstract mathematics, or inventing a new scientific miracle, anything at all (provided that it will earn more in that second than I earn in an hour)—then the competition is not making me redundant.
“The Sophotech still needs me and receives the benefit of my labor. Since I am going to get the benefit of every new invention and new miracle put out on the market, I want to free up as many of those seconds of Sophotech time as my humble labor can do.
“And I get the lion’s share of the benefit from the swap. I only save him a second of time; he creates wonder upon wonder for me. No matter what my fear of or distaste for Sophotechs, the forces in the marketplace, our need for each other, draw us together.”
If this demonstrates anything, then it is the pointlessness of forecasts!
And of course not forgetting the genius that was Dick Feynman.
Mr Ecks & Tim Almond: spot on. And AI progress is breathtakingly slow – I used computers to evolve neural nets in the late 60s; people are still doing that and the results are still bloody useless.
Mind you, common-or-garden automation has made huge strides. But that just means lathe operators have become personal trainers, planning consultants etc.
“If this demonstrates anything, then it is the pointlessness of forecasts!”
In time you’ll come to see it differently.
@Recusant,
The urbanisation thing got underway a good two centuries before the wholesale mechanisation of agriculture. It was enabled by increases in industrial productivity, not agricultural productivity.
Of course, once people worked out they could take machines out of the factory and put them on wheels to work fields with, the 90% of the population left working in agriculture found rather less work to go round. And their wages were not improved by the new competition from technology, we had instead a depression to show for it (and overworking of US farmland with crashing agricultural prices, though there is some debate as to the extent to which this was cause and which effect).
The point being – that yes we did eventually find other stuff for those displaced agricultural workers to do, but it took so long that those people suffered pretty seriously in the meantime.
So, if we get robots that can, pretty much overnight, do 90% of human work we will not overnight have lots of aromatherapy massage specialists and millions of former plumbers and office workers lounging around in The Caxton, we will instead have lots of unemployed people with no income and most production ending up in the hands of the robot owners. And on past experience it will take at least 10 years to fix that.
I think the gradual improvement in life for all in the industrial /agricultural revolution, is under appreciated. Every country with an industrial or agri rev saw a steady increase in population, that means less kiddies were dying. It’s a good thing when more of your progeny survive to procreate, for whatever reason, its much better than the alternative. There is no reason to suppose that future gains in technology will affect people any differently, the evidence so far is that life gets better, more people live for longer, eventually population stabilises, whereupon the static population sees an improvememt in quality of life, all gradual changes.
@ironman – best forecast I have seen this week was by Chris McCann who is apparently the founder of the eBible fellowship, and who forecast that the world would be destroyed by an apocalyptic blaze last Wednesday. Perhaps a bit extreme, but most forecasts are a triumph of hope over experience!
The problem with change is that it’s not a problem unless you are the one caught up in it. Agree with the poster above that while it will be sorted out it will take 10 years or even be a generational issue. For those caught up in between the two states it may not be a comfortable experience.
One thing I do wonder though is if only a handful of people can profit from controlling the robots them who are they going to sell to?
“So, if we get robots that can, pretty much overnight, do 90% of human work we will not overnight have lots of aromatherapy massage specialists and millions of former plumbers and office workers lounging around in The Caxton, we will instead have lots of unemployed people with no income and most production ending up in the hands of the robot owners.”
Yep. And it won’t matter, because the robots will be able to produce everything you need for free – so what do you need an income for?
And should the robot owners be so foolish as to charge for the wares their robots produce – why? what do they need money for, when robots provide them with everything they want? – then the unemployed rump of society would not be able to afford any of them, and so there would be a business for them supplying one another, doing it the old-fashioned non-robotic way.
The suffering of the starving unemployed that everyone is worried about would itself create a demand that would create the jobs they need. If people are starving, then there is a job there for a farmer. It’s precisely the fact that none of them are starving because the robots are feeding them that makes the farmer redundant, but then, if nobody is starving, what was you worrying about?
A redundant sector of society who between them have the skills needed for bare survival will always survive. If you can live off the land, for example, then you can obviously manage without a job or any income. The problem we have with the present-day unemployed is that between them they don’t even have the skills to be able to support themselves. Without the rest of society to provide for them, they’d die. But once robots rise to that level of human self-sustainment, an alternative non-robotic society ‘living off the land’ with no income becomes possible. They would simply set up their own currency and trade among themselves.
I don’t expect it to happen that way – that’s just an easier way to demonstrate that a comfortable life would still be possible outside the robot economy.
What I expect would actually happen is that the cost of living gets cheaper and cheaper, so that more and more people drop out of the rat race early – retire, or work part time – knowing they can survive for the rest of their lives on their savings. Producers produce for sale only as much as people can afford to buy – robot producers no less. So robots will have to expand their supply gradually enough to leave room for the human workers, and sell at a price the humans can always compete with. The transition to the ‘everything’s free’ society would be relatively smooth.
More than the economic aspect, what really baffles me about sharing our future with robots is the recent concern and opposition expressed about us having sex with them: the implication apparently being that they’ll all come with a fleshlight attachment that we mustn’t be allowed to use.
I’m showing my age here, but for my part any orifices and cavities my robot may possess will be used to store spare TV remotes and convenient caches of Werther’s Originals and fruit Polos.
“But let’s stick with the logical construct: the robots are making lots of cheapo stuff and we don’t get any of it.
So, where does that leave us?
Well, that leaves us right where we are today. We need to go to work each day, exactly as we do today, in order to produce, collectively, all the goods and services which we get to consume.”
Your logical construct leaves out the reality that the robots (or robot economy) and “us” (non robot economy) will be competing for resources.
Since these proposed robots are supposedly better than us at everything, that competition probably won’t last long.
If robots really do make huge progress, then – rather than lose the competition for resources – isn’t it more likely that we would simply become pets? At least those of us that were potty trained or could do tricks?
Or we become the robots.
They don’t have land to live off, either.