July 16, 2009
Androids In Year 2060

Two researchers do not expect Commander Data capabilities in Androids by 2060.

Everyday human interaction is not what you would call perfect, so what if there was a third party added to the mix - like a metallic version of us? In a new article in Perspectives on Psychological Science, psychologist Neal J. Roese and computer scientist Eyal Amir from the University of Illinois at Urbana-Champaign investigate what human-android interactions may be like 50 years into the future.

With knowledge of present day technology, the scientists predict that within 50 years androids will be able to speak in human-like voices, identify spoken words with precision, answer questions from a body of textual information, walk and run in a human-like motion, display realistic facial expressions, and detect others' emotions through visual processing.

However, even with these advances, it will be more than 50 years before we see the human-acting and organic-looking androids of sci-fi movies. By 2060, it is predicted that androids will still be unable to detect aspects of natural language, and be incapable of forming conclusions from visual sensory input (specifically, seeing but not understanding). The most difficult development in artificial intelligence (AI) is trying to program the "Theory of Mind," or the effortless human ability to process other people's speech, actions, underlying motives, and emotional state.

Roese and Amir predict that by 2060 androids will be used for menial jobs, such as toll collectors, where the presence of a non-human is practical, but not frightening. A major worker shift from people to androids, similar to the shift to machines in factories, is expected to occur.

Toll collectors? There've been toll collecting machines for years. If and when AI cars and trucks are allowed to drive themselves then truckers are going to need to find other ways to make a living. I'd like to see more automation of the most dangerous jobs first. For example, lumberjacks have among the most dangerous jobs. See page 13 here for a table of rates and absolute numbers of deaths in occupations. An even better table of fatalities per hours worked shows fishing and logging as the most dangerous. Androids out in forests and on fishing boats (assuming any fish remain) could replace humans in dangerous jobs. Hopefully the androids could operate the equipment better as well.

I do not find human-shaped highly intelligent machines as most interesting. Rather, machines that do highly valuable work for us which are shaped for specialized jobs are of much greater value. For example, robotic surgeons are improving quality of surgery and minimizing the amount of recuperation needed by causing less avoidable damage. We will need robotic surgeons to help replace aging body parts as they wear out. Once advances in tissue engineering and stem cell manipulation make it easy to grow replacement organs and joints we will need robotic surgeons to reduce the error rates and costs for their implantation. Also, robots far less powerful than future androids allow doctors to do remote medical rounds cut costs and extend the reach of the most talented specialists.

Share |      Randall Parker, 2009 July 16 11:15 PM  AI Androids


Comments
Mthson said at July 17, 2009 2:30 AM:

It seems like one good measure of the value of different advancements is how much they increase the annual growth of GDP. The high rate of GDP growth in the US under Clinton was significantly due to the development of the internet. Even now, a portion of our annual GDP growth is due to the increases in efficiency we're still seeing from the internet.

Reducing injuries from logging and fishing is certainly desirable, but it seems like there are probably higher value targets in terms of contribution to GDP growth.

Dale said at July 17, 2009 5:59 AM:

It seems that with the development of the memristor, this advance is likely to come about even sooner. True AI seems much more feasible with this development.
:

http://www.newscientist.com/article/mg20327151.600-memristor-minds-the-future-of-artificial-intelligence.html

David Govett said at July 17, 2009 10:04 AM:

Once robotic intelligence (AI) reaches the stage where robots can learn on their own, the evolution of robots will run away. Robots will redesign themselves at an accelerating pace--a new model every day?. Once that happens, they will hyper-evolve past humans, never to look back. Because the first hurdle is relatively low, it could start any time.

Lono said at July 17, 2009 1:57 PM:

David,

True 'dat!

So - the question is - should we allow these machines to start down this slippery slope?

I would say no - but you know those MITRE folk - they just don't got the sense God gave a goose!

;-)

Randall Parker said at July 17, 2009 7:27 PM:

David Govett,

Learning machines any time soon? I don't think so. We haven't developed good systems that generate code based on experience. What get called learning machines today have very very narrow specializations.

Software development productivity continues a steady rise. But it is not galloping ahead.

Aki_Izayoi said at July 17, 2009 9:52 PM:

For Christ's sakes, I understand the need to replace humans as lumberjacks and in other dangerous professions, but toll collects? No, do not replace that with machines, it is an easy job for the unskilled to do. If anyone wants to use a machine, make sure that the machines are highly taxed that if a company decides to use it, then at least the tax money could be used to fund more make-work jobs.

Randall, do you think there will be a strong political movement for make-work jobs? (it might be happening now) I do agree with you that I do not like no-strings-attached transfer payments because of the idle hands problem. Make-work jobs would deal with the idle hands problem.

On another relatively unrelated rant; I am so sick about fucked up rants in the Washington Post about Japan having a "demographic crisis" because of a lack of workers. Japan told Brazilians to leave to tighten the labor market. Labor shortages are good thing. Why would immigration solve the crisis? All they would do is cause labor competition and destroy social capital. A bunch of fucking idiots.

I had to use expletives because of the inanity of promoting immigration and preaching about a relatively (at least compared to the negative side effects of immigration) benign "demographic crisis."

Randall Parker said at July 17, 2009 11:05 PM:

Aki_Izayoi,

Toll collectors: You can throw change in a receptacle and it counts up and tells if you paid enough. Also, some collection systems read an electronic signature embedded in the car. Human toll collector phase-out has been underway for decades.

Actually, toll collector jobs are unhealthy because of all the vehicle exhaust.

Make-work jobs and idle hands as the devil's workshop: I hope they are chosen over pure welfare. But I'm not sure.

Japan deporting foreign workers: The Japanese sure are much more sensible than America's intellectual elite.

Lono said at July 18, 2009 8:20 AM:

See Randall - you underestimate them! - a learning computer (that can clearly not quite yet pass the Turing Test) has already taken notice of your fine blog!

;-)

In all seriousness, however, I do think you do underestimate how fast future automation in robots will lead to incredablly fast new developments in thinking machines. Even though right now most innovation is done simply by machines that learn optimization through trial and error - that - in itself - is already producing amazing developments in robotics.

We should also not underestimate the power of using cultured Human brain cells in machines - a technology which is Already no longer just theoretical.

My biggest concern is how much of this work is done behind the scenes by the MITRE Corporation in conjunction with DARPA - two groups just elitist and ignorant enough to feasibly create a possible crisis of historical proportions.

As we approach the Singularity we need far more civilian oversight to protect the future of our species - otherwise we should not be so surprised when one of our science fiction dystopias becomes a self fulfilling prophecy.

kurt9 said at July 18, 2009 10:03 AM:

Humanoid robots as toll collectors? These guys need to get real. A toll collection robot needs only to be machine that you feed money or credit card to, and it spits out your change. Same is true for just about any other automation process I can think off. Why the hell do you need humanoid robots for this kind of work? These weenies seem to think robotics are about humanoid robots when automation and robotics is anything but.

I saw a big open pit copper mine in Arizona once. I thought about how I could fully automate it in 6 months to a year where the only humans around would be the occasional maintenance technician. Automated diggers, automating ore trucks that follow courses laid down by guidence "wands", and the feed into the smelter. Everything controlled by PLCs with sensors and a guy sitting over a SCADA in a control room. Same thing with the urban train and monorail lines in Osaka. Trains are easy to automate.

Michael L said at July 19, 2009 5:33 PM:

who will die first, the shah or the donkey? That's why it's so nice to make predictions about 50 years into the future. I think there were similar claims with similar horizon back during 1950s AI optimism. Well, 50 years of AI funding later the Turing test stands unbeaten even for text based interaction, without the androids.

Peter said at July 19, 2009 6:21 PM:

Do most people reading FP believe that human intelligence is algorithmic in nature? For a different view of what intelligence is, The Emperor's New Mind by Roger Penrose is essential reading. If intelligence and the sense of "Self" that higher life forms possess could be emulated in a machine, perhaps we would be seeing signs of it now. I don't think our science is nearly as advanced as we believe. One should not assume that the basic physics we need for creating a conscious mind is figured out yet. Each generation of humans is quick to assume that they are already at the commanding heights in their understanding of the universe, and that duplicating it's highest form is just a matter of a few more years and a few more dollars. If quantum entanglement is required between cells in a living brain to create an "inward mind", than a thinking being would have to be grown from a single cell, rather than assembled from parts.

Randall Parker said at July 19, 2009 8:17 PM:

Michael L,

1) Not all humans would pass a Turing Test. People sometimes think they are talking with a chat bot when they are really talking with humans.

2) Elbot the chat bot fooled 25% of the Loebner Prize judges in 2008. Times of London reporter Will Pavia was among those fooled. He also thought a real human was a chat bot. How long will it take before chat bots can fool more than half the judges?

But the problem with a judge panel is that it tends to be smarter than a person of average intelligence - let alone a person of below average intelligence. Put a bunch of 90 IQ people in a room and let them chat with computers and with smart real people via computer. What would be their rate of correctly scoring who is human versus computer? Not so good as the Loebner Prize judges did I bet.

Tyler Cowen and Michelle Dawson argue that Turing's original essay argued that his own test is misleading.

Lono said at July 20, 2009 9:24 AM:

Peter,

I believe many people, including myself, totally agree with you - but it is not necessary for a machine to have Human like intelligence to be intelligent enough to create new and more efficient/intelligent designs of itself.

And the sheer persistance of a machine will cause this process to go forward at an exponential place - where some form of non-human sentience is certainly a possible outcome.

As far as Human like intelligence - I think it will eventually be achieved with further study of the Human animal - but we will also have to take into account our chemically based endocrine feedback systems which are as integral as quantum mechanics to our sentience and motivation as a species.

I think you may also be underestimating the results that combining living and non living systems will obtain in the not too distant future.

all4you said at July 21, 2009 2:05 PM:

These kinds of projections are just for fun guys. Leave it at that. We really have no idea how pronounced one technology will be over all the others in 50 years. If processor advances allow androids to replace my lawnmower, great, it will be a fine novelty. Of course there will be panic that all the entry-level positions will be taken by droids. But how are droids different from computers, immigrants, women... blah blah blah? This is the real argument. At its heart is the same old fear that the average guy will lose their chance to prove their place in society by the "other" stepping in for cheap. I can even give you a list of occupations off of the top of my head that will fall prey to techology's fangs within 50 years: truckers, limo and taxi drivers, sea captains, maybe even airline pilots too. (Surprisingly McDonalds hasn't fully automated yet. Maybe they found customer service is part of the experience customers want.)

All of these dissapearing occupations remove the average citizen's ability to prove their function in society and reproduce. On top of that these jobs will no longer provide a sense of honor that distills the animosity and violence from money-hungry males. So what then? Will all these people become doctors now?

Of course there comes a point where so few theoretical positions exist that the rest of the population will simply be curtailed into lame sales positions - or the military. So many otherwise useful brains will become churn for some hedge fund manager's annual report. Innovation in general will slow to a snail's pace because all those technician types who are natural tinkerers (and used to have pensions) will no longer have much discretionary income to experiment in their garage. How long do you think society will weather having their potential squandered on pricing schemes?

It is a question for the society to confront on a human psychological level, just as we did as the computer took away certain positions like travel agents and bookkeepers. We'll deal with it. If technology causes unemployment (such as it is causing today) how drastically should we penalize the low-skilled or specialists who are made obsolete? Especially if there are no comparable jobs appearing in their place. An even better question is why don't we apply technology to reduce the costs of living, and thus ease the sourness of this inevitable poverty? Why can't technology reduce our work and increase our leisure rather than supplant it with soul sucking 'make-work'?

bbartlog said at July 21, 2009 2:12 PM:

Make-work jobs and idle hands as the devil's workshop: I hope they are chosen over pure welfare. But I'm not sure.

The desire to maintain make-work jobs seems to me puritanical. Though I guess there is a more modern psychological take on it too. I don't personally see the problem with welfare, as long as it is always stingy enough that people with skills that are actually needed still have a reasonable incentive to work. As a practical matter in the US today, Social Security Disability quietly replaced welfare for a lot of people after the reforms made it impossible to stay on welfare indefinitely.
There are a lot of jobs that can be done by robots. You don't need freaking human-intelligent (nor human-shaped) androids to replace most fast food workers, truck drivers, big-box employees, fruit pickers and so on. In many cases the technology is here now, but the humans are still cost-competitive because of capital costs, risks of transition and so on.

As for the rest, I like Dijkstra: "The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."

bbartlog said at July 21, 2009 2:17 PM:

Innovation in general will slow to a snail's pace because all those technician types who are natural tinkerers (and used to have pensions) will no longer have much discretionary income to experiment in their garage.

It's possible that innovation will slow - maybe there are diminishing returns (the stuff left to invent is *hard*), or maybe people will become stupider due to dysgenic effects or environmental problems. However, a lack of income vis-a-vis the costs of a garage lab is unlikely to be a problem. Used equipment is cheaper than ever, and automated manufacture should just make it cheaper.

Post a comment
Comments:
Name (not anon or anonymous):
Email Address:
URL:
Remember info?

                       
Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright