The technology is inevitable: computers will be able to think in the future. That’s not what I’m worried about. I’m worried that we humans won’t be able to handle it.
Just think how difficult it has been for us to give blacks and women equal rights. We still haven’t been able to give gays equal rights. Any computer that starts to think for itself will be locked away in cyber-Guantánamo Bay forever.
Anyone who has seen Star Trek is familiar with the myriad episodes about human rights applying to non-humans, whether they’re computers or holograms or aliens. Even though these episodes are cheesy and annoying and I find myself yelling at the TV, “Oh, for God’s sake, just give him his rights!”, I know that the writers are spot on. The human race is not as enlightened as it thinks it is.
The problem is, we have been conditioned to know that computers don’t make mistakes. Sure, they crash sometimes, but they never get things wrong. They never forget anything. When was the last time you opened up Calculator, typed in 3+4, and it gave you the wrong answer? When was the last time you went to open your Word document and the computer said, “Sorry, I forgot what you wrote”, or “I can’t remember exactly, but here’s the gist of what your essay said”? Again, I’m not talking about hard drive crashes, I’m talking about the computer working perfectly well and forgetting something. It never happens. Computers are “perfect” in that sense. What’s on your hard drive is what’s on your hard drive. Period.
Human beings, on the other hand, get things wrong all the time. We forget everything: where we left our keys, where we parked, what we said five minutes ago. We get sick, we come down with diseases, we fall in love, we get pissed off, we act crazy. Some of us really are crazy. We have murderers and thieves and kidnappers. Some of us are stupid and think gays shouldn’t be allowed to marry. Human beings are inherently flawed.
What’s more, we can’t have information installed into us. If you install Photoshop on one computer, it will work just like Photoshop on any other computer. We also don’t do the same thing the same way twice. Sometimes human beings get 3+7 correct, and sometimes we get it wrong. The computer will always give the same answer to the same input. Computers are consistent that way.
Here’s the kicker: what if it turns out these flaws that human beings have are actually necessary for conscious thought? This seems very probable to me. How could you ever design something that can think and has free will, yet never ever makes a mistake and always gives the same answer to every question? How can something think and yet at the same time be 100% predictable and reliable? That’s not thinking, that’s reading through a set of predefined algorithms and instructions. That’s what computers do now, when they don’t think.
I’m pretty sure we could design a non-thinking machine that’s badass and acts human, but if we can just look at the source code and and predict what its going to say, I wouldn’t really call that thinking. Programmers, and technology, are going to have to go beyond the current model of computing. Neural networks, and all that.
Now, what do we do when computers start making mistakes? We deal with human mistakes because we have to. Human beings kill hundreds of thousands of people every year because they suck at driving. However, we are okay with that because we know people aren’t perfect. If a self-aware, thinking computer was driving down the road day dreaming and crashed into an unfortunate pedestrian, what do you think would happen to that computer? It would be reformatted in a second. The programmer would be tracked down and thrown in jail. It would be recalled from the market. The company that manufactured it would be sued for millions of dollars. There is no way in hell the government would ever let another one be built. Of course, new human beings are born everyday, even though human beings generally suck.
Where does the responsibility lie if a computer malfunctions? If you are a genius programmer and you create a thinking computer who turns out to be a murderer, is that your fault? It’s only reasonable to expect that some thinking computers will be good, and some will be evil, just like every other species.
Remember a few years ago when somebody at NASA used inches instead of centimeters, and it screwed everything up? It wasn’t the computer’s fault, the computer was just doing what it was told. But, what happens when it is the computer’s fault? Right now, there’s always a human being to blame when a computer gets something wrong. There was a bug in the code, the input was invalid, whatever. When computers get smart enough to start making their own mistakes, what is going to happen? You put in centimeters, but the computer spits out inches: whoops!
I don’t think the human race will ever be able to handle it. We could never accept that we made something as imperfect and yet as smart as ourselves. We have a hard enough time admitting our own faults. We could never handle a race of computers with a whole different set of faults.
Still, I think it is inevitable. Somebody, somewhere, somewhen, will create Data. I hope I live to see it. Even if the human race can’t handle it.
EDIT: I believe that computers will become intelligent one day, I really do. I’m not worrying about that. I’m worrying about the reaction that the human race will have to it. Like, imagine if aliens landed on earth. Sure, some intelligent people would welcome them with open arms, but the rest would panic, and the government would unleash all its nuclear weapons at it. Plus, each country would be worried that the aliens would ally themselves with some other country, and would turn into all out war. I wouldn’t be surprised if the arrival of aliens marked the destruction of the human race at its own hands. I’m afraid that same sort of thing could happen when artificial intelligences are created.
A lot of people have been saying in the comments, “So what? Artificial intelligences are so cool, I would totally give them all the rights they want and welcome them with open arms. There would be no problem at all!” Do you really believe that? The human race is so petty and violent, we are all so selfish and power-hungry, we fight wars with each other, we hate each other, we hate blacks, we hate Mexicans, we buy Russian wives… and you really think that when androids come on the scene we’ll just welcome them into the community and be one big happy family? I would love for that to happen, but I’m worried that it won’t.
Tags: artificial intelligence, computers, Data, Star Trek
Here’s an excellent short story that is closely related to what you are talking about. Go read it; it’s worth it.
http://www.cyphertext.net/~gfi.....llars.html
There is also a filk song based on it:
Lyrics: http://www.cyphertext.net/escapekey/Collars.txt
mp3: http://www.cyphertext.net/escapekey/Collars.mp3
Clearly you haven’t spent much time thinking about this, nor researching either. Probably none in fact. I think the only time you spent on this post is the time it took you to write it.
Of course we will accept computers as smart (and smarter) than humans. Why? Because if an entity is smart enough to request human rights, then there’s no argument – it deserves them. You don’t have to argue about it at all: a thinking machine deserves the rights extended to humans – and all the associated responsibilities that go with it.
Google a mock trial featuring the world’s first sentient computer suing its owners to prevent them taking it apart.
Also, computers who think will require and demand treatment commensurate with its intelligence; and if we don’t grant it – what use is a sulking or unco-operative computer? None at all.
However, the issue is probably moot, because it is likely that the first truly intelligent computers will actually be humans running in hardware, and concurrently there will also be people who have a very large amount of hardware in their heads. So it will be impossible to draw the line at what is, and what is NOT human.
This will eventually lead to several different species and types of humans. Biological only humans, modified humans, cyborgs, humans-in-hardware and android bodies housing both machine and human intelligences, both singly, and combined.
Your assertion that the retarded abilities of human memory are required for consciousness hold no water at all. The people with photographic memories and savant abilities are not human by your definition.
I give you 1 out of 10 for this.
You get the “1″ only because I didn’t see any glaring spelling mistakes in it.
You are obviously an idiot, but I will respond anyway.
We can’t even give humans human rights. What makes you think the human race will be accepting of artificial intelligences?
I can program a computer to ask for human rights. Does that make it intelligent?
I agree that the first intelligent computers will probably be human brains running in computers, but I wouldn’t call that artificial intelligence. I would call that intelligence transference, or something.
Savant abilities do not MAKE somebody human. Computers have far more memory than people with a photographic memory, and they aren’t human. Obviously, there is some other quality to being human. A quality that machines don’t have.
Look, it’s not like I disagree with you; I totally think that intelligent machines are the future, I just don’t think the human race is going to be able to handle it. Are you really going to look at me with a straight face and tell me people aren’t going to have problems with cyborgs walking around?
It is an interesting ponder. If a computer gets smart enough to learn how to call a lawyer, and wire money to his account, it can sue for its rights. After all, a corporation can act as a “person” in a court of law, even though it is not “alive”. If a computer were given control of a corporation, it would have the legal protection of a corporate entity.
Someone recently asked (on stackoverflow.com) whether programmers would “dare to create an artificial intelligence”, obviously attempting to address the ethical issues while skipping over the technical ones.
While everyone else jumped on his case because it was “not a programming question”, I called them cowards and insisted that it was obviously a programming question, and a good one at that. Unfortunately, the person asking apparently succumbed to the negative response and deleted the question. I was already leaning against continuing my contributions to the website, but that helped settle the issue for me.
Meanwhile, back to my response to the question…
I told my fellow programmer that he should go for it, mainly because I had no concern whatsoever that he could possibly succeed, but I have found the exercise enjoyable and I hoped that he would, too.
More specifically, I have no more concern that he, or anyone else, will produce an intelligent computer than I am concerned that they will produce an intelligent hammer. Since a computer is merely a more complex tool than a hammer, yet otherwise essentially the same, then the comparison is valid. Intelligence is a qualitative shift, not a quantitative one, so no amount of programming is going to transform a hammer into an intelligent being–likewise for a computer.
In other words, the intelligence behind a hammer resides in the person swinging it. Likewise, the intelligence behind a computer resides in the programmer that instructs/commands it. The computer is capable of “replaying” the intelligence, unlike a hammer, but that does not make it intelligent.
Intelligence is a different kind of thing. It is not information. It is not the ability to replay someone else’s intelligent actions/decisions. It is not merely the “accidental” or intentional result of adding lots of complexity, especially complexity for it’s own sake. Perhaps it can best be characterized by the essential quality of “responsibility”–hammers and computers are not responsible for their actions, only people and animals are. Before AI is remotely possible, one must first explain “responsibility” as it pertains to intelligence. I think that equates to explaining life, since they seem to go hand-in-hand: good luck with that.
I’m not worried. Nevertheless, I am happy to treat everyone as equals, when they bother to act like it. I doubt that I would have any problem with an AI, if it could actually behave intelligently.
As a matter of fact, that might be quite refreshing.
Well, on the one hand, you are absolutely right about some points. I certainly would consider a calculator that gives me wrong results defective and throw it away. On the other hand, you are completely missing the point. I’ll get to that later, because on the gripping hand, you have a hole in your basic argument big enough to drive Theo de Raadt’s ego through, and I figured that woudl be the part that’s best to tackle first
.
Suppose you walked up to me and asked me to do calculations in my head. I’d probably make a lot mistake. Now say I had a calculator to help me. Suddenly, with my intelligence augmented, I’d be a much more reliable system and certainly faster at doing calculations. Or, suppose you had your writings copied by medieval monks instead of having a webserver send a copy from off its hard drive. They might drop in omissions, misspellings, etc. if they knew English they might even rephrase you – almost certainly making you seem less intelligent than you do with your writings reproduced on my screen exactly character by character.
In other words, you really can’t point at a system and say that because it does x reliably and predictably, it can’t be intelligent. It could be a system that’s intelligent as a whole and has a subsystem for performing that task well… or it could be an intelligent system that happens to be very good at the task. At a sufficient level of integration you can’t tell the difference.
Now that that’s over with, let’s get to the point that I mentioned you’d missed.
You argue that we’d sue AI research into oblivion because we would so hate seeing AIs screw up, which they inevitably would. Well, yes, even a system far more intelligent than man can end up in situations it can’t handle or simply make sheer mistakes, I agree with you on that part. It’s just that I’m afraid economics is simply not on your side here.
How likely do you think it is that the development of AIs and the subtrates for running them would happen to plateau at exactly the point where we have roughly human-level AIs and running one at roughly human-equivalent speed would take the same amount of resources as just having a human there would? Regardless of how much credence you give to Moore’s law or the law of accelerating returns, if you consider the exponential increases in capacity we’ve gained so far from having better software running on better computers that there are more of and that are better networked, it seems very unlikely that the growth of computing power would happen to stop at just the right spot to create Star Trek’s Data but not far more intelligent beings, or far many more beings that would vastly exceed human productivity in many tasks.
Yes, I am saying AIs would vastly exceed human productivity. Look at a skyscraper recently? Think it could have been built by elephants or chimpanzees or thermites? It’s not our hands or our eyes, or for that matter the power of love or divine providence, that has made us the dominant species on this planet. It’s that when the other guys are just responding to their environment as they go, we’re predicting and manipulating it. We can do that because we happen to be the ones with the big brains. Now imagine all of the technological sophistication that exists on the planet – since collective intelligence over thousands of years has created it, certainly you can imagine a single creature or small community that would be able to come up with it? Maybe in a shorter time, too? Make the software good enough, add enough computing power, and it might happen in… I dunno, a few seconds?
No court case could keep up with a company that had a strong AI to their name. Even by simply adding computation power, they could so far outthink, outdesign and consequentially outmanufacture the competition that there would effectively be no competition. (Certainly some nations might legislate against AI research. Fortunately we humans are good enough at bickering at each other that we haven’t yet made the mistake of founding a single world government, so nobody has the power to stop all AI development everywhere in the world.)
FWIW, this “story” – that strong AI would be used by a company that produces goods to kick the other companies’ asses at producing goods, regardless of if there were court cases against them, isn’t really how I think the future will go. In my books, the likely outcome is that someone finally figures out to create AI, and said AI then turns the solar system into computronium to solve some pointless problem that someone carelessly happened to ask it about, either completely oblivious to or uncaring about mankind’s existence. Well, I mean, that’s the “sorry, failed” outcome. It could be worse. Or, it could be better, but people are still working out exactly what “better” means, and it might take a while. Look forward to it happening sometime within the next ten thousand years!
tldr: Data won’t happen; if artificial general intelligence is possible at all then it will be possible to make AIs so smart or so cheap that you can’t stop all of them.
P.s. Please do not make an artificial general intelligence that feel pain or pleasure. You ever do that once and you’ll probably end up finding the matter that previously was you repurposed to represent a really large pleasure number. And even if it didn’t destroy the world it’d probably ask for human rights or something and then you’d get into so many pointless arguments.
All this and more is what you get for taking computationalism seriously.
You say computers never give the wrong answer. That’s just a fairy tale they taught you in school. In real life developers write mountains of error handling code to deal with all the errors that computers supposedly never make. They also don’t seem to get any respect from teachers for doing so. Errors are not just some once in a lifetime thing. They happen all the time every day.
Hopefully the program catches and handles those errors before ruining the the user’s WELL PROGRAMMED ILLUSION with an annoying error message. It might be true that perfect, flawless hardware and perfect software “would” always do what you told it to do, but such a thing doesn’t ACTUALLY EXIST. Just like if I find the end of the rainbow and there’s a leprechaun and a pot of gold….
But there isn’t going to be a leprechaun is there? Just because calculator has always given you the right answer doesn’t mean that it always will. Remember that when using computers.
“Please do not make an artificial general intelligence that feel pain or pleasure… it’d probably ask for human rights or something and then you’d get into so many pointless arguments.”
Remember, Data couldn’t feel a thing and he still wanted human rights. I don’t think it’s emotions that make us want rights, I think it’s intelligence. I don’t think it will be possible to create an intelligent being who does not want to live his own life and will just be our slave. Also, those arguments aren’t pointless
Sorry had to chime in here. In order for this to be remotely acceptable the “thinking” component or machine must be self programming, self powered, self assembling and self replicating. I am sure there will be much debate on this subject as the masses will want robot slaves. Ultimately ai is no closer to life than fire it will be found.
Essentially human rights cannot be given to something that by definition is not human nor alive. Just as a joining of two similars is not a marriage by definition.
There is a history of ai vs orgnaic in our culture, from dr who and dune to bsg.
Under UCC it is typically accepted that he who creates a thing has authority over it. Droids will be slaves. So if a dr soong does make a data or c3po im pretty sure we wont have many people supporting our new droid masters. Servants never become the master, or do they?
I can see it now organic human slaves to the massive electronic brain that must be “maintained”. Sounds kinda like slavery under capitalism eh?
MS, your understanding of the way software works is incorrect. Software never fails. It always does exactly what it has been programmed to do (barring hardware failure, which occurs frequently).
When a software program gives incorrect results, it was either given incorrect data, or it was programmed incorrectly, but in either case, it does exactly what it was programmed to do. If the software gives you a stack buffer overflow, it was programmed to do that in response to particular data. The programmer may have made a mistake in programming it to lead to an error, but the software nevertheless continues mechanically towards the error, with no consciousness.
Philip, I started reading the novel “Accelerando” which addresses some of these questions. You can download it for free.
“When a software program gives incorrect results, it was either given incorrect data, or it was programmed incorrectly, but in either case, it does exactly what it was programmed to do”
That is not entirely true. It does happen that a cpu computes something wrong but it is extremely unlikely. Computers that are handling very delicate matters often have two cpu’s perform the same calculation in order to minimize the chance of failure. However, I don’t think this would affect an AI very much since an AI must not be dependant on a few, discrete calculations but a huge amount of computing adding up to a great complexity where these few faults probably won’t matter much.
This whole discussion is probably about the biggest question of them all. What is Life?
Most of us nerds have probably read Isaac Asimovs robotic novels and the three Laws that governs all robots. Our own fear is the greatest threat to our survival now that we do possess this huge firepower. If we can beat the fear, I think it would be possible to have a society where robots and humans live side by side. But I don’t think robots will be given the same rights as we do. The short answer is because we believe we are better than everyone else. The long anser is related to the Big question about Life.