Robots Entering the Workforce
Posted by Jeriaska on September 28th, 2007
SIAI Interview Series – James Hughes
The following transcript of the SIAI Interview with James Hughes has not been approved by the author. Video and audio are available at the Singularity Institute website.
“I am fundamentally suspicious of the Friendly AI project, because I think it is based on millenialist expectations that we are going to have a knight on a white horse come in and solve all of humanities problems.”
Robots Entering the Workforce
“Why do you think self-willed machine intelligence is likely?”
I see evidence of greater-than-human information architectures, which have self-willed features to them already in the world. If you look at organizations, if you look at the Catholic Church, you see an organization that has institutional memory, it has desires, it has plans over time, it has ways of reproducing itself. It has specialized functions within it, just like a meta-organism. This desire to create the pure being, pure intelligence in code that does not have any of the messiness of human beings I think is ignoring the fact that it is already all around us. The question is, Will there ever be one that does not involve human beings? Would that be a superior thing or not? That is where I think I strongly diverge from the rest of these guys.
“Why do you believe in the potential for machine intelligence?”
Because I see it happening in information architectures that are distributed and partly biological, I think it could also be created in non-biological architectures. That is why I believe in the feasibility of it. I think it is also an irrational thing. I mean, I would like to live in a future in which there is robotics. I see all the progress that we have made and I read people like Kurzweil who project these kinds of trends. I fundamentally am a materialist about the nature of consciousness. That is I think another precondition of believing in some kind of Singularity or artificial intelligence. I do not believe that consciousness requires a supernatural gift from the divine. Or if it does, robots will get it.
I have actually had this argument with Christian fundamentalists. It turns out that many Christian fundamentalists completely misunderstand the nature of the soul. Theologians for thousands of years have said the soul is not eternally created at the beginning of time, and it does not survive the death of the body, because God creates Adam from dirt and puts breath into dirt. So, that is a material process, to create man. The resurrection at the end of time is supposed to be the resurrection of the body. So, the soul is a conscious being’s relationship to the divine. If you were to create in a machine a being who had the same kind of sense of awe and appreciation for the natural world and whatever other spiritual qualities we want, if you were able to create that kind of phenomenology in a machine, it would have a soul. Since I am that kind of a materialist, I do not think that it is a problem to create artificial intelligence.
“Is Friendly AI possible?”
I think Friendly AI is possible, but not very likely. The reason I think it is possible is because, again, as a materialist, I see that there are friendly people. So I know that it is possible to have a conscious, sentient being who is friendly, and I do not think that there is any fundamental barrier to creating that same kind of consciousness in a machine. I think the problem is that the way that we at least imagine machine intelligence is going to be instantiated is going to be far more malleable, far more mutable than a biologically instantiated consciousness is now. I think the biological consciousness is going to be far more mutable in the future and so we may become as problematic as the AI’s. At any rate, say you create an AI and you give it these values and morals, and if one thing gets out of jigger and it takes a reductio ad absurdam to one set of values versus another set of values, then you have got a disaster on your hands. The more powerful it is, the bigger the disaster. That is why I am fundamentally suspicious of the Friendly AI project, because I think it is based on millenialist expectations that we are going to have a knight on a white horse come in and solve all of humanities problems.
“Will AI affect the workforce?”
That is the general thesis of structural unemployment as a consequence of automation. Utopians have advocated that for a long time, the notion that we might be able to get rid of drudgery by replacing human labor with machines, engineered animals, or whatever. That expectation has come and gone. We have seen, in the 1960′s, a group of economists wrote to President Kennedy and said, “We are going to expect in the next ten, fifteen years, there is going to be massive unemployment as the result of automation. We have to start preparing for that now.” Well, they were wrong. And one of the reasons that they were wrong is that the economy does create new jobs to do new things that people want done that were not done before. Once you get robots into mining, then there are industrial jobs, and service industry jobs.
The problem that we have now is that if we get robust AI, we will be able to replace with far superior AI, human labor in service and intellectual occupations. And after that, there ain’t nowhere to go. There are only a few emotional occupations where people will probably continue to prefer humans, although if you watch the movie AI, it makes quite plausible the notion that robotic prostitution will be superior to human prostitution. We have already seen that robotic psychiatry is a helluva lot cheaper and almost as effective as human psychiatry. So I don’t know if there is any intellectual, creative occupation, anything we can do better than machines, ultimately. When we get to that point, everyone is going to be unemployed. We have to figure out, are we just going to let everybody die? Or are we going to figure out how to feed everybody? And if we are going to figure out how to feed everybody, then you can’t have to do work to get paid. That is a whole new social contract.
“What role might augmentation and intelligence amplification play in future progress?”
If you go back and reread Vernor Vinge, he says there is artificial intelligence, which could lead us into this black hole beyond which humanity may not survive. Or, we could engage in widespread intelligence amplification, augmentation of human intellectual capacities, which might get us pretty close to the same damn thing, and we would know a little more about the result because we know ourselves a little bit better. If you read Kurzweil, The Singularity is Near, there is chapter after chapter on genetic engineering and human cognitive enhancement through nanorobotics. It is only when you get to these code worshippers… and they’re my friends, okay? They’re my friends, but they’re code worshippers. They think it is the beauty of math that is going to save the world, and that we should not have any biological origins, any kind of human gene Darwinian program involved in the whole thing because the beauty of math is going to save humanity. The beauty of math does not tell us what to do! Ethics, values, empathy, it all comes from being an embodied human being. So if you want an AI architecture that is driven by human values, I say start augmenting human beings. If you figure out how to make AI, glom it onto a human being.
“Will AI’s have legal rights?”
It does not say anyplace in the Constitution that you have to be human in order to be a legal person. We recognize corporations as legal persons. I think that there is already a legal precedent. The one thing is that the 14th Amendment says you have to have been born or naturalized in the United States. Maybe if AI’s are not born in the United States, then we will have to naturalize them in the United States. But if we recognize them as legal persons, just as we have already done with corporations, then what are the preconditions for being a person? I think that is the legal discussion that needs to be had. It is even more possible and approximate that we are going to have genetically modified chimpanzees, orangutans, bonobos who meet legal personhood criteria.
“Why support the Singularity Institute?”
As I said, I think that there is insufficient engagement with the actual world of computer science, computer security, cognitive science. If you were to scholar.google.com and put in Friendly AI to see what the scholarly publications are in computer science, cognitive science or philosophy of mind journals, there’s zip. Because no one has ever heard the concept, much less taken it seriously. So if this community actually wants to impact the people who are trying to design expert systems, artificial intelligence, and cybersecurity, they have to directly engage. Those people are in academia, government, think tanks, and so far this has been a subcultural phenomenon. Now, I know there are people in the audience here who are not a part of this subculture. So, that is the first step. But, then it takes many, many more steps.
“Will the Singularity ever become mainstream?”
I think all ideas go through certain kinds of predictable stages. Some get stalled and never reach maturation. But one of the first stages is what I think of as the “salon stage.” It’s where a group of people who generally agree with one another, sometimes they violently disagree with one another, but they are all really smart and like to get together and talk about ideas, like the French salons in the 1700s. You could have a wonderful conversation about whether women should have the vote. “It’s such a great idea, women having the vote. Let’s have a conversation about it!” And then you have the suffragette movement in the late 19th century and women are starving themselves to death and people are blowing shit up, and you have the Workingman’s International having strikes, and people are getting shot in the streets. Then you move from being a salon, being a tea party, to actually engaging with the politics of your day. So, the degree to which the public and government take seriously the kinds of radical social changes and radical threats that the SIngularity Institute is talking about is the degree to which they either have to get serious about political engagement or they are going to be wiped off the map. The only people who are going to be talking about this are the Christian right and people in the Pentagon.
“Are we heading toward a Singularity?”
Well, I do believe in discontinuous moments in history, revolutionary moments in history when everything changes very quickly. I think there is a certain amount of ahistoricism about the singularitarians and their understanding of human history. There have been periods when things moved very, very quickly, like the late 1800s. There was a lot of technological innovation, a lot of cultural change, political change. So, I do believe there are periods of change and periods where things are not so rapid. And I also believe there are periods like the transition from being a non-symbol using species to being a symbol using species, in which case you do not have anything to do with what came before. That might be the kind of transition we are in, or it may be more of a discontinuity where we go into a new phase change as a society, like going from a wage and work-based society to a society with basic income guarantee and no necessity to labor. That is going to be a very fundamental change, whether super AI achieves its goals or not, if we had robots doing all the labor. So, yes, I do believe there is going to be a fundamental discontinuity. I see it happening before 2050. But, why? I’m not a supergenius who can map out all the different trends. I don’t think anybody can, really.
“Will there be marginalization?”
I run a think tank called the Institute for Ethics and Emerging Technologies, and one of our chief theoreticians is a very staunch and a very harsh critic of singularitarianism and transhumanism in general. One of the things that he is very critical of is the effort to subculturalize these kinds of phenomena. What is the advantage if you are going to the Pentagon and you want to talk about the threat from artificial intelligence attacking major information infrastructures, what advantage is it that you say, I’m a singularitarian and I represent a singularitarian organization? That does not give you anything. It does not give you any credibility, it only destroys your credibility. From a practical politics point-of-view, he is right. I worked for many years as a socialist organizer. Exactly the same problem. If you go into the local democratic candidate’s office and say, “I represent the democratic socialists of America. We want to support your candidacy and we are going to publicly endorse you!” They will say, if you want to support my candidacy, don’t publicly endorse me and don’t tell anybody that you’re a socialist. So, it is exactly the same thing with something as wacky and far-out as Singularitarianism. Now, the other part of that is, the reason I volunteered for that candidate was because I was a socialist and I wanted to advance universal healthcare and workers’ rights. I would not have gone if I had not become a socialist, so Singularitarians are developing a cadre of activists who are going to hopefully go out and try to engage in a mission in computer science, academia, politics policy, if they do it right. And they are going to face that same tension. What is the use of being a Singularitarian when you are trying to do that?
