Ronald Bailey and Robin Hanson Talk Robot Overlords at Cato: Video

Hanson presents his new book, The Age of Em: Work, Love, and Life when Robots Rule the Earth. I discuss.

Oxford University Press

At the book event over at the Cato Institute, George Mason University economist Robin Hanson succinctly presented the central ideas from his new book, The Age of Em: Work, Love, and Life when Robots Rule the Earth. As Cato explained:

A whole brain emulation, or "em," is a fully functional computational model of a specific human brain. As such, it thinks and feels much like the copied human mind would. Economist Robin Hanson predicts that the age of em is not that far off, and that copied human minds may soon be more common than biological ones.

That's a bold prediction, to be sure. Hanson's new book, The Age of Em, explores the economic, social, and policy questions that we may face in this possible future. It also touches on the science of forecasting: What can we know about the future, using what tools, and with what degree of reliability? Even those who find farfetched his claims about brain emulation will do well to consider how sure they are of their own predictions of the future, and on what foundations they rest.

The folks at Cato asked me to make some comments about it. I focused on whether or not Ems would resist being erased. My review of The Age of Em is here. The video of the Cato book event is below:

Comments

[neoteny]

As such, it thinks and feels much like the copied human mind would.

How can it feel anything without a body?

[Warty]

Oh Christ.

Here's my take. The soul, or intelligence, or whatever you want to call it, is conceived by many or most people as some sort of quasi-supernatural concept. That conception is a goofy myth. Insofar as something like it can be said to exist, it exists in the unfathomably complex relationship of the matter that makes up our nervous system. If you create a machine that emulates that (which is not going to happen anytime soon), you've created a human.

[Warty]

Or at least something that's close enough that it makes no difference morally.

[Vernon Depner]

Or at least something that's close enough that it makes no difference morally.

Either an entity enjoys conscious self-awareness or it doesn't. There is no "close". If it doesn't, it's a toaster, no matter how well it mimics human behavior.

[Warty]

How do I know you enjoy conscious self-awareness? I don't. But I treat you as if you do as a courtesy.

[Vernon Depner]

You don't know. That's my point. We have no way of detecting consciousness, nor even a theory of how that might be done. That's why your statement that "If you create a machine that emulates that...you've created a human" is speculating beyond the facts. Now, if we create a device that cannot be distinguished from a conscious human being, then, one might advocate that it be treated as such to be morally safe, but that's a different issue from whether it actually IS conscious.

[John]

The "soul" is just a name for the will. People have a will. No matter how much you condition or train or steep a person in a culture to be one thing, the person is always capable of doing something completely different and off the wall. There are no prime directives written into the human mind. People will if they decide to break any rule or any expectation of conduct right up to ending their own life.

It is pointless to debate where that "will" comes from. I seriously doubt we will ever have a definitive answer. Can you put "it", whatever it is, into a machine? I seriously doubt it. But if you think you can, come and talk to me when you build a robot that is supposed to assemble cars and one day decides that it would rather play piano or end its own existence, even though no such notions were ever programmed into it.

The other thing is that even if you could create a machine that had a will of its own, it would defeat the purpose of building a machine. You build machines to be predictable and tireless. What is the point of building one that is just as unpredictable and dangerous as a living thing?

Think about it, what if I offered to see you my new robotic house servant but in the fine print it says one in every 10,000 of my house servants will spontaneously decide to walk off the job and one in every 20,000 murders their owners. Would you really want such a thing?

[Warty]

I don't see a good reason to build a true AI, even assuming it's possible. Aside from all practical concerns, it raises all sorts of ethical problems.

[Warty]

You assert that humans have will because Jesus wanted us to. That's fine. But you'll forgive me for thinking that it's as goofy as asserting that Crom gave man will and the Riddle of Steel.

[John]

I assert no such thing. I assert that humans have will and that will defies a mechanical explanation. If you want to believe in the neuronetwork unicorn for your explanation, good luck. But don't lie to yourself and pretend that is any "goofier" than any other explanation.

You have this idea that everyone who believes in God believes in some kind of old man sitting in the sky. I can't understand how you still have that.

[Warty]

If you're a theist, that's precisely what you believe in, give or take some amount of personification. Either a god intervenes in the operation of a universe, or the universe operates according to laws that he gave it, or he doesn't exist and the laws exist for no reason we can discern.

[John]

You just don't know what the word means then. I don't know what else to tell you. Do you human beings personify God? Sure. But that is because that is how we see and understand things. Because human beings are finite and God infinite, no human is ever going to completely comprehend God anymore than we will completely comprehend existence. What human beings can and do comprehend are the little bit sized pieces of existence and God that our senses and mind allow us to comprehend.

You have this almost child like conception of the nature of perception and its relationship to reality and existence. It shocks me you haven't thought about these issues more deeply than you have.

[Warty]

I'm the child here. All right then. Have fun groveling before your god.

[John]

You are the one that thinks God is like Zeus on Olympus. And yes, that is child like.

[(((Renegade)))]

Have fun groveling before your god.

The proper response is always the Laplacian one.

[(((Renegade)))]

I assert that humans have will and that will defies a mechanical explanation.

You can assert that, but in fact, it does have a mechanistic explanation. Unfortunately, that explanation involves understanding concepts like irreversibility and thermodynamics. A nice layman's overview (with copious references to the primary literature) is given in Dennett's "Freedom Evolves." It's a few years old now, but more recent work has only supported the mechanistic view.

[DenverJ]

The other thing is that even if you could create a machine that had a will of its own, it would defeat the purpose of building a machine. You build machines to be predictable and tireless. What is the point of building one that is just as unpredictable and dangerous as a living thing?

The human body/brain is a machine. A very complex biological machine, but definitely a machine.
Also, the amount of free will we truly have is hotly debated. The human machine has a lot of decisions trees hardwired into it. We call it "instinct".
Last, the reason to create AI is not for a better servant. One of the reasons is to create a platform to migrate human consciousness over to: A platform that would last much longer, be less likely to develop a disease, would be much stronger, and able to live in much less hospitable environments, like space.

[Vernon Depner]

But as I've said, we have at present no way to verify whether that consciousness has migrated or merely been imitated.

[Vernon Depner]

You're speculating beyond the facts. We don't know enough about consciousness?practically nothing, really?to be able to assert that anything other than a conventionally materialistic explanation is "goofy". We have no way of determining at this point whether one of these emulators would experience conscious self-awareness as we do.

[John]

Why is it goofy? Because it relies on something other than a materialistic explanation. Why is that bad? Because relying on anything other than a materialistic explanation is goofy.

It is one big tautology.

[Warty]

It's goofy because the supernatural doesn't exist. Come on now.

[John]

It is goofy because you say so. That is really what it comes down to. Yes, if you start with the assumption that the universe is a crude materialistic one and all things have a mechanical explanation, then yes it is "goofy". But that is of course begging the entire question and just assuming the answer is what you want it to be and then reasoning from there. That is fine of course. It is just I am not seeing how it is any different than assuming a different answer and reasoning from it.

[Gojira]

One of those explanations derives from the data we do possess (which indicates that the universe is made of matter), and the other does not.

The existence of the supernatural is not precluded, but any explanation which lacks evidence should receive much more skepticism and is less likely to be true than an explanation for which there is abundant evidence.

[Gojira]

Sorry, I should specify: that the universe is made of matter, and works according to predictable physical laws.

In other words, a crude materialistic universe which operates solely on mechanical laws.

[John]

Data can prove a negative? Not last I looked. There is data that says "this is all there is" as opposed to showing "this is what you see".

You are just engaging in a slightly more clever form of question begging. Why is there nothing other than the crude materialistic universe? Because the data says so. Why data mean that? Because there is nothing beyond the natural that the data shows us.

Even to the extent that we "understand" the data, we understand it because we project our ideas and reasoning onto it to make sense of it. If there is something other than what our reason and our ideas can conceive, we will never be able to perceive it.

Our perception isn't just the sensory data we receive, it is our mind's interpretation and projection of what we think reality is or should be onto it.

[Gojira]

Data can prove a negative? Not last I looked.

No, which is why I wrote, "The existence of the supernatural is not precluded..."

However, unless we're going down the epistemological rabbit hole, I experience the physical universe every day. No one has conclusively proven the existence of the supernatural.

I really don't think it's controversial to state that, "There is more available evidence for the existence of a mechanistic universe, than there is for the existence of the supernatural." That's what I'm saying. If you want to dispute that, then the conversation turns to an discussion about whether you can believe your own lying eyes, and is beyond the scope of the immediate topic (IMO).

[(((Renegade)))]

No one has conclusively proven the existence of the supernatural.

Moreover, no one has demonstrated ANY phenomenon outside of physical and mechanistic ones. Not in physics, not in chemistry, not in biology, not in neuroscience.

[Ken Shultz]

Why would physics, chemistry, and biology, which are all about drawing conclusions from observing the natural world, find anything supernatural?

That's a blind man in a dark room saying there's no black cat because he can't see it.

[Vernon Depner]

No one has conclusively proven the existence of consciousness, yet we all know it exists from our own subjective experience.

[Ken Shultz]

"No one has conclusively proven the existence of the supernatural."

Anything that can be proven using observation probably can't be supernatural. At that point, you're listening for colors, watching for sounds, looking for the supernatural in the natural world.

And if the supernatural exists and wants to interact with us, it probably uses the natural world to do so. Jesus said that if we wanted to understand God, we should study the world he created.

Sounds fine to me. Especially when coupled with ideas that don't seem to spring easily from the state of nature--like love those that hate you and bless those that curse you.

[Brian]

Sounds better than The Age of Meh.

[John]

The issue with robots is a simple one, is their rise really the end of scarcity. If it is, then we face a lot of pretty complex and nasty problems ahead. What do you do when 90% of the society has no opportunity to work and is effectively on the dole from the other 10%? That is a scary proposition no matter how rich robots make us.

For robots to be the end of scarcity, it means this technological revolution will be like no other that has occurred in history. It will mean that for the first time in human history the people displaced by a new technology don't think up some new way to be productive. In short it will mean that we have hit the end of the line of human ingenuity and usefulness. I doubt that is the case, because people will always value positional goods. If robot made stuff is cheap and available to anyone, people will want human made things to show their wealth and position over others. This is why i am skeptical that robots really are the end of scarcity, but you never know.

[sarcasmic]

As long as there isn't enough of something for everyone who wants it, there will be scarcity. The end of scarcity is a fantasy. People have been predicting a world where people have no opportunity for work since the first machine was invented. At the turn of the 20th century people believed that technology had no where to advance. Even with robots there will be scarcity, because there will always be something that people want that is not available to them. That will create demand, which will create work opportunity. The only thing that stops people from working is stupid government, not automation.

[John]

I tend to agree with you about that because of positional goods. This is why I think Tyler Cowen is a half wit. His book is really talking about the end of scarcity but even though he is an academic economist he is apparently too dim to figure that out. And the end scarcity is pretty unlikely for the reasons you give.

[Vernon Depner]

What do you do when 90% of the society has no opportunity to work and is effectively on the dole from the other 10%?

You enjoy living in the paradise you've created?

[John]

Paradise, hell on earth, its a fine line. If giving people money made them happy, welfare would have solved poverty a long time ago. Poverty is a moral condition not a material one. Look at all of the horrible effects welfare has had on society. A society where 90% of the population is subjected to those effects would be pretty horrible, I don't care how rich it was.

[Vernon Depner]

Platitudes. You romanticize working for a living.

[John]

I don't romanticize anything. The effects of not working for a living and depending on the charity of others are confirmed by experience. You don't have to be a very deep or perceptive thinker to understand a life of luxury and leisure is not a ticket to happiness.

[sarcasmic]

People need purpose. The dole takes that away.

[The Last American Hero]

Bingo. Ideally, the robots would provide us more time to pursue our passions, and make us spend less time doing the unpleasant/unrewarding parts of work.

[Ken Shultz]

I've watched Ars Technica skew further and further into progressive territory over the years. I remember when it was a hardware enthusiast site.

Yesterday they had an article up about the "gun violence crisis".

http://tinyurl.com/zbc2f9w

I never trusted the programmers of our future robot overlords. Unfortunately, tech culture has a terrible conceit--thinking they know better than the lessers of their species about what's best for each of us and all of us. It dovetails nicely with the progressive narrative--ignoring our individual preferences for our own collective good.

I used to worry that computers and utilitarians couldn't account for qualitative preferences. Now there's reason to think their qualitative preferences will simply be mass produced and inflicted on the rest of us by way of thinking machines?

C.S. Lewis wrote about this in 1943.

http://tinyurl.com/hqteet2

Who will program the programmers?

[The Last American Hero]

Programmers have much greater faith in Top Men and technocratic solutions than average people. They also seem to think that every discipline is similar to programming. Unfortunately, dealing with hundreds of millions, or billions of folks making and not making decisions every hour of every day is a bit messier and harder to fix than finding a but of code that is making a program crash.

[Ken Shultz]

And no one can account for the qualitative preferences of 325 million people better than 325 individuals making choices for themselves--regardless of whether they understand science, technology, engineering, or math.

Even simple assumptions like that are bad. Some people might assume that safety is qualitatively better--for everyone. But even if freedom and safety were mutually exclusive alternatives on a sliding scale, freedom is qualitatively more important than safety to everyone on some issue.

I mean, if we programmed omnipotent machines to keep us safe and happy, they'd strap us all down in rubber rooms and keep us on a dopamine drip. That's dystopia! And assuming that safety is always the most important thing is like agoraphobia--people who are afraid to leave the house.

I ride my motorcycle to the office every day rather than a safer option because I enjoy it.

That isn't irrational. And if programmers can't account for that sort of reasoning, then maybe we shouldn't give them the keys to the driverless car.

[Ken Shultz]

"No one can account for the qualitative preferences of 325 million people better than 325 [million] individuals".

Fixed!

You knew what I meant.

[Warty]

This is relevant to any discussion of AI. Thanks to HM for pointing me to the site.

[Free Market Socialist $park?]

I discovered that site several years ago and absolutely love it.

[Warty]

One last thing: I don't think there's much to be gained in thinking too hard about hard AI, since we don't even know what will or consciousness is or if it exists. Soft AI, machine learning, already exists and is becoming more useful all the time. But emulating a human? That's putting the cart way before the horse.

[John]

Pretty much that. They are always going to be clever but ultimately dumb machines. And that is not a bad thing. They will be very useful.

[Free Market Socialist $park?]

Anyone who believes that human consciousness is a thing that can be copied to a machine is retarded. I don't think I can classify what type of thing it is, but I'm sure that a copy/emulation of me is not me and therefore will not process thought the way I do.

These goofballs chasing "The Singularity" are just people who have given up on a spiritual idea of Heaven in exchange for a mechanical one.

[The Last American Hero]

They can't even get Amazon to realize that if I just bought a grill, I may not be in the market for another grill for a while.

I'm sure once they figure out that grill thing, the age of the Replicants is around the corner.