In the long run, we are telepathic androids | Hacker News Reader

In the long run, we are telepathic androids

economist.com

51 pointsnopinsight34 comments

Open articleView on HN

There is a string of interesting articles linked inside. I have laid them out here for quick access.

http://krugman.blogs.nytimes.com/2012/12/08/rise-of-the-robo...

http://www.economist.com/blogs/freeexchange/2013/03/labour-m...

http://www.motherjones.com/media/2013/05/robots-artificial-i...

http://www.economist.com/news/finance-and-economics/21572741...

Ignoring all of the text past the block quote:

These are not new ideas; this was all previously thought of by The Big 3 Sci-fi writers (Robert A. Heinlein, Arthur C. Clarke, Isaac Asimov), with Isaac Asimov making most of the robotics predictions. Paul Krugman (one of the articles linked to) chose economics because it was the closest available field to Asimov's Psychohistory: https://en.wikipedia.org/wiki/Paul_krugman#Personal_life

[Asimov spoilers after the quote, with non-spoiler links after]

" What’s going to happen is massive income transfers to flesh and blood human beings. These income transfers will come to be seen as a right-of-birth. This will make complete social sense once you realize that most of the beings on earth will be robots and therefore not-of-birth.

    Birth is something that happens to a minority of beings who are special, flesh and blood humans.

    The concern, as I see it, is over accepting the dual truth that robots will in all likelihood be sentient beings with an inner life just as ourselves, and they will live in grinding inescapable poverty."

[Begin Asimov spoilers]

Asimov envisioned that 50 Spacer colonies would eventually leave the solar system and colonize other systems. Their societies would develop according to the group psychosis and phobias the original floating colonies possessed, adjusted to fit their new environments. (The 50 colonies resembled a spectrum of psychosis where the social phobia of the most extreme Spacer worlds progressed to the point they only managed physical contact for sexual reproduction, which they eventually phased out in favor of asexual reproduction, I digress). Earthlings feared robots and relegated them to agricultural and infrastructure tasks while shunning them from their agoraphobic society.

https://en.wikipedia.org/wiki/Solaria is the last and most extreme of these Spacer worlds. The decision to become social isolationists with a heavy dependency on robotic labor resulted in the demise of their society, and was the type of social decay that Asimov feared might occur in humanity following the adoption of robots.

[End Asimov spoilers]

I highly recommend https://en.wikipedia.org/wiki/Isaac_asimov for anyone interested in this topic. There are some online sources for his work, and there is a large availability at http://www.abebooks.com/

Other Links:

http://www.asimovonline.com/asimov_home_page.html https://en.wikipedia.org/wiki/Isaac_Asimov_bibliography http://archive.org/details/IsaacAsimov-TheFoundationTrilogy http://openlibrary.org/authors/OL34221A/Isaac_Asimov https://en.wikipedia.org/wiki/Foundation_series https://en.wikipedia.org/wiki/Robot_series_%28Asimov%29

After years of reading Yudkowsky/LessWrong, I find it hard to take these kinds of articles seriously. They're talking about a future where AI actually exists, and not letting their imagination get even close to the true repercussions.

When you suddenly have an intelligent species running around that's vastly more intelligent than humans, it's hard to even imagine what it will mean, but "there will be vast transfers of wealth" is not even close. More like "the world will be unimaginably different", or more likely "the world will be destroyed".

For more in-depth looks at these things: LessWrong.com and intelligence.org.

The title is better than the article.

Humans and robots are merging more than they are competing. Its a symbiotic relationship.

I don't understand, why the AI should have sentience? Will it be required by some AI laws, or will it be initially a convenient interface to communicate with AI?

I think we do not have the slightest idea, what sentience is. ( Beyond a purely subjective description.) So we have one example of a general intelligence, humans, which are sentient, and no example of a general intelligence which is not. So the baseline expectation should probably be, that a AGI is sentient.

That.

And even if they get sentience, why will they get an ego, or greed (or any kind of desire that is not satisfying our desires)?

The machines probably just won't think they are slaves, because they probably won't be able to define themselves.

I agree. It seems to me that the Kevin Drum article is much clearer on this topic. AI doesn't need to be that good in order to oust humans from their jobs. His first line example is automated cars, not sentient robots.

In the case where most jobs can be automated like driving can be automated, we're not talking a technological singularity, we're just talking a massive transfer of wealth to holders of capital. Personally, I think that's the more realistic scenario, and the one that it's easiest to think through the consequences of.

I think the implication is not that future AIs -should- have sentience, it's that they -will- have sentience as a byproduct of their advancement. So the issue is how we're going to deal with that byproduct.

Why AIs should have sentience as a byproduct? For example let's say that I want AI to make an accounting software, first the AI will look for accounting laws, then will analyze my requirements and look for similar software already developed, then it will generate the code and offer me the software, where in these steps it will need sentience?

I'd argue your presupposition is false. Your bury sentience in your definitions of "look for accounting laws," "analyze my requirements," "look for similar software", "generate the code," "offer me the software."

Each of those steps require a feeling for what is right, wrong, appropriate, inappropriate, and culturally acceptable. MakeAccountingSoftware() is a little more involved than GraphPlan().

Does the feeling of right and wrong means sentience?

You can have codified right and wrong, but that's widely different. Sharia right and wrong varies from US Constitutional right and wrong varies from US-in-practice right and wrong.

I'd argue you need sentience for any meaningful right and wrong. There are a few universal rights we can't even agree on as a society, so having Evil Corp Government Agglomerated strictly defining right and wrong is, well, wrong.

The downside is a perversion of sentient AI could end up with Evil Sentient Sharia AI that shoots non-burka wearing females or an Evil Sentient Hasidic AI that shoots any non-beard-n-curls-n-hat wearing dudes or Evil Sentient Spaghetti Monster AI who force feeds carbs to everybody not wearing a strainer on their heads.

In short: determining what is right and wrong requires sentience. Just following what you are told is right and wrong does not.

For me the real question is; what is sentience? I have the hunch that sentience is more of a concept or a quality rather than a physical (let alone metaphysical) thing. Daniel Wegner once wrote a nice book about this, saying that we just feel sentient: http://mitpress.mit.edu/books/illusion-conscious-will

They probably even cannot have sentience.

Do we want our tools to have sentience? I work with a lot of sentient humans and it is not always fun, organized, or productive.

The problem with sentient beings is that they have their own ideas about what should happen next. I don't want to have to have a ton of meetings with my computer to convince it to run a new piece of code.

One problem with having super-intelligent non-sentient AIs (and by sentient, I just mean 'seems sentient to other humans'), is that they won't necessarily have human's best interest at heart.

The usual example of this concept is the paperclip maximizer: http://wiki.lesswrong.com/wiki/Paperclip_maximizer

Yeah, I'm not at all convinced sentience is a safeguard against interspecies competition. Look what we have done to other species...

AIs will be strong / have sentience because the easiest path to creating AIs that are that capable is to throw lots of computing power at the problem: simulate and then emulate specific human brains and use those resulting in silico entities as a starting point for further alterations and enhancements towards specific end goals.

That level of computing power will be available long before the nature of sentience is understood, probably within the next 15 years, and there are already heavily funded scientific projects that are starting down that road.

That doing it this way is enormously unethical isn't going to stop people from going down this road: the potential economic gains of creating knowledge workers to order are vast, and there are already philosophies (e.g. identity as pattern) that can be adapted to suggest that instantiating and destroying - or even just running - large numbers of in silico sentiences doesn't amount to mass murder. Therefore some groups will adopt those philosophies and prosper from the gains.

> emulate specific human brains

This is the most hilarious of all possible paths to artificial intelligence. A form of the cargo-cult phenomenon, creating the structures that we see lead to intelligence and sentience with no understanding of the natural underpinning thereof. It will probably work about as well as the original air fields as well.

Imagine what a simulation of a human mind would do suddenly devoid of all natural sensation and experience. Madness, almost certainly.

Why would it have no sensation and experience? Give it some from a simple simulated environment and body. Assuming it takes a freshly dead brain, vitrified, ultra-microtomed, and scanned in to be reconstructed for simulation. The mind wakes up in a sparsely decorated hospital looking room, remembers having just been out for a walk and crossing the street, then something about an oncoming car. A friendly looking doctor appears through a window.

"Morning," the doctor says, "you've been asleep a while and probably feel foggy and confused. You were injured, but your body healed a while ago. We can get to all your questions later. But first we want to test your thinking and rule out any brain damage... could you go to that computer and type out brief descriptions of the images that it shows you?"

When the simulated mind gets frustrated with typing about pictures, and stops producing good output, clear the running simulation and load it from the start again, with new pictures. This gives a strong-AI image labeling system. For more in depth tasks more simulated world interaction is needed between waking up and getting some work out of the mind, but you can also try lots of times and variations to get that right and once the mind is at the work-table store the state right before the mind begins working. Use that as the starting point each time.

Which means that the programmers of tomorrow will essentially be behavioral psychologists.

Now might be a good time to point out http://intelligence.org/?r=1

I'm at a loss. This is the dumbest thing I have read in a long time and I kept re-checking the address bar in amazement to see that this is really on The Economist. I'm sort of shocked to see such obscene detritus published by them. This was such a garbled mess. We're using Moore's Law as a predictor of strong AI? Really? Does anyone there have an education in just basic CS, let alone AI? We're just totally skirting over very serious issues as to whether strong AI is even realistic, especially by 2040. The Moore's Law stuff is just laughable. And before we can even get our head of the boring technical stuff, away we go, crafting these predictions.

Ignoring the timeframes laid out in the article, if you accept the proposal that AI will surpass human level cognition at some point in the future (big IF, I know...) then issues put forth here are something we _will_ have to deal with.

"Massive transfers of income to flesh and blood human beings."

There are so many conflating forces in action at once here. The One True Sentient AI appearance will probably be followed by The End of Practical Scarcity (food, clothing, material goods) which will require a cultural upheaval.

People extend current society metaphors to post-TOTSAI and immediately assume you'll have Google CEOs getting trillions of dollars from TOTSAI while everybody else lives in hovels. That's not a tenable world outcome.

(There are clever book-length writings to be done on these topics.)

Current technology is already growing towards establishing a fast-twich ambient psychic network amongst ourselves anyway. It'll be used both for greatness and the utterly most trivial drivel imaginable (a real time conference of a million one direction fans? imagine twitter, but worse.)

There are many, many jobs that you could eliminate without inventing sentient AI. Anything that requires implementing a plan, rather than coming up with the knowledge, should be fair game. Doctors, most of contract law, teaching, lots of logistical jobs. I imagine we're going to get there awhile before we get to strong AI.

Let's break it down:

Medical — Diagnosticians (Gregory House, M.D.) should absolutely be replaced by non-sentient decision/aggregation/clustering systems. Most medical things don't require talented surgeons. We can have PAs do almost all of the hands-on stuff. For the foreseeable future, we'll need actual doctors, nurses, and surgeons for emergency procedures.

For routine medical procedures, general purpose robots with enough dexterity to do autonomous medical operations don't exist yet. What we could have is a cluster of talented medical robot operators (in india? somewhere cheap) doing a dozen operations a day by telepresence with hospitals outfitted with standard medical robot equipment.

Law — Yes, they can be gone.

Teaching — Tricky. We don't have ractives or illustrated primers yet. Teaching is everything from helping a 5 year old write the letter S to yelling at undergrads who almost blow up a chemistry lab due to inattention.

Logistical jobs — automated cars, automated busses, automated truck driving. Yes, let's do that. I trust your 500 sensors evaluated 100 times per second are safer and more accurate than any human.

Right. My larger point though is that there are lots and lots and lots of jobs that can be eliminated without strong AI. We may succeed in destroying the demand side of the economy long before we solve the problem of making the supply side pointless.

A great deal of being a good teacher is being able to excite and motivate students. This goes from kindergarten through at least undergrad. And it applies across the socioeconomic spectrum.

"For routine medical procedures, general purpose robots with enough dexterity to do autonomous medical operations don't exist yet. What we could have is a cluster of talented medical robot operators (in india? somewhere cheap) doing a dozen operations a day by telepresence with hospitals outfitted with standard medical robot equipment."

Not per se, but we're coming damned close in this area. Surgical robots are becoming quite dexterous, and even though most of them require remote operation, it's not because of limitations in the dexterity or precision of the machines -- it's because of limitations in the AI.

Also, as with most things in medicine, the elephant in the room is liability. Who's liable if a fully autonomous machine malfunctions in a surgery, or is improperly maintained, etc? The hospital? The manufacturer? The machinist/programmer? The doctors who were present in the same room as the machine? Nobody wants to be stuck holding the bag on that one.

There is something about our community that makes for a population of presumably smart people at programming but very impractically minded about many, many other things. Law requires much more creativity and gray area sort of thinking than you suggest with replacing it with some "non-sentient" AI. Medicine? Are you serious? I'm a great fan of science fiction but it's almost like when you combine a little bit of technical knowledge with an over indulgent diet of science fiction, you get a, in my view, very naive, strange sort of thinking.

"Assuming Moore's Law keeps churning away at its normal exponential pace"

In the long run, you are all dead, as in, ceased to exist.

Reply on news.ycombinator.com