= 27d567bc2d48d9f40e9ccc20ea370b15)
A few thoughts on a Friendly AGI (safe vs friendly, other minds problem, ETs and more)
Friendly AI is an idea that I find to be an admirable goal. While I'm not yet sure an intelligence explosion is likely, or whether FAI is possible, I've found myself often thinking about it, and I'd like for my first post to share a few those thoughts on FAI with you.
Safe AGI vs Friendly AGI
-Let's assume an Intelligence Explosion is possible for now, and that an AGI with the ability to improve itself somehow is enough to achieve it.
-Let's define a safe AGI as an above-human general AI that does not threaten humanity or terran life (eg. FAI, Tool AGI, possibly Oracle AGI)
-Let's define a Friendly AGI as one that *ensures* the continuation of humanity and terran life.
-Let's say an unsafe AGI is all other AGIs.
-Safe AGIs must supress unsafe AGIs in order to be considered Friendly. Here's why:
-An unsafe AGI is likely to be built at that point because:
-Some people will find the safe AGI's goals unnacceptable
-Some people will rationalise or simply mistake that their AGI design is safe when it is not
-Some people will not care if their AGI design is safe, because they do not care about other people, or because they hold some extreme beliefs
-Therefore, If a safe AGI does not prevent unsafe AGIs from coming into existence, humanity will very likely be destroyed.
-The AGI most likely to prevent unsafe AGIs from being created is one that actively predicted their development and terminates that development before or on completion.
-So to summarise
-Oracle and Tool AGIs are not Friendly AIs, they are just safe AIs, because they don't suppress anything.
-Oracle and Tool AGIs are a bad plan for AI if we want to prevent the destruction of humanity, because hostile AGIs will surely follow.
(**On reflection I cannot be certain of this specific point, but I assume it would take a fairly restrictive regime for this to be wrong. Further comments on this very welcome.)
Other minds problem - Why should be philosophically careful when attempting to theorise about FAI
I read quite a few comments in AI discussions that I'd probably characterise as "the best utility function for a FAI is one that values all consciousness". I'm quite concerned that this persists as a deeply held and largely unchallenged assumption amongst some FAI supporters. I think in general I find consciousness to be an extremely contentious, vague and inconsistently defined concept, but here I want to talk about some specific philosophical failures.
My first concern is that while many AI theorists like to say that consciousness is a physical phenomenon, which seems to imply Monist/Physicalist views, they at the same time don't seem to understand that consciousness is a Dualist concept that is coherent only in a Dualist framework. A Dualist believes there is a thing called a "subject" (very crudely this equates with the mind) and then things called objects (the outside "empirical" world interpreted by that mind). Most of this reasoning begins with Descartes' cogito ergo sum or similar starting points ( https://en.wikipedia.org/wiki/Cartesian_dualism ). Subjective experience, qualia and consciousness make sense if you accept that framework. But if you're a Monist, this arbitrary distinction between a subject and object is generally something you don't accept. In the case of a Physicalist, there's just matter doing stuff. A proper Physicalist doesn't believe in "consciousness" or "subjective experience", there's just brains and the physical human behaviours that occur as a result. Your life exists from a certain point of view, I hear you say? The Physicalist replies, "well a bunch of matter arranged to process information would say and think that, wouldn't it?".
I don't really want to get into whether Dualism or Monism is correct/true, but I want to point out even if you try to avoid this by deciding Dualism is right and consciousness is a thing, there's yet another more dangerous problem. The core of the problem is that logically or empirically establishing the existence of minds, other than your own is extremely difficult (impossible according to many). They could just be physical things walking around acting similar to you, but by virtue of something purely mechanical - without actual minds. In philosophy this is called the "other minds problem" ( https://en.wikipedia.org/wiki/Problem_of_other_minds or http://plato.stanford.edu/entries/other-minds/). I recommend a proper read of it if the idea seems crazy to you. It's a problem that's been around for centuries, and yet to-date we don't really have any convincing solution (there are some attempts but they are highly contentious and IMHO also highly problematic). I won't get into it more than that for now, suffice to say that not many people accept that there is a logical/empirical solution to this problem.
Now extrapolate that to an AGI, and the design of its "safe" utility functions. If your AGI is designed as a Dualist (which is neccessary if you wish to encorporate "consciousness", "experience" or the like into your design), then you build-in a huge risk that the AGI will decide that other minds are unprovable or do not exist. In this case your friendly utility function designed to protect "conscious beings" fails and the AGI wipes out humanity because it poses a non-zero threat to the only consciousness it can confirm - its own. For this reason I feel "consciousness", "awareness", "experience" should be left out of FAI utility functions and designs, regardless of the truth of Monism/Dualism, in favour of more straight-forward definitions of organisms, intelligence, observable emotions and intentions. (I personally favour conceptualising any AGI as a sort of extension of biological humanity, but that's a discussion for another day) My greatest concern is there is such strong cultural attachment to the concept of consciousness that researchers will be unwilling to properly question the concept at all.
What if we're not alone?
It seems a little unusual to throw alien life into the mix at this point, but I think its justified because an intelligence explosion really puts an interstellar existence well within our civilisation's grasp. Because it seems that an intelligence explosion implies a very high rate of change, it makes sense to start considering even the long term implication early, particularly if the consequences are very serious, as I believe they may be in this realm of things.
Let's say we successfully achieved a FAI. In order to fufill its mission of protecting humanity and the biosphere, it begins expanding, colonising and terraforming other planets for potential habitation by Earth originating life. I would expect this expansion wouldn't really have a limit, because the more numourous the colonies, the less likely it is we could be wiped out by some interstellar disaster.
Of course, we can't really rule out the possibility that we're not alone in the universe, or even the galaxy. If we make it as far as AGI, then its possible another alien civilisation might reach a very high level of technological advancement too. Or there might be many. If our FAI is friendly to us but basically treats them as paperclip fodder, then potentially that's a big problem. Why? Well:
-Firstly, while a species' first loyalty is to itself, we should consider that it might be morally unsdesirable to wipe out alien civilisations, particularly as they might be in some distant way "related" (see panspermia) to own biosphere.
-Secondly, there is conceivable scenarios where alien civilisations might respond to this by destroying our FAI/Earth/the biosphere/humanity. The reason is fairly obvious when you think about it. An expansionist AGI could be reasonably viewed as an attack or possibly an act of war.
Let's go into a tiny bit more detai. Given that we've not been destroyed by any alien AGI just yet, I can think of a number of possible interstellar scenarios:
(1) There is no other advanced life
(2) There is advanced life, but it is inherently non-expansive (expand inwards, or refuse to develop dangerous AGI)
(3) There is advanced life, but they have not discovered AGI yet. There could potentially be a race-to-the-finish (FAI) scenario on.
(4) There is already expanding AGIs, but due to physical limits on the expansion rate, we are not aware of them yet. (this could use further analysis)
One civilisation, or an allied group of civilisations have develop FAIs and are dominant in the galaxy. They could be either:
(6) Dominators that tolerate civilisations so long as they remain primitive and non-threatening by comparison.
(7) Some sort of interstellar community that allows safe civilisations to join (this community still needs to stomp on dangerous potential rival AGIs)
In the case of (6) or (7), developing a FAI that isn't equipped to deal with alien life will probably result in us being liquidated, or at least partially sanitised in some way. In (1) (2) or (5), it probably doesn't matter what we do in this regard, though in (2) we should consider being nice. In (3) and probably (4) we're going to need a FAI capable of expanding very quickly and disarming potential AGIs (or at least ensuring they are FAIs from our perspective).
The upshot of all this is that we probably want to design safety features into our FAI so that it doesn't destroy alien civilisations/life unless its a significant threat to us. I think the understandable reaction to this is something along the lines of "create an FAI that values all types of life" or "intelligent life" or something along these lines. I don't exactly disagree, but I think we must be cautious in how we formulate this too.
Say there are many different civilisations in the galaxy. What sort of criteria would ensure that, given some sort of zero-sum scenario, Earth life wouldn't be destroyed. Let's say there was some sort of tiny but non-zero probability that humanity could evade the FAI's efforts to prevent further AGI development. Or perhaps there was some loophole in the types of AGI's that humans were allowed to develop. Wouldn't it be sensible, in this scenario, for a universalist FAI to wipe out humanity to protect the countless other civilisations? Perhaps that is acceptable? Or perhaps not? Or less drastically, how does the FAI police warfare or other competition between civilisations? A slight change in the way life is quantified and valued could change drastically the outcome for humanity. I'd probably suggest we want to weight the FAI's values to start with human and Earth biosphere primacy, but then still give some non-zero weighting to other civilisations. There is probably more thought to be done in this area too.
Simulation
I want to also briefly note that one conceivable way we might postulate as a safe way to test Friendly AI designs is to simulate a worlds/universes of less complexity than our own, make it likely that it's inhabitants invent a AGI or FAI, and then closely study the results of these simluations. Then we could study failed FAI attempt with much greater safety. It also occured to me that if we consider the possibilty of our universe being a simulated one, then this is a conceivable scenario under which our simulation might be created. After all, if you're going to simulate something, why not something vital like modelling existential risks? I'm not sure yet sure of the implications exactly. Maybe we need to consider how it relates to our universe's continued existence, or perhaps it's just another case of Pascal's Mugging. Anyway I thought I'd mention it and see what people say.
A playground for FAI theories
I want to lastly mention this link (https://www.reddit.com/r/LessWrongLounge/comments/2f3y53/the_ai_game/). Basically its a challenge for people to briefly describe an FAI goal-set, and for others to respond by telling them how that will all go horribly wrong. I want to suggest this is a very worthwhile discussion, not because its content will include rigourous theories that are directly translatable into utility functions, because very clearly it won't, but because a well developed thread of this kind would be mixing pot of ideas and good introduction to common known mistakes in thinking about FAI. We should encourage a slightly more serious verison of this.
Thanks
FAI and AGI are very interesting topics. I don't consider myself able to really discern whether such things will occur, but its an interesting and potentially vital topic. I'm looking forward to a bit of feedback on my first LW post. Thanks for reading!
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
= f037147d6e6c911a85753b9abdedda8d){kind=logo}
Comments (44)
Maybe they are just less pessimistic.
It's not that non dualists don't.
Which is to say that physicalists accept consciousness and subjects so long as they are just matter doing stuff.
An eliminativist doesn't. Other physicalits do, and consider themselves proper.
That doesn't follow. Taking consciousness seriously isn't that exclusive. Even qualiaphobes don't think torture is OK.
Thanks for engagement on the philosophical side of things. I'll politely beg to differ on a couple of points:
Pessimism/optimism doesn't seem like an appropriate emotion for rational thought on the nature of universe or mind? Perhaps I misunderstand.
I fairly certain strict Monists don't believe in a subject as a metaphysical category the way Dualists do.
I don't think they do. For example, in the case of "subjective experience" they'd not regard the subjective part as meaningful in the way a Dualist does. Only Dualists see the subject and object as legitimate metaphysical categories - that's what the "Dual" in Dualism is!
There's plenty of other reasons for that beyond accepting the subject. It doesn't follow that opposition to torture implies belief in consciousness.
Again thanks for engaging on the philosophy, it is actually much appreciated.
I can (and do) believe that consciousness and subjective experience are things that exist, and are things that are important, without believing that they are in some kind of separate metaphysical category.
I understand, but I just want to urge you to examine the details of that really closely, starting with examining "consciousness"s place in Dualist thought. What I'm suggesting if many of us have got a concept from a school of thought you explicitly disagree with embedded in your thinking, and that's worth looking into. It's always alluring to dismiss things that run contrary to the existence of something we feel is important, but sometimes those rare times when we question our core values and thought that we make the most profound leaps forward.
I urge you to be less of a dismissive, lecturing dick when talking about consciousness.
What concept? The concept of consciousness or the concept of consciousness as fundamental?
Maybe we have a concept of consciousness because we are conscious.
Maybe we have a concept of gods because we are gods? It don't think that logic works. If someone is physicalist then they can't assume consciousness a priori. In which case, how can observation of brains and behaviours justify a concept like consciousness? The only way it can arise is out of a mind-body separation (Dualism).
Maybe we gave a concept of rocks, because there are rocks.
It isn't a question of all sentences of the form "we have a concept of X because X exists" being analytically true. It is a question of having evidence of specific things. The other minds problem is the other minds problem because we all have evidence of our own minds.
1 I am aware of my own consciousness
2 my own consciousness must be an outcome of the physical operation of my brain
3 similarly operating brains must be similarly conscious
See other post. Cheers for discussion.
I mean pessimism about the prospect of physicallisticaly explaining consciousness.
Materialists and physicalist don't believe in subjectivity and consciousness the way that dualism and idealists do...and that doesn't add up to only dualists believing in consciousness. What is unclear is why an AI would need to believe in consciousness in the way a dualist does,a s a separate ontological category, in order to be ethical.
I wouldn't say that AI needs to refer to consciousness at all to be ethical. I think it will be better if we design it without reference to the problematic concept, to be much safer. Consciousness is a pretty deeply embedded word in our culture's thinking, so much that we don't remeber the philosophical context in which it is derived. I just want to get people to ask, where does the concept come from? Until people decide to ask that, this is probably all crazy talk to them.
I can't claim to have read every one's thought ever, so perhaps someone claiming to be a Monist believes that, but I do know you can't both be a Monist and believe in a Mind-Body duality. That's by definition. And I also know you can't justify something along the lines of "consciousness" from observation of people's behaviour and brains - if we do its because we're bringing the concept along with us as part of a Dualist perspective, or as part of latent Dualism. The only way you can establish a need for the concept is through Mind-Body separation - otherwise you've already got all the stuff you need to explain humans - brains, neurons, behaviours etc. The need to plonk "consciousness" on the top of all that is the latent Dualism I'm talking about in some Physicalists.
The reason a AI would have to believe in consciousness in a Dualist way is the same - because it will not be able to induct such a thing as a "consciousness" from observations. If somehow we managed to cram it in there and give the AI faulty logic, apart from the general unpredictability that implies (an AI with faulty logic?), the AI may realise the same philosophical problem at some point and classify itself or others as without consciousness (some variant of the other minds problem), thus rendering them to the same importance as paperclips or whatever else.
Nobody said otherwise. You keep conflating consciousness with ontologocally fundamental consciousness.
Justify which concept of consciousness? Justify how? We believe others are conscious because we are wired up to, via mirror neurons and so on. But that's a kind of compulsion. The belief can be justified in a number of ways. A physicalist can argue that a normally functioning brain will have a normal consciousness, because that us all that is needed, there is no nonphysical element to go missing. Dualism is no help at all to the other minds problem, because it posits an indetectable, nonphysical element that could go missing, leaving a zombie.
You are conflating consciousness as a posit needed to explain something else with consciousness as a phenomenon to explain. Whatever I believe, it seems to me that I am conscious, and that needs explaining.
Because observations can't give even probablistic support? Because the physicalist argument doesn't work? Because it wouldn't have a concept of consciousness? Because it isn't conscious itself?
About what? If it judged you to be conscious would it be making a mistake?
I'm saying that the only sound logic justifying belief in consciousness arises out of Dualism. (please note I'm not trying to convince you to be Monist or Dualist). Or to put it another way, Physicalism offers no justifcation for belief consciousness of either type.
If consciousness is a thing we should be able to forget about it, and then rederive the concept, right? So in that spirit, if you're Monist, ask yourself what was the point where you discovered or learnt about consciousness. What moment did you think, that thing there, let's call it a consciousness. You didn't look into a brain and find an organ or electrical pattern and then later decide to give it a name right? If you're like 99.99% of people, you learnt about it much more philosophically. Yet, if you're a Physicalist, your belief in objects is derived from empirical data about matter. You observe the matter, and identify objects and processes through that observation. Study of the brain doesn't yield that unless you bring the concept of consciousness along with you beforehand, so consciousness for Physicalists is really in the same class as other hidden objects which we also can imagine and can't disprove (I'm looking at you Loki/Zeus etc).
I'll leave it at that and let you get the last word in, because even if you're willing to consider this, I appear to be offending some other people who are becoming kinda aggressive. Thanks for discussion.
And I keep saying that being conscious means being aware that you are conscious, and that is empirical evidenceof consciousness.
I can abandon and then recover the concept of consciousness, because there is stuff of which I am aware, but which other people are not aware, stuff that is private to me, and "consciousness" is the accustomed label for the awareness, and "subjective" is the accustomed label for the privacy.
Metaphysical beliefs, physicalism and so on, are not relevant. What is relevant is where you are willing to look for evidence. Most physicalists are willing to accept the introspective evidence for consciousness. You seem to think that the concept of consciousness cannot be recovered on the basis of evidence, and what you seem to mean by that is that you cannot detect someone else's consciousness. You have implicitly decided that your own introspective, subjective first person evidence doesn't count. That's an unusual attitude, which is why you have ended up with a minority opinion.
I don't have external OR introspective evidence of Loki and Zeus. The fact that you consider consciousness to fall into the same category as Zeus is another indication of your disregard of introspection.
I want to reply so bad it hurts, but I'll resist thanks for the convo.
Yeah, well, replying could lead to updating or something crazy like that.
sigh I agree with that sentiment. However, conversations where parties become entrenched should probably be called off. Do you really feel this could end in opinions being changed? I perceive your tone as slightly dismissive - am I wong to think this might indicate non-willingness to move at all on the issue?
I don't mean to imply anything personal. I still feel you're overlooking an important point about the ideas you're referring to having fundamentally dualist foundations when you take a proper look at their epistemology. You refer to introspection but introspection (aside from in casual lay usage) is properly a Dualist concept - it implies mind-body separation, and it is not in any way empirical knowledge. Even more prominently the use of the word "subjective" is almost the definition of bringing Dualism into the discussion, because the subject in subjective comes directly from Descartes separation of mind and body.
If someone wishes to be Monist, wouldn't they start by not assuming Dualist concepts? They'd start with the reliable empirical evidence (neuroscience etc) and approach thought interaction as a interactions between internal brain states. They wouldn't conceptualise those interactions using Dualist terminology like "consciousness", at least in any discussion where there was precise science or important issues to be considered.
I thought this was a well explained and epistemologically straightforward part of my post. The general reaction has to me appeared to have been immediate rejection without questions or attempts at clarifications. Actually I'm disappointed that the part people mostly want to reject so utterly is the only part getting much attention. Every sense I get is that the people that have replied can't even consider the possibility that there is problem with the way consciousness is thought about in FAI discussions. That worries me, but I can't see the possiblity of movement at this point, so I'm not terribly enthusiastic about continuing this repeating over and over of the various positions. I'm happy to continue if you find the concept interesting, and I guess I'm at least getting comments, but if you feel that you've already made up your mind let's not waste any more time.
I imagine having an AGI that is safe but not friendly would involve seriously limiting its abilities. However, a friendly AGI would be able to compete just fine with an unsafe AGI. Whether or not the FAI wins the fight is vastly more important than what it does during the fight, so its goal system will do little to change how it fights.
Interesting comment, cheers. What about a situation where the first action of a hostile AI would be to irradiate the Earth, killing all life and removing the FAI's reason to continue struggling, even though it itself was tough enough to survive? It seems options like this would tip the balance very much in UFAIs favour?
The FAI would precommit to struggle fruitlessly in an irradiated world, in order to prevent the UFAI from having a reason to irradiate the world.
If by struggle fruitlessly you mean do whatever it can to hurt the UFAI, then you have a point.
It would recreate humanity. I imagine that if the UFAI does enough damage that the FAI can't remember what humanity was, it would be enough damage to lobotomize any AI and defeat it easily.
I guess that depends on the level of capability that an AGI had at that point. It supposes the level of knowledge had was enough for recreation, which may be much higher than what's required for mere protection. It's hard to perceive as that level of capability is way beyond our own. I'll certainly give it some thought.
It also depends on how you define "human". I'd hope the FAI is willing to upload us instead of wasting vast amounts of resources just so we're instantiated in a physical universe instead of a virtual one.
It's worth noting that the AI only has to be advanced enough to store the information. Once it's beaten the UFAI, it has plenty of time to build up the resources and intelligence necessary to rebuild humanity.
I personally imagine that AGI will arrive well before its possible to store a full down-to-the-subatomic-level map of a person in a space that's any smaller than the person. "Just store the humans and bring them back" implies such a massive storage requirement that its basically not much different from making a full copy of them anyway, so I wonder if such a massive storage device wouldn't be equally vulnerable to attack.
I'm also keen to see us continue as a biological species, even if we also run simulated brains or people in parallel. Ideally I can see us doing both if we can establish a FAI. The best bet I can see so far is to make sure a FAI arrives first :-)
You don't need accurate down to the subatomic level. You just need a human. The same human would be nice, since it means the FAI managed to keep all of those people from dying, but unless it's programmed to only value currently alive people, that's not a big deal.
Also, you make it sound like you're saying we won't develop that storage capability until well after we develop the AGI. It's the AGI that will be developing technology. What we can do just before we make it is not a good indicator of what it can do.
Because neural pathways and other structures of the brain are pretty small, I think you'd need an extremely high resolution. However, I guess what you're saying is that a breeding population would be enough to at least keep the species going, so I acknowledge that. Still, I'm hoping we can make something that does something in addition to that.
Your second point depends on how small the AGI can make reliable storage tech I guess.
In the perhaps this whole point is moot because its unlikely an intelligence explosion will take long enough for there to be time for other researchers to construct an alternative AGI.
Their children will be fine. You don't even need a breeding population. You just need to know how to make an egg, a sperm, and an artificial uterus.
It might encounter another AGI as it spreads, although I don't think this point will matter much in the ensuing war (or treaty, if they decide on that).
Thanks for the link. I reposted the idea currently on my mind hoping to get some criticism.
But more importantly, what features would you be looking for in a more serious version of that game?
I think I'd like the comments to be broadly organised and developed as the common themes and main arguments emerge. Apart from that a little more detail. I don't think it has to go into much implementation specifics, because that's a separate issue and requires a more highly developed set of math/CS skills. But I think we can make use of a broader set of smart brains by having this kind of discussion.
I would have liked it if the first part of the post (up to "Other mind problems") were posted separately. It has a clear independent point and could/should be discussed independently, don't you think?
Fair point. Because I'm new here I think I was concerned my content would not be of enough interest as separate posts. Probably this wasn't really the best decision. I might break parts of the article off into more developed pieces if there's enough interest.
I think another important point is how simulations are treated ethically. This is currently completely irrelevant since we only have the one level of reality we are aware of, but once AGIs exist, it will become a completely new field of ethics.
And perhaps most importantly: When people disagree on how these questions should be answered, how do you react? You can't really find a middle ground here since the decision what views to follow itself decides which entities' ethical views should be considered in future deliberations, creating something like a feedback loop.
Yeah that's an important topic we're going to have to think about. I think its our natural inclination to give the same rights to simulated brains as for us meatbags, but there's some really odd perverse outcomes to that to consider too. Basically, virtual people could become tools for real people to exploit our legal and ethical systems - creating virtual populations for voting etc. I've written a little on that half way down this article: http://citizensearth.wordpress.com/2014/08/23/is-placing-consciousness-at-the-heart-of-futurist-ethics-a-terrible-mistake-are-there-alternatives/
I think we'll need to have some sort of split system - some new system of virtual rights in the virtual world for virtual people and meatbag world rights for us meatbags, basically just to account for the profound physical differences between the two worlds. That we can preserve the species and still have an interesting virtual world. Waaay easier said than done though. This is probably going to be one of the trickiest problems since someone said "so, this democracy thing, how's it going to work exactly?"