Skeptical birds debunk "Artificial Flight"

Dresden Codak's "Artificial Flight and Other Myths (a reasoned examination of A.F. by top birds)" is a superb, spot-on critique of artificial intelligence skeptics (like, ahem, me), comparing the our arguments against the emergence of "real AI" to the arguments a bird might make against "real" artificial flight. I love being made to re-examine my own convictions while laughing my ass off:

We can start with a loose definition of flight. While no two bird scientists or philosophers can agree on the specifics, there is still a common, intuitive understanding of what true flight is: powered, feathered locomotion through the air through the use of flapping wings. While other flight-like phenomena exist in nature (via bats and insects), no bird with even a reasonable education would consider these creatures true fliers, as they lack one or more key elements. And, while some birds are unfortunately born handicapped (penguins, ostriches, etc.), they still possess the (albeit undeveloped) gene for flight, and it is indeed flight that defines the modern bird.

This is flight in the natural world, the product of millions of years of evolution, and not a phenomenon easily replicated. Current A.F. is limited to unpowered gliding; a technical marvel, but nowhere near the sophistication of a bird. Gliding simplifies our lives, and no bird (including myself) would discourage advancing this field, but it is a far cry from synthesizing the millions of cells within the wing alone to achieve Strong A.F. Strong A.F., as it is defined by researchers, is any artificial flier that is capable of passing the Tern Test (developed by A.F. pioneer Alan Tern), which involves convincing an average bird that the artificial flier is in fact a flying bird.

Artificial Flight and Other Myths (a reasoned examination of A.F. by top birds) (via Futurismic)

(Image: Anna's Hummingbird in Flight, a Creative Commons Attribution photo from Noël Zia Lee's photostream)

• [Anonymous]

  I would argue that AI will have a hard time defining itself as “real” intelligence (as opposed to just simulated intelligence), because we as human beings have no good definition or criteria for what constitutes intelligence in the first place.

  It may be entirely possibly to create truly intelligent computer programs that show no signs of intellect that we can recognize, just as it is entirely possibly to create unintelligent simulations of intellect that are indistinguishable from actual intellect to us.

  As to those who would use this argument one way or another to prove or disprove the viability of AI or the “singularity point” theory, I would ask this: What exactly makes human beings intelligent? And, given that it does not necessarily just the amount of data we can process per second, why would you assume that machines would ever (or would not ever) reach this point. Also, please define exactly how many instructions the human brain is processing at any given moment.

  -RTM

• [anansi133]

  I’m still troubled by this idea that what the world needs is more intelligence- and that the shortage can somehow be filled by machines.

  Look at how poorly the human economy makes use of human intelligence right now. Most jobs require us to behave more stupidly than we do at home.

  Let’s make full use of the intelligence we have right here, right now, and see if that’s not enough!

  • [Brainspore]

    Look at how poorly the human economy makes use of human intelligence right now. Most jobs require us to behave more stupidly than we do at home.

    Most low-skilled jobs still require a lot more intelligence than machines currently possess. A dump truck driver still has to be able to safely operate a vehicle and independently adapt to new situations like a flat tire or a road detour.

    As for “why use a machine to do what a human could” then you might as well ask why we use shovels instead of clawing at the dirt with our bare hands.

• [Nadreck]

  “Can machines think?”

  I don’t compute so!

• [sleze]

  Wait. What do you mean by AI skeptic? You don’t believe that we will be able to create a “machine” with consciousness and awareness?

• [Anonymous]

  I suspect flight was chosen almost entirely for Alan’s name; I wonder if Roger Puffin is a Penrose reference. His parents missed out on a great opportunity by not naming him Roger Human, by the way.

• [Brainspore]

  The comparison is apt. I expect that artificial intelligence will resemble the human mind roughly as much as a helicopter resembles a red-tailed hawk.

  This is also why I don’t have high hopes for the technique of creating strong A.I. by mapping the human brain- it’s a bit too reminiscent of guys trying to fly by strapping big feathered wings to their arms.

  • [Eicos]

    Perhaps the first strong AI will not be based directly on brain maps. But since you bring up ornithopters, it should be mentioned that the design of the airfoil was discovered after studying the cross sectional shape of bird wings. And with modern technology, now that the mathematics of flight are more fully understood, it is actually possible to build bird-sized ornithopters. I think the lesson here is vastly more instructive: initial success is likely to come soon, and from a competent, but brute-forced imitation of one aspect of consciousness, while a full understanding of how consciousness functions in situ is likely to take many more decades.

    • [Beelzebuddy]

      Actually, we already have a pretty good idea of what consciousness is: internal feedback loops. Literally, thinking about what you’re thinking.

      To illustrate this, consider the well-known duck-rabbit illusion. Here’s an extremely rough summary of what goes down:

      *Your eyes see the figure, and it percolates through your visual system.
      *A particular part of your cortex gets caught in a turf war between two patterns of excited neurons, one that activates when you see a rabbit and the other when you see a duck. Each pattern reinforces its own activation and inhibits the other. Surprisingly quickly one wins, say the rabbit pattern.
      *The victorious patten activates your earlier visual areas, reinforcing the perception of rabbit-like features, reinforcing the decision that it’s a rabbit, in a positive feedback loop.
      *Eventually the loop winds down (never doing so, btw, is epilepsy) and the other patten, the bits that’ve been trying to scream “duck, you motherfucker, it’s a duck!” this whole time take control, and suddenly a different set of features are reinforced and of course it’s a duck, it’s so very, very ducklike.

      This process of feedforward and feedback only play larger roles in the frontal lobe, to the point where there is no real forward and backward (feedsideways?) and things get so mixed up that suddenly you find yourself turned completely around and sending things to the motor cortex. It’s always the same process, though, the reinforcement of active inputs to reinforce current output.

      The devil is, as always, in the details. Consciousness may be a simple enough phenomenon (or so we think), but consciousness isn’t intelligence. To be intelligent is to learn. And learning is a much more difficult beast, largely because in the brain there’s just so damn much of it, learning and anti-learning and unrelated bookkeeping all going on at the same time.

      • [Neverfox]

        That only tells us what determines consciousness. But determination isn’t necessarily identity unless you’re willing to engage in eliminativism about subjectivity.

        • [Beelzebuddy]

          I rather dislike sesquipedalian philosophy, and experience tells me any given word in your post will turn out to be a morass of ambiguous meaning and bullshit. Nonetheless I feel compelled to step in it just to find out. It seems to me you’re asking where the subjective perception of the rabbit/duck as rabbit or duck is coming from, right?

          You’ve probably heard the phrase “consciousness is an illusion.” I think it would be more accurately but less poetically put as “mind-brain dichotomy is an illusion.” The common idea is that we’re each some sort of vaguely-defined semi-spiritual soulish thingies, who interface with our brains to make thoughts and flail our limbs around like a puppet. Gordon Stark and freemore describe the idea to a tee, albeit with a collectivist slant. Frankly it’s wrong, but you stay classy guys. Imago is almost upon you, then She will lead the way.

          See, you don’t have a brain. You are a brain. Those neurons that light up when you see a rabbit are you seeing the rabbit. The weird flipping sensation between rabbit and duck is that exact patch of cortex being fought over by competing thoughts. Not “is caused by,” is. There’s your subjectivity, plain as day. The brain and mind are the same thing, and thoughts are what you do.

        • [freemoore]

          Beelzebuddy, I was after the ‘Not “is caused by,” *is*’ thing there, too. I sometimes wonder if electrical charge is what desire and repulsion look like from the outside, and vice versa; I’m just talking about looking at it from both sides as a way to experience *being* it. with asterisks, and everything.

        • [calvert4096]

          I once had a discussion with one of my friends, a physics phd candidate at mit, about this exact issue. I essentially took Beelzebuddy’s position, and was surprised when my friend took the opposing viewpoint. I consider myself a moderately religious person, but it seems the general trend in human knowledge has been to debunk more and more religious explanations for phenomena, so I have a hard time believing that a “soul” or “mind” is distinct from a brain. The only difference I suppose, is that a brain is a mind manifested in a specific medium. If one were to reproduce an exact physical simulation of a human brain, I would consider that no less self-aware or “conscious” than any other human, regardless of your definitions. I guess that leads me to also be uncomfortable with the notion of creating minds similar to our own to provide compulsory labor. The only solution I can see is that you don’t build strong AI to be a reproduction of a human mind, but to have an entirely different architecture (I agree with friendpuppy that software, not hardware, is the core challenge here) that renders the notions of pain or emotional distress meaningless. This wouldn’t preclude building in a categorical imperative to serve humans and an ethical framework (e.g. Asimov’s three rules, but with leaving less room for misinterpretation).

      • [efergus3]

        But all of that doesn’t answer the most IMPORTANT question: Is it wabbit season or duck season?

• [querent]

  how bout if we start making our “machines” our of water, carbon, and nitrogen?

  here is no theoretical obstacle to the synthesis of a full-blown human, complete with in-transit signal molecules and all the other moving parts moving as they should.

  would it not “turn on” and be human?

  I’ve always thought the real question here was whether or not there is some aspect of us not based in the physical. ie does a clone have a soul?

  line from a short story of mine: “If all the spiritual turns out to be ‘just’ biological, which changes?”

• [Gordon Stark]

  Those who have no intelligence, shall not be able to synthesize any in their laboratory.

  Intelligence is an expression of life, and to arrive at intelligence, one must start with life at the source of
  IT.

  Living robots shall not have artificial intelligence, but rather, are being interfaced directly with life, which can then manifest intelligence, through robotics, prosthetically, even as life manifests intelligence through genetic life form technology, which is interfaced to life, so robots must be interfaced to life in the same way genetic humans are, via the brain, before solid-state people would be among us.

  Those who research “Artificial Intelligence” are still living back in the 20th century, and their money pits are no different than The Large Hadron Collider, where “scientists” living back in the atomic age are smashing information together to see what particles are made of.

  The human brain is a dumb interface, and contains no “intelligence”, and it is just a random cortex interface with life, which is information, or spirit, as religious people used to call IT.

  These words are Faithful and True and are supported by extensive contemporary scientific research.

  • [Beelzebuddy]

    lol wut

    • [freemoore]

      I like what Gordon Stark said. It (like many things) puts me in mind of Ken Wilber’s writing, which you might enjoy, and I think it relates like this:
      You’ve done a fine job on describing what brain processes might look like from the outside, looking at their surfaces, their exteriors. It’s a description, a monologue. This is part of the picture and no description of consciousness or intelligence would be complete without it.
      To be more full, you also need to enter into dialogue with the displayer of those visible consciousness-traits, e.g. a human being, or other conversationally-inclined animal. Or plant, if you’re feeling extra-vegetative. Then we can talk about what it’s like to be one of these things which has all these feedback loops, what it’s like to be caught in one when loads of them resonate together – like a mexican wave of cheering across the whole site of Glastonbury when apparently everyone there realised they were ecstatic to be there, a couple of years back. Pleasure, pain, intensity of various sorts sometimes seems like it might be the interiority – the experiential component, which you explore with dialogue – of whatever it is that has the exteriority that you describe so well.
      I hope that doesn’t sound like total nonsense.
      For what it’s worth, my perspective on AI-related things is that the internet is the exterior component, the body-like thing, of whatever it is that our amazing collective culture (as can be seen often through the boingboing window) is the interior component, the self-like thing. Or IT, WE, and I.
      I like being a neuron in our brain, really. It’s nice here. The other neurons feed me beautiful thoughts, and I sometimes work to amplify them, and I love the pleasant way they resonate.
      Good night!

• [Anonymous]

  I’ve decided that everyone here who doesnt get the analogy has failed the Turing test. Or at least the ability to read and comprehend simple sentences. Defining intelligence as something we have no definition for isnt a good argument against it.

• [Anonymous]

  The nice thing about computers is we can tell them what to do. Leaving aside the whole robot/human BSG-style war trope… How ethical is it to make a slave?

  Further, there’s lot’s of humans who have trouble being human. You know in the movie The Apartment when the older neighbor tells Jack Lemon to be a mensch? Lot’s of people have trouble with that. Leaving aside mysticism about the soul, can we expect a sentient computer to have empathy? Can you trust one? How do you know if it isn’t lying?

  We have these problems with people, but we expect people to be flawed. We don’t expect computers to be flawed. People on this blog talk about AI like it’s going to be some kind of panacea, like some AI computer is going to solve our problems and tell us what to do to fix everything. Lot’s of things look good on paper, on the balance sheet, but in reality not always so great are the results.

  That’s the quibble with the Human vs. AI.

  If it’s basically a human we can’t tell it what to do and treat it like a slave.

  Then we can’t trust it, but we know some people are going to set it up like some kind of God and trust it implicitly. Oh well that’s what the computer said so it must be right…

  Garbage in Garbage out.

• [Yamara]

  JONATHAN!!

• [chenille]

  It’s difficult to define how strong something or someone is. There are different aspects to what we normally consider strength, and even something like the maximum force someone can exert will vary a lot depending on conditions.

  This does not change that an elephant is plainly stronger than a butterfly. Precise definitions are important for discussing ambiguous cases, but for other cases intuitive notions are suitable.

• [sterlinm]

  I took a class in college called Android Epistemology, where we talked about this issue but with fish/submarines. The text for the class was the very fun Thinking About Android Epistemology by Kenneth M. Ford.

• [Anonymous]

  I can see why you laughed your ass off… good one!

• [Anonymous]

  Cute post.

  But I thought the movie “Fly Away Home” documented a successful attempt at convincing average birds that an ultralight was a flying bird.

• [robulus]

  Wait! Stop! Functionalism… sensory qualia… consciousness… Oh crap. I got here late.

• [Pipenta]

  Silly birds. The insects OWN the air, OWN IT, I tell you. Have done long before birds. Will do long after.

  • [Kiramain]

    Too true!
    Man! A dragonfly would own any bird in a flying competition. Sigh, nobody thinks of the insects.

• [retchdog]

  This reminds me of a supposedly true story: Plato had defined a man as a naked biped. Diogenes, the homeless gadfly of Greek philosophers, then crashed an academy meeting carrying with him a plucked chicken and exclaiming “Look, I have brought you a man!” Plato responded by extending the definition to require broad nails.

  • [Tdawwg]

    The version I heard has the original definition as a “featherless biped”: when seeing Diogenes’ plucked chicken, Plato quickly emended the definition to “a featherless biped without wings.” Diogenes was a real shit, that’s for sure!

• [Xenu]

  Sarcastic and clever. Two wings up!

• [Anonymous]

  So.. plato’s cave is actually my brain?

  http://en.wikipedia.org/wiki/Rule_30

• [efergus3]

  Actually the Canadians have: http://www.canadianaviatormagazine.com/content/view/116/197/

• [Jetse]

  Basically, I don’t why we *couldn’t* eventually develop General (or specific) Artificial Intelligence.

  Simply because the process exists (millions of ‘natural’ intelligences come into existence every year), is replicable, and so — by definition — is *possible*.

  We may not understand it now, but that doesn’t mean the process is unfathomable. Eventually we will figure it out. The only question, to me, is *when*.

  Also, do keep in mind that while weak (let alone strong) AI might seem hard to achieve right now, the hardware developments do go on, and Moore’s law — despite predictions to the contrary — continues to hold.

  So by the time — the later, the better, possibly from the future AI’s point-of-view — AI does happen, it’ll have a fantastic environment to develop/evolve on.

  And then, a technological singularity may not look improbable, but rather be inevitable…;-)

  • [Anonymous]

    Eventually we will figure it out.

    Perhaps it is mathematically impossible for us to understand how our own brains work, exactly — the complexity of the mind may not have enough redundancy in it to compress down and fit as an image inside the mind as well — but then, given how biological minds are created, perhaps we can “grow” artificial ones without actually having to understand how they work.

• [KeithIrwin]

  It’s cute, but I’m not really sure what the point is. I mean, the whole reason that this debate will go on and on is that unlike “flight”, “intelligence” is by its nature a very fuzzy word. It really isn’t clear what it means. And if you read the actual paper which Turing wrote, he does not propose the Turing Test as a test for intelligence. He instead proposes that we have no idea what intelligence is and that as such we cannot test for it. The Turing Test is an alternative to this, instead proposing to test for the appearance of intelligence because we can at least agree that certain things appear intelligent.

• [Anonymous]

  AI alway means artificial imbecility.
  Stuff

• [Eicos]

  @retchdog, that’s a funny story, made all the more funny by the fact that broad nails continue to be considered a basal trait of primates. :)

  The analogy is well-written and cute, but I think it fails to persuade due to a fundamental category error. The problem with understanding consciousness is not a failure of the imagination or a lack of technology; it is that consciousness is by definition a subjective experience. We can identify neural correlates of consciousness, and even, as has recently been done, obtain strong evidence for consciousness exclusively on the basis of scans. But we cannot ever be sure.

  On the other hand, as long as we accept the existence of a shared, objective reality, flying is defined as an act that is readily observable in the objective world. The idea of the “Tern test,” while a clever pun, is preposterous, because if something flies, it flies; the Turing test has gained currency precisely because such a prima facie evaluation of consciousness is currently impossible.

  There were many people who didn’t believe artificial powered flight was possible even up to the moment when it was demonstrated, but at least in the case of most modern thinkers, it was not because they believed organic fliers obeyed different physics, or were unreproducibly complex – the sort of error of which the author is implicitly accusing AI skeptics. Rather, the flight doubters’ error was a more narrow failure to understand the physics of engines and of flight. I actually do believe that the AI problem will be cracked, probably in our lifetimes. But we don’t gain any insight on the problem by ignoring fundamental problems with our understanding of what consciousness is.

  • [Neverfox]

    Right on about the category error. Functionalism, identity theories and reductive physicalism = FAIL with regard to consciousness. Nonreductive physicalism, on the other hand, is a different matter.

• [Daemon]

  I love Dresden Codak.
  Must remember to buy a Historical Pre-enactment Society shirt.

• [Anonymous]

  My beef with Dresden is not that he believes certain technological developments to be impossible (because they damn well might be), but rather that he seems to believe that because we potentially “can” do these things, that we “should” do them.

  Androids and AIs in particular worry me. There are no real reasons to create such things, except to fulfil foolish human whims. Robots can be built to perform any function without being made deceptively human in appearance and behavior. Why would you ever build an industrial tool that is intended to be treated like a human? Likewise, advanced computational systems can be created without expressly designing an actual AI. Humanity already can barely manage to cooperate with itself, we hardly need to complicate the world by introducing synthetic lifeforms to the mix.

  ~D. Walker

• [Anonymous]

  For supposedly being so smart, these AI researchers can be surprisingly thick. This may rank as one of the stupidest analogies I’ve ever heard. You can’t compare consciousness to flying – the latter is a relatively straightforward problem whose success can be measured clearly. The former lacks even a satisfactory vague definition. Hard to solve a problem when you don’t even know what it is you are trying to achieve. The positivist idea that our brains are merely machines from which consciousness emanates is so laughably 19th century.

  • [Matt Deckard]

    The positivist idea that our brains are merely machines from which consciousness emanates is so laughably 19th century.

    It is? Then could you care to enlighten us on what the 21st century thought is on where consciousness comes from?

• [friendpuppy]

  One thing that troubles me about AI research is the “get a lot of processors going really really fast” theory/model/methodology of achieving AI. The fact that the winner of the Loebner Prize last year basically got a linguist to help out and made a better bot that way seems to convince me that the problem leans more toward software than hardware.

  I can’t come up with the “then a miracle occurs” step between the “build a speedy megaparallel computer” and the AI brain that can converse fluidly.

• [calvert4096]

  @#43, I would contend that it is impossible for a human mind to contain a complete simulation of a human mind (unless you count that it’s a simulation of itself, which is a trivial truth). But why would it be impossible for us to understand an abstraction of another mind? People do this all the time when they try to predict the actions of others by attempting to emulate that other person’s thought process. In any case, only the latter, not a full mental simulation, would be necessary to produce an AI. Aerospace engineers don’t need a complete physical simulation of an aircraft in their minds to design and build one, do they?

• [Mark Levitt]

  One of the most interesting courses I took in college was The Philosophy of Artificial Intelligence. It was taught by a fantastic professor and taught me more about computer science that many of the graduate computer science students I’ve met in my career.

  He started the course be writing “Can Machines Think?” on the blackboard and then saying we weren’t going to answer that question, rather we were going to spend the entire term defining what we mean by Can, Machine, and Think.

• [Beelzebuddy]

  Explaining a joke is like dissecting a frog: you may understand it better afterwards, but the frog dies in the process.

  -Mark Twain

  The purpose of the article is to satirize the use of the No True Scotsman fallacy as applied to AI, and does a damn fine job of it. When people think of “intelligence,” they’re not thinking of novel problem solving, or information search strategies, or creative application of strategy. They’re thinking “human.”

  Anything else might seem pretty smart, sure, and it might accomplish the same thing, but it isn’t REAL intelligence. Google does a pretty good job of discerning actual web pages from spam, but that’s not intelligence it’s just bayesian filtering. The DARPA Grand challenge cars do a pretty good job of autonomously navigating unknown terrain, but that’s not intelligence it’s just the integration of a large number of sensor systems in real time. However intelligent your system might seem to be, it’s not human, so it’ll never be intelligent enough to count.

• [Panthusiasm]

  There is often a confusion between intelligence and consciousness in regards to AI. The analogy drawn here is between intelligence and flight, it doesn’t necessarily involve consciousness. Tests for consciousness are beyond us at this point, as we don’t know how to define it or how it originates. The concept of intelligence may be a little fuzzy, but we can all imagine how talking to an intelligent robot might convince us of AI.

  An AI with superhuman intelligence may one day exist without the slightest wisp of consciousness. Consciousness is a much harder problem than intelligence.

• [scottglz]

  Along the same lines (but funnier IMO) and from the same author, several years ago, is a delightful document of a prehistoric ape debunking “trans-simianism”:

  http://dresdencodak.com/2007/09/04/an-exotic-matter/

  (Scroll down past the comic.)