Eliezer Yudkowsky and Massimo Pigliucci Debate the Singularity
Research Fellow Eliezer Yudkowsky of the Singularity Institute for Artificial Intelligence and Philosophy Professor Massimo Pigliucci of the City University of New York debate the meaning of intelligence and the possible limits of AI.
The Singularity and the outer limits of physical possibility (08:38)
Do human brains run software? (09:58)
Consciousness, intelligence, and computation (03:14)
What could minds be made of? (13:08)
Is mind-uploading a dualist dream? (19:18)
Would the Singularity be a Vonnegut-style catastrophe? (10:56)
From Science Saturday at BloggingHeads, which features a lot of familiar faces, including Mike Treder, Razib Khan, Jamais Cascio, etc.
Eliezer (in response to Pigliucci invoking the idea of "limits"): "Just because there are physical limits doesn't mean that those physical limits are low. There is a limit to the power output of a supernova, but you wouldn't want to jump into one wearing only a flame-retardant jumpsuit."
At 10:00, Pigliucci appears to claim that the distinction between hardware and software in the brain means nothing, and that the hardware the human brain runs on is irrelevant.
I recently read a review of Pigliucci's recent book at the Chronicle of Higher Ed.
"Intelligence is producing correct answers and is good as gold however you get it." -- EY
Pigliucci first jumped into the Singularity argument most aggressively with this article.
"It's extremely difficult to see how you can carry the argument that understanding and predicting reality is like sugar... that you could get the same answers, written on a different kind of paper, and they would not be useful. Because that's the same kind of answer you're getting in the photosynthesis example."
Pigliucci seems to acknowledge that other forms of consciousness besides human should be possible, but is apparently skeptical of the feasibility of AI..?
Around 20:00, Pigliucci pulls the ridiculous "self-awareness" card, essentially saying it is so special it can never be implemented in a computer. Excuse me, but even a camera filming itself in a mirror is "self-aware" -- it holds an internal representation of its own state. "Self-awareness" is trivial. If what he actually means is "phenomenological consciousness" (why not say so?) then that's more tricky, but to assert that other aspects of thinking can be simulated but this cannot is simply an arbitrary statement borne of apparent deep reverence for consciousness. It may be that consciousness requires certain complex structural elements to emerge (which may even require chemical simulations in a computer, though I doubt it), but our reverence and sentimentality for it have no sway in the matter, in the same way that our reverence and sentimentality for life didn't end up proving that life is powered by an irreproducible vital force.
At 22:00, Pigliucci says that a special substrate may be necessary for self-awareness. At this point, I like to invoke a comment from Eliezer circa 1998 (may no longer represent his current opinions):
"Can we really program human-equivalent AIs?" Yes. The objections fail to consider this: We can cheat. First and foremost, seed AIs don't have to be human-equivalent. An acorn is not a tree. Second, we're allowed to steal code from DNA, observe developing brains... even build AIs out of human neurons if there's a fundamental Penrosian gap. Third, if unmodified humans don't rise to the challenge, that doesn't rule out transhumans or neurosurgically produced specialists.
Coding an AI isn't an ideological argument. If a method works, we'll use it.
At 22:45, Eliezer points out that Pigliucci's argument implies that zombies are possible -- accurate simulations that aren't conscious. Pugliucci fervently disagrees, but as far as I can tell Eliezer is correct.
Why do people think we can simulate every physical object in the known universe but miraculously not the human brain? Isn't it obvious how this argument is a repeat of vitalist-materialist arguments from a hundred years ago?
The act of jumping is encoded in biology, but no one claims we can't build a jumping robot. Why do people see intelligence as qualitatively different than jumping, when the same simple process evolved them both?
2010 H+ Summit @ Harvard
Remember, the 2010 H+ Summit is coming up on June 12-13... here is a blurb.
The 2010 H+ Summit: Rise of the Citizen Scientist (hplussummit.com) is an important 2-day conference that imagines the role of technology in developing the future, with consideration of various emerging technologies and transhumanist ideas.
This innovative summit, to be hosted on June 12-13th, 2010 at Harvard University Science Center by the Harvard College Future Society, with assistance from Humanity+, will feature over 60 incredible speakers, including futurist Ray Kurzweil, inventor Stephen Wolfram, and scientist Aubrey De Grey among many others.
Topics considered will include Human Enhancement, Artificial (General) Intelligence, Longevity, Whole Brain Emulation ("Mind Uploading"), Technology and Democracy, Bioethics, Science Fiction and Science, and Neuroscience among many others.
To learn more and register, visit hplussummit.com. Registration fees are structured to reward early adopters.
Aubrey de Grey at Singularity Summit 2009: The Singularity and the Methuselarity: Similarities and Differences
Aubrey de Grey at Singularity Summit 2009 -- The Singularity and the Methuselarity: Similarities and Differences from Singularity Institute on Vimeo.
Phil Bowermaster Responds to Annalee Newitz: “Five Arguments Against Four Arguments Against Immortaility”
Phil Bowermaster responds here. Me, I can appreciate the io9 post as a masterpiece of generalization from fictional evidence; including images, I count eleven specific appeals to fictional evidence. This appears to be an early form of co-processing, where content from an external device (in this case, poor television shows) heavily intertwines itself with the thinking processes of the writer, to the point where reality cannot be distinguished from fiction.
Apply for the 2010 SIAI Visiting Fellows Program
Now is your last chance to apply for a Summer 2010 Visiting Fellowship at the Singularity Institute. For a concise summary of what SIAI is about, read this new short introduction.
Cryonics Will Scare Your Head Off
Annalee Newitz apparently thinks cryonics is creepy.
Her favorite comment on the photo collection of dewars (scary!) was this articulate one:
Profound.
Question: is cryonics any more "creepy" than what we already do with bodies where metabolism has ceased?
Human beings are largely unaware about the gruesome nature of “deathâ€.
Humans also shy away from the mutilation that occurs during hospital surgery.
Hollywood films portray cryonics in a glamorous high-tech manner that makes it appear that one’s body can easily be placed into a capsule and frozen for future revival.
Reality is that cryopreservation involves complex surgery whereby tubes are inserted into major arteries and veins in order to deliver special anti-freeze solutions into the brain. The purpose is to reduce or eliminate freezing damage and other types of damage to brain cells. The process involves introducing stabilizing drugs and a special solution in the field and a major procedure in an operating room.
There’s nothing pretty about human cryopreservation, but as you’ll read, the alternatives are truly ghastly—and every alternative involves the head eventually separating from the body.
We deceive ourselves
When I worked as a licensed embalmer, I was quite talented at taking horrific human remains and making them look good temporarily. In order to do this, a tremendous amount of mutilation was done to each corpse.
First step is to wire or sew their mouths shut. Incisions are made in the neck, groin and other areas to access arteries to insert tubes that were used to force formaldehyde in. Veins are accessed (raised) to push blood out.
While formaldehyde delivered through blood vessels preserves tissues of the body, it does little to keep cavities (such as the stomach, bowels, lungs and cranium) from putrefying. To keep the body from decomposing before burial, we used a device that resembles a thick hollow sword to repeatedly penetrate the body cavities to vacuum out as much of the liquid contents as possible. We would then reverse the process by pouring formaldehyde directly into the thoracic and abdominal cavities and sometimes the brain. Sometimes the same sword (trocar) used to evacuate the bowels was shoved up the nose through the sinuses to suck out cerebral-spinal fluid in the cranium.
When I learned how to do this in mortuary school, I thought how undignified the entire process is. Without embalming, however, the outcome is even worse.
You know what's creepier than cryonics dewars? That the editor-in-chief of an ostensibly progressive, futurist blog could be so explicitly anti-transhumanist, anti-Singularity, and anti-life extension.
Consider the other side of the story before you condemn cryonics along with Ms. Newitz.
What Does a Buckyball Undergoing Unimolecular Disassociation by Use of Extremely High Levels of Vibrational Excitation Look Like?
How about a C60/C240 collision at 300eV?
H/t Machine Phase.
Jürgen Schmidhuber at Singularity Summit 2009 — Compression Progress: The Algorithmic Principle Behind Curiosity and Creativity
Jürgen Schmidhuber at Singularity Summit 2009 - Compression Progress: The Algorithmic Principle Behind Curiosity and Creativity from Singularity Institute on Vimeo.
Eyes Flashing, Robot Presides at Japanese Wedding
Speaking of technology intruding into the traditional role of religious authorities...
Coverage and a gallery can be found at AP.
A Christian Perspective on the Singularity Movement
This was published late last year at Metanexus by the founder of the foundation, William Grassie: "Millennialism at the Singularity: Reflections on Metaphors, Meanings, and the Limits of Exponential Logic". Here's a quote to pique interest:
This is a very technical discussion in computer science, but the short of it is that many problems simply don’t compute. There are also other theoretical and practical limits to computation. These are called intractable problems because they “require hopelessly large amounts of time even for relatively small inputs…†Computer encryption depends on this second fact. It may be that the genome, in dynamic relationship with proteins and its environment, is in some sense “encrypted.†It may be that the mind-brain is similarly “encrypted.†In which case, we will never be able to fully understand, let alone reliably control life and mind no matter how exponentially our scientific knowledge grows nor how fast technological know-how accelerates
Here's another quote:
Of course, anytime we talk about the future, our hopes or our fears, we are in the realm of religions.
...
Nowhere is this religious dimension of the Singularity movement more readily apparent than in their uncritical enthusiasm for life extension research, as if this was an obvious good.
Wanting to live longer than 90 years = religion.
When theists call the Singularity movement "religious", they are essentially saying, "Oh no, this scientifically-informed philosophy is intruding on our traditional turf!"
For the tl;dr version, see this quote from Sister Miriam Godwinson:
"Men in their arrogance claim to understand the nature of creation, and devise elaborate theories to describe its behavior. But always they discover in the end that God was quite a bit more clever than they thought."
Sister Miriam Godwinson, "We must Dissent"
Dangers of Molecular Nanotechnology, Again
Over at IEET, Jamais Cascio and Mike Treder essentially argue that the future will be slow/boring, or rather, seem slow and boring because people will get used to advances as quickly as they occur. I heartily disagree. There are at least three probable events which could make the future seem traumatic, broken, out-of-control, and not slow by anyone's standards. These three events include 1) a Third World War or atmospheric EMP detonation event, 2) an MNT revolution with accompanying arms races, and 3) superintelligence. In response to Jamais' post, I commented:
I disagree. I don't think that Jamais understands how abrupt an MNT revolution could be once the first nanofactory is built, or how abrupt a hard takeoff could be once a human-equivalent artificial intelligence is created.
Read Nanosystems, then "Design of a Primitive Nanofactory", and look where nanotechnology is today.
For AI, you can do simple math that shows once an AI can earn enough money to pay for its own upkeep and then some, it would quickly gain the ability to take over most of the world economy.
Have Giulio or Jamais read "Design of a Primitive Nanofactory" or Nanosystems?
Knowledge of where we are today in nanotechnology, plus Nanosystems, plus "Design of a Primitive Nanofactory", equals scary.
Where we are today: basic molecular assembly lines
The most important breakthrough: a reprogrammable universal assembler
Shortly thereafter: a basic nanofactory
Shortly thereafter: every nation with nanofactory technology magnifies its manufacturing potential by a factor of hundreds or more.
Chris Phoenix gets it. Jurgen Altmann gets it. Mark Gubrud gets it. Thomas Vandermolen gets it. Eric Drexler seems to have gotten it a long time ago. Michio Kaku, Annalee Newitz, and many others have called molecular nanotechnology "the next Industrial Revolution".
When will others get it? Here's a quote from the CRN page on the dangers of molecular nanotechnology:
Molecular manufacturing raises the possibility of horrifically effective weapons. As an example, the smallest insect is about 200 microns; this creates a plausible size estimate for a nanotech-built antipersonnel weapon capable of seeking and injecting toxin into unprotected humans. The human lethal dose of botulism toxin is about 100 nanograms, or about 1/100 the volume of the weapon. As many as 50 billion toxin-carrying devices—theoretically enough to kill every human on earth—could be packed into a single suitcase. Guns of all sizes would be far more powerful, and their bullets could be self-guided. Aerospace hardware would be far lighter and higher performance; built with minimal or no metal, it would be much harder to spot on radar. Embedded computers would allow remote activation of any weapon, and more compact power handling would allow greatly improved robotics. These ideas barely scratch the surface of what's possible.
Will weapons like these in the hands of every backwater terrorist and militia lead to a future that is "slow" or "boring? It could lead to a future where numerous major cities become essentially uninhabitable.
Here's a potentially illuminating quote:
“Revolutions are cruel precisely because they move too fast for those whom they strike.â€
Jacob Bronowski
John Horgan Attacks the “Artificial Brain” Projects
John Horgan, the eminent science journalist who previously called me a cultist, is back on track with a guest blogging article at ScientificAmerican.com titled, "Artificial brains are imminent...not!" And hey, guess what -- I totally agree with him. (Especially as far as the "cat brain" is concerned.) If AI comes about within the next two decades, I wager that it will be because we discovered the operating principles of intelligence and instantiated them in a machine, not because we copied a brain.
(Additional note: Markram has claimed that he has simulated a neocortical column with biologically realistic fidelity, but without demonstrating it more thoroughly, there is no way we can know if this claim is true. A commenter, Jordan, pointed out that Horgan misrepresented Markram's attitude.)
Here's a big quote from John's post:
Sejnowski is a very smart guy, whom I've interviewed several times over the years about the mysteries of the brain. But I respectfully—hell, disrespectfully, Terry can take it—disagree with his prediction that artificial brains are imminent. Sejnowski's own article shows how implausible his prediction is. He describes two projects—both software programs running on powerful supercomputers—that represent the state of the art in brain simulation. On the one hand, you have the "cat brain" constructed by IBM researcher Dharmendra Modha; his simulation contains about as many neurons as a cat's brain does organized into roughly the same architecture. On the other hand, you have the Blue Brain Project of Henry Markram, a neuroscientist at the Ecole Polytechnique Fédérale de Lausanne.
Markram's simulation contains neurons and synaptic connections that are much more detailed than those in Modha's program. Markram recently bashed Modha for "mass deception," arguing that Modha's neurons and synapses are so simple that they don't deserve to be called simulations. Modha’s program is "light years away from a cat brain, not even close to an ant's brain in complexity," Markram complained.
Talk about the pot calling the kettle black. Last year Markram stated, "It is not impossible to build a human brain and we can do it in 10 years." If Modha's simulation is "light years" away from reality, so is Markram's. Neither program includes "sensory inputs or motor outputs," Sejnowski points out, and their neural-signaling patterns resemble those of brains sleeping or undergoing an epileptic seizure. In other words, neither Modha nor Markram can mimic even the simplest operations of a healthy, awake, embodied brain.
The simulations of Modha and Markram are about as brain-like as one of those plastic brains that neuroscientists like to keep on their desks. The plastic brain has all the parts that a real brain does, it's roughly the same color and it has about as many molecules in it. OK, say optimists, the plastic brain doesn't actually perceive, emote, plan or decide, but don't be so critical! Give the researchers time! Another analogy: Current brain simulations resemble the "planes" and "radios" that Melanesian cargo-cult tribes built out of palm fronds, coral and coconut shells after being occupied by Japanese and American troops during World War II. "Brains" that can't think are like "planes" that can't fly.
Yes -- especially with respect to IBM. (As I added in above, I don't know all the facts about Markram's simulation because he hasn't demonstrated it.) Earlier, on May 3rd, when I criticized the IBM "cat brain" nonsense here on this blog, an IBM employee in the comments went ballistic with ad hominem attacks, completely avoiding any discussion of the science. Quick point here about John's claim that the "cat brain" is organized into "roughly the same architecture"... not really. Even the paper doesn't claim this.
There are many important reasons why this issue is worth addressing:
1) Overblown claims today leads to public disillusionment tomorrow. For people who care about the future of real artificial intelligence, like me, there is already enough disillusionment, and I'm not going to stand around while another cycle of overblown AI promises occurs.
2) Simple dishonesty. The people behind the projects know that the "brains" lack the low-level structure that makes even the most rudimentary forms of thinking possible.
3) Scientific support. Even though I'm not a scientist myself, I know that any neuroscientist exposed to the details of these simulations would realize that they are not even vaguely close to biological brains.
4) I believe that the future of AI is in uncovering and implementing the operating principles of intelligence rather than copying the brain, like how the Wright Brothers uncovered and implemented the operating principles of flight rather than copying a pigeon.
5) If we can't question this obvious farce now, then what will we do years down the road, when more subtle AI deceptions are foisted upon science-ignorant journalists and the public?
In his article, John Horgan even brings a smile to my face when he shows that he is not an ideologue:
Go back a decade or two—or five or six—and you will find artificial intelligence pioneers like Marvin Minsky and Herbert Simon proclaiming, because of exciting advances in brain and computer science: Artificial brains are coming! They're going to save us! Or destroy us! Someday, these prophecies may come true, but there is no reason to believe them now.
Someday, these prophecies might come true... did you hear that? Sweet validation. This statement is somewhat consistent with my research that shows the number of academics who have published papers or books arguing that AI is impossible in principle appears to be roughly zero. (What Computers Can't Do is the only one that comes to mind, but I haven't read it.) Mr. Horgan is not an academic, but he is a smart guy, so I would expect him to acknowledge the possibility of AI in principle, while not being duped by claims that neural networks that display oscillations represent a major advance in whole brain emulation.
The assumption that each neuron is an on-off switch that performs exactly 200 operations per second is wrong on so many levels. The answer is more subtle -- there could be single neurons that require tens of thousands of operations per second to simulate, or perhaps entire neural aggregations that can be modeled with just one operation per second. The truth will be complex, nothing like the superficial calculation, "100 billion neurons times 200 spikes per second equals 10^17 ops/sec of computing power equals the human brain". In the comments, Jordan pointed out that Markram uses an entire CPU for each neuron, so that may be on the right track, but we still can't confirm because nothing has been released.
Note that my skepticism of these brain-building projects is inconsistent with the popular misconception of Singularitarians as insufferable nerds awaiting a Techno Rapture. So is my skepticism of claims of accelerating technological progress.