Hi again Matt, The claim that a human brain is a computer is and always has been a scientific hypothesis used to guide the science of artificial intelligence. It's not a law of nature.
That hypothesis may be true. We cannot use 'everything that can be computed can be computed by a Turing machine', a theoretical result in computer science, as any proof that a human brain 'is a computer', where 'computer' means the classical formal definition of abstract symbol manipulation (automata theory). Einstein made an epistemologically equivalent claim about general relativity early last century: That gravitational waves are real. It was a hypothesis until the evidence came in a couple of years back. The hypothesis that X = 'natural brains are computers' is in precisely this state. Any assertion of existing proof is false. It is not proven empirically. Any work done assuming it is proven is theoretical science done on the premise of an assumption that X is empirically proved. It isn't. The reason for this paper is to put form to the real empirical testing that is needed, and to thereby highlight what is missing from the science. There is real testing missing from the science. It is where you compare the 'computed model of brain physics' (symbol manipulation of the formal kind) with what the brain physics does. When and in what way might these two things part company? You do this in a context directed at the origins of intelligence. Theory and empirical work meet as is usual everywhere else in science except AI. Putting the appropriate subset of brain physics on the chips has never been done. Analogue computers do not do that. Digital computers don't even come close! That is the absence of scientific proof I am talking about. It's directed at an engineering goal "creation of an artificial version of natural brains". It's simultaneously directed at empirical proof of hypothesis X. It simultaneously provides the correct context of the empirical work as being in neuroscience and physics, not computer science. Computer science prediction/ formalities are the theory and created a hypothesis. Neuroscience and physics and related engineering provide the empirical proof. This is empirical work that has never been proposed let alone tried. It's been orphaned by a cultural meme (assumption X is proved) trained into all the communities and that had no context for it's revealing until now. Note: I predict that: 1) when the science is done properly it will prove that X is actually true, but provide a real context for expectations of performance and the symptoms of underperformance. 2) it will also prove that under the conditions of use of digital/analogue computers used universally in AI so far, achieving real AGI with digital/analogue computers of the familiar kind is so technologically onerous (huge) nobody in their right mind would ever use it. The model of the brain needed is spectacularly huge. That is, X is true but practically useless in achieving full equivalence. But I cannot claim this to be proved for the exact reason X cannot be claimed true. I don't know. Neither do you. Neither does anyone on this list. Neither does the wider community dedicated to the AI project. If you like, I can help you see this by an attempt to get you to see what 'empirical work' in computer science is, and how it is not empirical science when applied to the neuroscience of artificial (versions of natural ) intelligence. Can I do this? It will definitely help the paper. Colin On Thu., 27 Jun. 2019, 6:19 am Matt Mahoney, <[email protected]> wrote: > A computer is any device which can simulate a universal Turing machine up > to some memory bound. Real computers are finite state machines. But with > sufficient time and memory they can perform any halting computation because > all halting computations use a finite amount of tape. > > My brain is a computer because I can model a universal Turing machine with > the help of pencil and paper to augment my ~100 bits of short term memory. > > Colin can reasonably argue with my definition of a computer. By my > definition, Alexa and the controllers in my TV and microwave are not > computers because you can't load your own programs into them and run them. > Of course expanding the definition of a computer doesn't help his argument. > > I think most of us agree that everything the brain does, like see, hear, > speak, move, pass the Turing test, etc. can in principle be done on a > suitably programmed computer with sufficient speed and memory. Whether or > not this makes the brain a "computer" is really an uninteresting > philosophical argument. > > On Wed, Jun 26, 2019, 3:48 PM Mike Archbold <[email protected]> wrote: > >> Probably what most people mean by computer is roughly the usual common >> sense digital or perhaps quantum computer. There are also theoretical >> hypercomputers. I guess I would define computer simply as something >> that follows the usual form: >> >> input -- function -- output >> >> Input and output might be feedback oriented -- the output is part of >> the input. Or, there may be no output. Maybe a computer just inputs >> and makes a computation. I think the only hard requirement is that >> there is some kind of function happening based upon the input and >> there will almost certainly be output. >> >> So, I my opinion is a brain qualifies as a computer -- under the >> looser definition. >> >> Mike A >> >> >> On 6/26/19, Steve Richfield <[email protected]> wrote: >> > Colin, >> > >> > You are apparently unaware of the varied history of analog computers, >> that >> > include things like electrolytic computers that operated in small tanks >> of >> > conductive liquid. These were used to design motors and transformers >> using >> > the similarity of electric fields to magnetic fields. >> > >> > Also, the hand-crank mechanical tide computer now in a case at NOAA >> > headquarters, that saw more than a century of usefull full time service. >> > >> > Like one man's terrorist is another man's freedom fighter, a computer is >> > anything that usefully computes, REGARDLESS of the intent or perception >> of >> > others. >> > >> > By this standard, a brain clearly IS a computer. What argument can >> there be >> > that it is NOT a computer? >> > >> > Steve >> > >> > On Tue, Jun 25, 2019, 5:23 PM Colin Hales <[email protected]> wrote: >> > >> >> >> >> >> >> On Wed., 26 Jun. 2019, 4:25 am Steve Richfield, >> >> <[email protected]> >> >> wrote: >> >> >> >>> Stefan, >> >>> >> >>> I probably have more neuroscience background than anyone else on this >> >>> list, possibly excepting Colin, having worked as a research assistant >> in >> >>> the Department of Neurological Surgery at the University of >> Washington, >> >>> so >> >>> I suggest caution when challenging me in that venue. >> >>> >> >>> I might also be the only one on this list who has actually held a job >> as >> >>> a mathematician at a university department (University of Washington >> >>> Physics and Astronomy Department). You appear to be clueless in this >> >>> venue, >> >>> at least as it intersects with neuroscience. >> >>> >> >>> My long-term goal is to fuse mathematics, neuroscience, and computer >> >>> science into a single effort leading to AGI and beyond. >> >>> >> >>> Please excuse my talking down to this audience - in my possibly >> hopeless >> >>> efforts to guide this fusion. >> >>> >> >>> My main impediment seems to be the few talented people in each of >> those >> >>> disciplines who ignorantly dismiss the value of those in the other >> >>> disciplines. >> >>> >> >>> People like you, who see things that obviously are part of the >> function >> >>> of neurosystems, but can't (yet?) grok their mathematical >> significance, >> >>> so >> >>> they can apply this understanding to their programming. >> >>> >> >>> There are others like you in the other disciplines - like Colin, who >> >>> immediately dismissed the prospect of ANY sort of mathematics >> operating >> >>> in >> >>> neurons, etc >> >>> >> >> >> >> Hi Steve, >> >> >> >> Can I issue my own millionth ARGHHH!!?? and a word of clarification. >> >> There's a great big crack of misaligned meaning into which your >> >> perspective >> >> on my position has fallen. >> >> >> >> The brain's signalling is 100% computation. Agreed. Right there with >> >> everyone. >> >> >> >> What I am saying is that it's NOT A COMPUTER. Sorry to shout. But this >> >> keeps being missed. The confusion... Is of (a) natural signalling >> physics >> >> doing computation with (b) (the physics of) a computer exploring the >> >> numerical/symbolic properties of a mathematical abstraction of the >> >> natural >> >> computation. >> >> >> >> Both are mathematical activities. >> >> >> >> I do not deny the presence of mathematics at all! >> >> >> >> What I question is the unproved HYPOTHESIS (sometimes called 'substrate >> >> independence', it has various names), that (a) and (b) are or can be >> >> functionally indistinguishable. >> >> >> >> Neither do I deny that potential equivalence! >> >> >> >> What I vehemently demand is that if anyone claims the (a)/(b) >> equivalence >> >> hypothesis is proved, they are wrong. This is because the correct >> science >> >> that tests the equivalence has not started. >> >> >> >> Why? Because if it had started we'd be comparing what computers do with >> >> what an inorganic version of the brain's signalling physics does. It >> >> involves a real test for potential disparity between (a) and (b) .... >> >> That >> >> the hypothesis is false. Instead of the universal assumption of the >> >> equivalence, and confinement to the use of computers. >> >> >> >> The science process that examines potential falsehood of (a)/(b) >> >> equivalence is the appropriate form of the empirical science involving >> an >> >> artificial version of brain-physics-as-computation, has never been >> >> attempted or even proposed. >> >> >> >> You may strongly believe that (a) and (b) are equivalent. You may be >> >> deeply unable to see how (a) and (b) could possibly not be equivalent. >> >> >> >> These are just opinions and have no place in science. Neither of these >> is >> >> an argument that they are/are not equivalent. >> >> >> >> If you want to prove it: DO THE EMPIRICAL SCIENCE CORRECTLY >> >> >> >> I have a design for a chip that does the exact kind of computation >> >> performed by brain's, done with the same physics. It's not a computer. >> >> Someone else could have their own chip design. >> >> >> >> Have I made myself clear? I'm not saying computers can or cannot do >> >> anything. I am saying that in the context of the brain, the equivalence >> >> of >> >> brain-based computation and a computed model of what the brain does is >> >> being inappropriately assumed true without any scientific proof of the >> >> kind >> >> demanded in every other science of a natural phenomenon. >> >> >> >> I hope a have made the state of affairs clear: the science is all >> messed >> >> up. It's gone on way too long. >> >> >> >> The catch phrase? >> >> >> >> "Brains? Yep, they are 100% computation and 100% not a computer" >> >> >> >> If you can get the difference, it will transform AGI and out it back on >> >> the right path after 65 years of (perhaps justified, but clearly now >> past >> >> its use-by) deformity. >> >> >> >> Cheers >> >> Colin >> >> >> >> >> >> >> >> *Artificial General Intelligence List <https://agi.topicbox.com/latest >> >* >> >> / AGI / see discussions <https://agi.topicbox.com/groups/agi> + >> >> participants <https://agi.topicbox.com/groups/agi/members> + delivery >> >> options <https://agi.topicbox.com/groups/agi/subscription> Permalink >> >> < >> https://agi.topicbox.com/groups/agi/T87761d322a3126b1-Mf5fa96e49ab106773fb4c745 > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/T87761d322a3126b1-Mfb0ed9df022c8de55957404c> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T87761d322a3126b1-M0be9de2af3644bd0b0a5c654 Delivery options: https://agi.topicbox.com/groups/agi/subscription
