1. We have added the ability to collapse/expand forum categories and widgets on forum home.
    Dismiss Notice
  2. Photobucket has changed its policy concerning hotlinking images and now requires an account with a $399.00 annual fee to allow hotlink. More information is available at: this link.
    Dismiss Notice
  3. All Civ avatars are brought back and available for selection in the Avatar Gallery! There are 945 avatars total.
    Dismiss Notice
  4. To make the site more secure, we have installed SSL certificates and enabled HTTPS for both the main site and forums.
    Dismiss Notice
  5. Civ6 is released! Order now! (Amazon US | Amazon UK | Amazon CA | Amazon DE | Amazon FR)
    Dismiss Notice
  6. Dismiss Notice
  7. Forum account upgrades are available for ad-free browsing.
    Dismiss Notice

Will A.I. reach human intelligence?

Discussion in 'Science & Technology' started by Narz, Apr 24, 2010.

?

Will A.I. reach/surpass human intelligence?

  1. Yes, within 25 years.

    15.6%
  2. Yes, within 100 years.

    26.7%
  3. Yes, but not for a long, long time.

    24.4%
  4. Probably not.

    11.1%
  5. No.

    15.6%
  6. Not sure.

    6.7%
  1. Kyriakos

    Kyriakos Alien spiral maker

    Joined:
    Oct 15, 2003
    Messages:
    45,330
    Location:
    Thessalonike, Greece
    There are also articles- i heard this on the news though, no source- that claim that the future computers will be programmed not in a computer language as those available today, but in ancient greek, making use of its structure which is said to be more complicated than other languages.

    Αυτόματον Ειμί ;) (i am Automaton)
     
  2. ParadigmShifter

    ParadigmShifter Random Nonsense Generator

    Joined:
    Apr 4, 2007
    Messages:
    21,810
    Location:
    Liverpool, home of Everton FC
    I'm guessing you heard that news on April Fool's Day?

    Computers are programmed in languages that the computer finds easy to understand.
     
  3. peter grimes

    peter grimes ... Moderator

    Joined:
    Jul 18, 2005
    Messages:
    13,143
    Location:
    Queens, New York
    That's BS.
     
  4. ParadigmShifter

    ParadigmShifter Random Nonsense Generator

    Joined:
    Apr 4, 2007
    Messages:
    21,810
    Location:
    Liverpool, home of Everton FC
    Nah, it's Beta Sigma.
     
  5. peter grimes

    peter grimes ... Moderator

    Joined:
    Jul 18, 2005
    Messages:
    13,143
    Location:
    Queens, New York
    LOL

    What timing! I was searching for link to an article I read a year or more ago that discussed a computer set-up that was able to derive mathematical theorems without any guidance from people, when I came across this:

     
  6. Spoonwood

    Spoonwood Grand Philosopher

    Joined:
    Apr 30, 2008
    Messages:
    4,791
    Location:
    Ohio
    I voted "not sure". What does one mean by "human intelligence"? We'd first have to have a more-or-less clear idea of this to evaluate such a claim. Not sure anyone really has this. On top of this, does human intelligence refer to one thing or many things? Sure Einstein had intelligence. But, if he tried to compose a symphony, he would display the qualities of an "idiot", and thus wouldn't come as intelligence, he'd qualify as an idiot. Or say Newton had tried to analyze the origin of species. Newton would end up an idiot. Or Darwin trying to develop fluid mechanics. Darwin ends up an idiot then also.
     
  7. PeteAtoms

    PeteAtoms FormulaRandom

    Joined:
    Aug 13, 2002
    Messages:
    3,722
    Location:
    Land of Ooo
    Or a bunch of Chinese dudes with two-way radios, but not really...
     
  8. warpus

    warpus In pork I trust

    Joined:
    Aug 28, 2005
    Messages:
    46,612
    Location:
    Stamford Bridge
    Now that's what I'm talking about! This is true AI.

    The "AI" in our games and portable devices is not real AI.
     
  9. Souron

    Souron The Dark Lord

    Joined:
    Mar 9, 2003
    Messages:
    5,947
    Location:
    (GMT-5)
    Actually that's one thing that we have thought computers to do. It's called an expert system, and it models human's ability to reason.

    And really, it's no surprise, since reason is well understood. Anything that can be totally described can be mimicked by a computer (given enough resources).
     
  10. ParadigmShifter

    ParadigmShifter Random Nonsense Generator

    Joined:
    Apr 4, 2007
    Messages:
    21,810
    Location:
    Liverpool, home of Everton FC
    Indeed. Akinator, the web genius, is very good

    http://uk.akinator.com/

    EDIT: OMG! First person I thought of Richard Bacon, on the radio now, got easily ;)

    He took more than 20 questions to get Marcel Marceau though.
     
  11. Souron

    Souron The Dark Lord

    Joined:
    Mar 9, 2003
    Messages:
    5,947
    Location:
    (GMT-5)
  12. Spoonwood

    Spoonwood Grand Philosopher

    Joined:
    Apr 30, 2008
    Messages:
    4,791
    Location:
    Ohio
    Expert systems generally come as domain specific. Humans can and do reason across domains. Do you know of any expert systems that do this? The Wikipedia article also indicates that expert systems also have trouble adapting to changing environments. Human can usually reason across different environments without changing their knowledge basis. So, how do expert systems model human reasoning well in general again?

    I certainly don't agree with this. What makes you say this? I could speculate on what you mean, but I'll allow you to describe what you mean here. If reason comes as well understood, then where do we have a complete theory of commonsense reasoning?

    What can actually get totally described? I simply don't see us having the capacity to do this even for biological cells (which contain moles of molecules, which contain even more atoms, which contain even more sub-atomic particles... which means well over trillions of particle-waves in action).

    And how can actually have enough computational resources for mimicking the reasoning of a human mind in real-time? Bremmerman's limit basically implies we can't.
    http://www.wmiles.com/2010/01/bremermanns-limit
    http://rogercostello.wordpress.com/category/bremermanns-limit/
     
  13. Souron

    Souron The Dark Lord

    Joined:
    Mar 9, 2003
    Messages:
    5,947
    Location:
    (GMT-5)
    There's nothing in principle preventing expert systems from being general. It's simply a matter of scale.

    trueknowledge.com is an attempt at a universal expert system.
    WolframAlpha has similar goals, though it is not clear to what extent it uses expert system logic. Mostly WolframAlpha is build on Mathematica.

    I mean that the field of logic is well understood. What do you mean by "commonsense reasoning", and what is not completely understood about it?

    Anything in the field of logic or math can be. Anything that can be approximated by a mathematical framework can be mimicked to the extent that that framework is accurate.

    I think you somehow mistook my statement that reason can be mimicked to mean that we should build biology copying AI. I do not advocate that. Here's my thoughts on it anyway though.

    I agree that barring huge quantum computers, accurately portraying biological cells is unfeasible, unless a simplified model is used.

    I'm not sure what you're trying to say here. The human brain is a computer that can reason in real time. Similarly a functionally equivalent silicon replica can be made to work in real time. As I recall, the problem is that we cannot fit that many logical neurons on a chip yet. But we're on the cusp, and should be able to before moore's law is no more. Also, current computers are much faster that the human brain, so it may be possible to sacrifice time, and make the same space mimic multiple neurons over a fix time interval.

    This is not very useful though, because we don't have a map of all our neural connections, or a complete understanding of how they evolve. We'll need that to copy a brain.

    Of course that assumes that the lowest component modeled is a neuron. Going deeper would soon require quantum mechanics, which cannot be modeled in less than exponential complexity on a conventional computer.
     
  14. warpus

    warpus In pork I trust

    Joined:
    Aug 28, 2005
    Messages:
    46,612
    Location:
    Stamford Bridge
    I agree with Spoonwood.

    We don't really understand the way humans reason. We understand bits and pieces, but we're far from having a good grasp on the subject. I also don't see artificial intelligence being able to cross from domain to domain and adapt nearly as well as we can, unless it's A.I. modelled after the human brain/a neural net.
     
  15. Souron

    Souron The Dark Lord

    Joined:
    Mar 9, 2003
    Messages:
    5,947
    Location:
    (GMT-5)
    We surely understand more of how humans reason than how brain neurons interconnect.

    What kind of human reasoning do we not understand?
     
  16. Tabster

    Tabster Chieftain

    Joined:
    May 22, 2007
    Messages:
    431
    Will the A.I intelligence have an imagination?

    I would like to believe there is more than just cold hard reasoning to human intelligence.

    Would the machine need somekind of emotional feedback system?
     
  17. peter grimes

    peter grimes ... Moderator

    Joined:
    Jul 18, 2005
    Messages:
    13,143
    Location:
    Queens, New York
    Very interesting question. Emotion is integral to our intelligence, but I'm not sure there's a reason to believe that it's necessary for any conceivable intelligence.
     
  18. Spoonwood

    Spoonwood Grand Philosopher

    Joined:
    Apr 30, 2008
    Messages:
    4,791
    Location:
    Ohio
    I find that funny. I mean under mathematics for "the square root of 501" it lists 22.3830292856 which works as wrong at two levels.
    1. The machine doesn't spot the contradiction in terms by talking about "the square root of 501", since the definite article "the" means exactly one and only one. 501 has TWO square roots in the reals, so "the square root of 501" ends up a contradiction in terms. Or it arbitrarily favors positive roots over negative roots. The machine doesn't spot this as an arbitrary assumption about roots.

    The cube root example works out worse, since you can't as easily produce complex roots from a positive real root.

    2. It produces a rational number, while a square root of 501 comes as an irrational number. So, the machine should write 22.3830292856..., but it doesn't.
    This one I also find kind of funny. If you trusted that machine and had no knowledge of American presidents, you would think it gave you a complete list of the presidents for the years during which Barrack Obama's age lay in {13, ..., 19} where the members of that set only take on the values of natural numbers. But, of course, the machine ONLY gives you the first and last presidents in that time period.

    I certainly dispute this. Look, here's one possible propositional logic which I doubt you'll find any logician talking much about, and I doubt they know how to "understand" it. Let 1 denote truth in the classical sense. Let 0 denote falsity in the classical sense. Numbers in (0, 1) will denote degrees of truth between 0 and 1. Let u stand for disjunction of propositions, i for intersection, -> for implication, ~ for negation, and <-> for equivalence. Then, one possible extension of the classical logic connectives which can serve for as a propositional logic comes as:

    i(a, b)=.4 if a=.2
    .1 if a=.3
    min(a, b) otherwise

    u(a, b)=pi-3 if a=e-2, where pi indicates the transcendental number pi, and e indicates the exact value of the exponentional function at 1.
    max(a, b) otherwise.
    ~(a)=1 if a=0, 0 if a=1, a otherwise.
    a->b=.4 if a=.5 or b=.5
    min(1, 1-a+b) otherwise
    a<->b=+sqrt(2)-1 if a=.4 or if b=+sqrt(3)-1, where +sqrt(n) indicates the positive square root of n.
    1-abs(a-b) otherwise.

    Who has studied and understood such a propositonal logic? Thing comes as one can produce as many propositional logics like the above as there exist real numbers.

    It may not quite work, but I'll assume "commonsense reasoning" as synonymous with dispositional logic. In other words, we have propositions which usually hold true, but not always. What doesn't come as completely understood comes as what "usually" means within a context. It doesn't just mean "most" or greater than or equal to 50%. The problem comes as that "usually" varies very, very widely across contexts. It might mean between 65% and 75% of the time (with say 70% as the best fit) in one context, while it might mean above 90% and below 93% in another context.

    Then completely describe a single irrational number. If you write ... or etc. or some similar phrase in your description I don't see the description as complete, I see it as only describing part of the entity in question, and then us inferring from that pattern how such a complete description would work given that we had infinite time and resources to go on with the description. But, of course, we never have such time and resources.

    As another challenge, describe all ways of describing "usually" as a fuzzy number of type-1. Or a type-2 fuzzy number.

    This means we do NOT have a total description. We have an approximation which, by definition, misses some part of the entity in question.

    Alright, say I grant that. Then, what kind of circuits does it have? What sort of logic gates in the "human brain" allow some electrical signals to pass through and others not to pass through? There exist more problems than just fitting logical neurons onto a chip. Computers only work with electrical signals. Neurons work with both electrical and chemical signals. Can we really imitate such a system just by electrical signals? I doubt it.

    Neurons don't work as the only carriers of information. Glial cells also do such. There exists a lot more going on in the brain than just neurons communicating. See the video "networks in the brain" here.
     
  19. Tabster

    Tabster Chieftain

    Joined:
    May 22, 2007
    Messages:
    431
  20. Souron

    Souron The Dark Lord

    Joined:
    Mar 9, 2003
    Messages:
    5,947
    Location:
    (GMT-5)
    What are you talking about? it gives four presidents. Ragan, Nixon, Carter and Ford.
     

Share This Page