Coffeehouse Thread

43 posts

What is the Future of Artificial Intelligence?

Back to Forum: Coffeehouse
  • User profile image
    Dziyakerxs

    "Citius, Altius, Fortius"

    The trend is to go "swifter" (more gigahertz, more cpu's), "higher" (larger disk space = larger virtual memory = larger cache = more paging), and "stronger" (bigger RAM)...


    What now for meta-heuristics?

    Are we seeing the end of Artificial Neural Networks? Genetic Algorithms? Simulated Annealing and Ant Colony Optimization among others? Is the "quad-core" the final solution to NP-Complete and NP-Hard problems? Is it over for us Computer Scientists and a renaissance for the Engineers?



    Its brain-power versus good-old muscle...

    Tell us what you think...

  • User profile image
    littleguru

    How could a quad core solve the NP issue? Just add a few more instances - increase N a little bit - and you are at the same issues. More power (at least speaking one core vs quad core) won't solve that issue.

    As it is with other AI stuff: more power always helped, but also smart algorithms need to be in place!

  • User profile image
    Dziyakerxs

    I hear you...

    But the fact of the matter is research into intelligent algorithms and heuristics is steadily declining.....

    Consider Deep Blue... Beat Gary right? As I've read in one article, they called it a "Hardware-Software Synergy". The hardware part being the paralled processors and the software part the alpha-beta pruning strategy.

    I think only a handful of programmers and/or engineers nowadays know how and what it is to solve complex problems, and real-world problems at that. Media (advertising) optimization and Credit-Risk evaluation just to name two.

  • User profile image
    kettch

    While it's not the solution, part of developing a good AI seems to be throwing more hardware at it. Until hardware can approach the ability of organic structures to do concurrent processing, I don't think that AI is going to get very far.

  • User profile image
    littleguru

    I have had a lot AI lectures at university and they never told us that high performance will solve the problems... They usually told us that if you are able to proof that NP problems can be transformed into a P problem, you earn like $ 1,000,000 or so Wink

  • User profile image
    Royal​Schrubber

    Dziyakerxs wrote:
    Consider Deep Blue... Beat Gary right? As I've read in one article, they called it a "Hardware-Software Synergy". The hardware part being the paralled processors and the software part the alpha-beta pruning strategy.


    Actually deep blue (and probably no other human made machine) is able to beat human champion in traditional asian game called Go.

    What Deep Blue does is it looks at trillion moves - if Kasparov was able to do that I am sure he would win easily. In Go there is much more moves than computer can calculate plus usual chess methods for position evaluation are useless here. Here tactic is important - real intelligence.

    True AI does not rely on power, quad core is at best only 4x faster which is miniscule factor - to make noticable difference you need factors of 30x and more - IBM can always throw four times more machines in server farm - quad core makes no difference at all.

  • User profile image
    kettch

    littleguru wrote:
    I have had a lot AI lectures at university and they never told us that high performance will solve the problems... They usually told us that if you are able to proof that NP problems can be transformed into a P problem, you earn like $ 1,000,000 or so


    No, it's not speed that makes the difference. The human brain is incredibly slow compared to even primitive processors. It's the ability to perform bajillions of parallel operations that makes the neural activity of organic systems difficult to match.

    It's possible that rather than spending money on powerful machines, a much larger cluster of low power processors would be better. However, you are still stuck with the problem of programming by hand what genetics already has.

  • User profile image
    littleguru

    kettch wrote:
    
    littleguru wrote:
    I have had a lot AI lectures at university and they never told us that high performance will solve the problems... They usually told us that if you are able to proof that NP problems can be transformed into a P problem, you earn like $ 1,000,000 or so


    No, it's not speed that makes the difference. The human brain is incredibly slow compared to even primitive processors. It's the ability to perform bajillions of parallel operations that makes the neural activity of organic systems difficult to match.

    It's possible that rather than spending money on powerful machines, a much larger cluster of low power processors would be better. However, you are still stuck with the problem of programming by hand what genetics already has.


    This is not completely true. If you add n cores (cpus in a cluster) you are getting management overhead. That reduces performance. Both is required: clustering + performance!

  • User profile image
    Dziyakerxs

    littleguru wrote:
    I have had a lot AI lectures at university and they never told us that high performance will solve the problems... They usually told us that if you are able to proof that NP problems can be transformed into a P problem, you earn like $ 1,000,000 or so


    Yeah... The P vs NP problem is one of the questions listed at this site:

    "http://www.claymath.org/millennium/"

  • User profile image
    Dziyakerxs

    kettch wrote:
    While it's not the solution, part of developing a good AI seems to be throwing more hardware at it. Until hardware can approach the ability of organic structures to do concurrent processing, I don't think that AI is going to get very far.


    I have to agree with you. I have experienced trying to solve problems approaching exponential computation times (not to mention the space needed to store those states)... And even if I was utterly sure that nothing more could be done to optimize the approach (by my judgment), running time for the algorithm is still quite un-wait-able for the uninformed and cannot-be-put-off user.

    -- And this was 2 years ago... On a common office machine.

  • User profile image
    Dziyakerxs

    RoyalSchrubber wrote:
    
    Dziyakerxs wrote:
    Consider Deep Blue... Beat Gary right? As I've read in one article, they called it a "Hardware-Software Synergy". The hardware part being the paralled processors and the software part the alpha-beta pruning strategy.


    Actually deep blue (and probably no other human made machine) is able to beat human champion in traditional asian game called Go.

    What Deep Blue does is it looks at trillion moves - if Kasparov was able to do that I am sure he would win easily. In Go there is much more moves than computer can calculate plus usual chess methods for position evaluation are useless here. Here tactic is important - real intelligence.


    From what I could remember, Deep Blue (the software [AI]) went down 6 plies, pruning the tree, and then deferred further calculation to the processors (Hardware Deep Blue).

  • User profile image
    Royal​Schrubber

    Dziyakerxs wrote:
    
    RoyalSchrubber wrote:
    
    Dziyakerxs wrote:
    Consider Deep Blue... Beat Gary right? As I've read in one article, they called it a "Hardware-Software Synergy". The hardware part being the paralled processors and the software part the alpha-beta pruning strategy.


    Actually deep blue (and probably no other human made machine) is able to beat human champion in traditional asian game called Go.

    What Deep Blue does is it looks at trillion moves - if Kasparov was able to do that I am sure he would win easily. In Go there is much more moves than computer can calculate plus usual chess methods for position evaluation are useless here. Here tactic is important - real intelligence.


    From what I could remember, Deep Blue (the software [AI]) went down 6 plies, pruning the tree, and then deferred further calculation to the processors (Hardware Deep Blue).


    That does not change anything - everything that we invented is inferior to human brains - what we do now is using cheap shortcuts with linear processing speed potential of our modern processors. It's not real AI - machine learns almost nothing in all cases.

    Btw, I think it's entirely possible that someone could create real thinking AI. Some argue that it's imposible to simulate trillions of neurons, but I think that some day a guy will come and use one trick that will simplify simulation to reasonable size. Bubblesort can hardly sort billion elements, but it's piece of cake for quicksort. That kind of trickery Wink
    Also to me it looks possible that this kind of magic has already been discovered and that there are guys who have real AI. It looks reasonable to hide this kind of technology - someone can earn billions of dollars with it (and potentially change the globe if he releases it)...

  • User profile image
    Dziyakerxs

    RoyalSchrubber wrote:
    
    Dziyakerxs wrote:
    
    RoyalSchrubber wrote:
    
    Dziyakerxs wrote:
    Consider Deep Blue... Beat Gary right? As I've read in one article, they called it a "Hardware-Software Synergy". The hardware part being the paralled processors and the software part the alpha-beta pruning strategy.


    Actually deep blue (and probably no other human made machine) is able to beat human champion in traditional asian game called Go.

    What Deep Blue does is it looks at trillion moves - if Kasparov was able to do that I am sure he would win easily. In Go there is much more moves than computer can calculate plus usual chess methods for position evaluation are useless here. Here tactic is important - real intelligence.


    From what I could remember, Deep Blue (the software [AI]) went down 6 plies, pruning the tree, and then deferred further calculation to the processors (Hardware Deep Blue).


    That does not change anything - everything that we invented is inferior to human brains - what we do now is using cheap shortcuts with linear processing speed potential of our modern processors. It's not real AI - machine learns almost nothing in all cases.

    Btw, I think it's entirely possible that someone could create real thinking AI. Some argue that it's imposible to simulate trillions of neurons, but I think that some day a guy will come and use one trick that will simplify simulation to reasonable size. Bubblesort can hardly sort billion elements, but it's piece of cake for quicksort. That kind of trickery
    Also to me it looks possible that this kind of magic has already been discovered and that there are guys who have real AI. It looks reasonable to hide this kind of technology - someone can earn billions of dollars with it (and potentially change the globe if he releases it)...


    I agree with you totally... That kind of magic will certainly change everything as we know it. But I don't think It'll be coming soon. We've all been offered that "dare to be great" moment. Its up to us on how and what to respond right?

    Going back, not a single person has had a really profound impact on the history of computing. Well perhaps Donald Knuth! Big Smile  The chair is up for grabs... Its a race to create the real thinking machine...


    Which brings me to ask you if you think that Machine Intelligence should be measured by Human Intelligence (take for example the Turing Test)... What do you think?

    Maybe thats for another post huh? Lets see...

  • User profile image
    Minh

    Does that question interest you?

  • User profile image
    jason818_25​3.33


    You guys are really smart, so i dont think i can really address all the things your talking about.

    but... you did introduce me to the turing test. i didnt know what it was till i read this post and desided to look it up on wiki. so now for a funny expierence.

    Once i accepted a phone call from my mom, but this phone call wigged me out. it actually scared me to the point to where i began to swet and get really afraid. and why did this call scare me? because i was a frade i was talking to a computer. it really scared me. It sounded just like her, it had her voice, her inflections, her vocabulary, and even some life expierences in the data base. i asked some questions to clarify if it was really my mom, but to no avail in coming to a solidified conclution. actually it took me a while befor i would even talk to her over the phone agian because i thought it was a computer on the phone and i didnt want to talk to a computer, i wanted to talk to my mom. the only way i could even use the phone (because i was beggining to think any one on the other end could potentialy be impersinated by a computer) so as a solution i imagined that if it was a computer on the other end that maybe it was being used as more of a answereing machine and so some how on the other end, my real flesh and blood mom would be getting the message i was sending, even if on her end of the line it wasnt me she was talking to but maybe if it was a computer impersinating me. ether way i figured, or hoped that she was getting the message i was trying to send. after a while i got over this fear and was able to accept i was talking to my mom. but then again, if i think to much about it, im not really talking to my mom when on the phone, but rather im talking to a bunch of wires and plastic, and threw the wire and plasic and alot of cable the message im sending gets sent over thousands of miles where my mom is in fact talking to wires and plastic her self.

    so whats the big deal of this and who cares about the turing test. who cares if a computer is as "smart" or intelegent as us? (personaly i would love it if i could speak to some one or something more intelegent than us- from the looks of things some times we arent all that bright-so if somethings brighter than us, gee wiz will you, help us). Personaly i dont think the turing test is much to talk about regarding inteligent life, but i do think it gives us volumes to evalute. If my mom is being impersinated and i have no clue what end is up, or what end is down, then i am in alot of trouble. i just might fly my plane strait into the ground. that would be bad. and im not instriment rated. yikes, that is really bad.  bla bla bla you woud say that im crazy.

    ps Im actually a bot and any one who was convinsed into thinking i was a human was fooled. =) just kidding. im human.

    Edit: the point I was trying to make was,. As technology evolves it creates challenges and problems unforeseen before hand. If AI is created to clear the Turing test, then what is to keep us humans from knowing what is a real person or what is not. I can think of many devious things to do with an AI that passes for a real human.

    Like the advent of vehicles or the internet, what safe guards will be placed to protect the public from such a technological advancement?

  • User profile image
    jsampsonPC

    I was playing poker with a group of people the other evening, one of them was 86 years old. He said, "Son, what do you do?" and I said, "I'm a software developer - I work with computers."

    He looked back and said, "It's amazing what these computers are doing today. Someday we'll be able to talk to them and they will do stuff." I just smiled and said.."yeah..." Smiley

  • User profile image
    Dziyakerxs

    Minh wrote:
    Does that question interest you?


    You mean the Human Intelligence vs. Machine Intelligence Issue?

    Well...

    It does indeed! Smiley

    As a matter of fact, I'm starting a thread discussing it after I put some more of my thoughts regarding the other replies here. It really is quite nice having found really smart people around here (and sharing really good ideas at that)...

  • User profile image
    Dziyakerxs

    jason818_253.33 wrote:

    You guys are really smart, so i dont think i can really address all the things your talking about.

    but... you did introduce me to the turing test. i didnt know what it was till i read this post and desided to look it up on wiki. so now for a funny expierence.

    Once i accepted a phone call from my mom, but this phone call wigged me out. it actually scared me to the point to where i began to swet and get really afraid. and why did this call scare me? because i was a frade i was talking to a computer. it really scared me. It sounded just like her, it had her voice, her inflections, her vocabulary, and even some life expierences in the data base. i asked some questions to clarify if it was really my mom, but to no avail in coming to a solidified conclution. actually it took me a while befor i would even talk to her over the phone agian because i thought it was a computer on the phone and i didnt want to talk to a computer, i wanted to talk to my mom. the only way i could even use the phone (because i was beggining to think any one on the other end could potentialy be impersinated by a computer) so as a solution i imagined that if it was a computer on the other end that maybe it was being used as more of a answereing machine and so some how on the other end, my real flesh and blood mom would be getting the message i was sending, even if on her end of the line it wasnt me she was talking to but maybe if it was a computer impersinating me. ether way i figured, or hoped that she was getting the message i was trying to send. after a while i got over this fear and was able to accept i was talking to my mom. but then again, if i think to much about it, im not really talking to my mom when on the phone, but rather im talking to a bunch of wires and plastic, and threw the wire and plasic and alot of cable the message im sending gets sent over thousands of miles where my mom is in fact talking to wires and plastic her self.

    so whats the big deal of this and who cares about the turing test. who cares if a computer is as "smart" or intelegent as us? (personaly i would love it if i could speak to some one or something more intelegent than us- from the looks of things some times we arent all that bright-so if somethings brighter than us, gee wiz will you, help us). Personaly i dont think the turing test is much to talk about regarding inteligent life, but i do think it gives us volumes to evalute. If my mom is being impersinated and i have no clue what end is up, or what end is down, then i am in alot of trouble. i just might fly my plane strait into the ground. that would be bad. and im not instriment rated. yikes, that is really bad.  bla bla bla you woud say that im crazy.

    ps Im actually a bot and any one who was convinsed into thinking i was a human was fooled. =) just kidding. im human.



    Well you know what? I actually care if we someday produce something that could eventually (or maybe even naturally) surpass our own intelligence, if ever it could be measured.....

    Let me ask you this...
    What would become of us humans?

    Our intelligence is the very key to our being on top of the food chain. Of course machines may never actually prey on us, but the outcome is so disturbing... Imagine a world where humans are treated as second to something or second to someone..... I hope we both have the same picture on our minds...

    Worst case scenario: (I hope you wont boil because of this...)

    You get home one day and you find two realities of your mother, one is your real biological mother, and the other, a copy... Not a clone, but an exact replica. Using the most advanced technology imaginable, somehow, someone has copied every aspect of your mother and put it in another container... And breathed life into it... You cant identifty which one is your real mom... You reason, you test, you ridicule, you criticize, you converse, you do everything you could imagine to know which of them is real... And then it hits you.....

    THEY ARE BOTH REAL...
    BOTH OF THEM IS YOUR MOM...
    BUT ONE OF THEM IS A MACHINE...


    Welcome to the twilight zone Jason. I'm sure just imagining it would make you go nuts.

    You've experienced the phone call... What's next?

Comments closed

Comments have been closed since this content was published more than 30 days ago, but if you'd like to continue the conversation, please create a new thread in our Forums, or Contact Us and let us know.