Page 8 of 9 FirstFirst ... 6789 LastLast
Results 71 to 80 of 81

Thread: AI: Superintelligence - paths, dangers and strategies

  1. #71
    Meae Musae Servus Hephaestus's Avatar
    Type
    INTP
    Join Date
    Dec 2013
    Location
    Ceti Alpha V
    Posts
    13,738
    Flying machines with flapping wings are called "ornithopters", not planes.

    Funny thing: everyone who made a video of their human powered ornithopter or anything they thought close enough to call one seems to think they're the first.
    Some days you're the bug; some days you're the feature.

    --Meditations on Uncertainty Vol ξ(x)

  2. #72
    know nothing pensive_pilgrim's Avatar
    Type
    INTP
    Join Date
    Dec 2013
    Posts
    4,672
    Quote Originally Posted by Starjots View Post
    Reminds me of The Selfish Gene by Richard Dawkins. Clearly, as you state, this sort of thinking has had a lot of impact.



    Since evolution is really just an algorithm with a few ingredients like overproduction, mutation, selection and so forth, certainly seems to me artificial evolution is possible.
    Of course it's possible, but if humans are just input/output survival and reproduction machines, it should be the only thing possible. That was my point.

    In consumer products the environment that does the selecting is the consumers, the replicators are the manufacturers/makers, the resource that enables reproduction is money and the mutation is the various 'new and improved' variations of products etc.

    The products are created by our conscious minds but don't behave like products of natural evolution. The 9 multigrain bread which has supplanted white bread is not an animal. The latest iPhone or Samsung smartphone is not a living thing either. I think the reason is the environment is humans, not the natural environment. Then again, domestic animals have been artificially evolved but they start from living creatures.

    So I'd say in a way our products do behave in a sense - they have populations, thrive, go extinct and so forth from an extrinsic perspective but have no or limited behavior intrinsically. Then again, technology has begun to make smarter products which might have behaviors but these in are supposed to be in our service.

    So..... could it be our collective human desire to select for more 'behavioring' and 'human like' qualities in products is enough to drive their coming into existence? I'd say maybe yes!
    I think the mechanism by which consumer products succeed or fail could be largely evolutionary. What really changes the formula is that new products aren't just variations of old ones, instead we can be inspired to create in entirely new ways. I suspect that this kind of creativity was also the spark of abiogenesis, and the reason why anything at all exists.

  3. #73
    Meae Musae Servus Hephaestus's Avatar
    Type
    INTP
    Join Date
    Dec 2013
    Location
    Ceti Alpha V
    Posts
    13,738
    Product success or failure is more influenced by marketing than evolution is. Marketing plays a nontrivial role in species that depend on either other species or other members of their species for sexual reproduction, but every species I can think of has a means to reproduce and potentially evolve without being successful marketing for sexual reproduction.

    Not so with tools.
    Some days you're the bug; some days you're the feature.

    --Meditations on Uncertainty Vol ξ(x)

  4. #74
    Senior Member Starjots's Avatar
    Type
    INTP
    Join Date
    Dec 2013
    Posts
    2,174
    Deep You

    See the OP that started this moribund thread. IN that article, an AI program that only knew the rules taught itself to play GO, the most complex game around, better than any other program. And naturally the program it beat had already beaten the best human GO player.

    Now the same thing has been done with Chess.

    With no input other than rules, an AI learned enough to beat all other chess programs. Humans aren't even in the same league in Chess.

    The Singularity is near. Just sayin.

    The

  5. #75
    just dont think about it mhc's Avatar
    Type
    iNtP
    Join Date
    Feb 2014
    Posts
    770
    Quote Originally Posted by Starjots View Post
    Deep You

    See the OP that started this moribund thread. IN that article, an AI program that only knew the rules taught itself to play GO, the most complex game around, better than any other program. And naturally the program it beat had already beaten the best human GO player.

    Now the same thing has been done with Chess.

    With no input other than rules, an AI learned enough to beat all other chess programs. Humans aren't even in the same league in Chess.

    The Singularity is near. Just sayin.

    The
    I find AI fascinating, however im a bit skeptical as to what AI actually means to the consumer today. please tell me that this is not just a program that has found calculated (by maths invented by humans *cough* *cough*) a path to "1" over "0" with every given and possible variable, calculated after every move.
    Just look at the blue sky

  6. #76
    Senior Member Starjots's Avatar
    Type
    INTP
    Join Date
    Dec 2013
    Posts
    2,174
    Quote Originally Posted by mhc View Post
    I find AI fascinating, however im a bit skeptical as to what AI actually means to the consumer today. please tell me that this is not just a program that has found calculated (by maths invented by humans *cough* *cough*) a path to "1" over "0" with every given and possible variable, calculated after every move.
    No I think not. Humans just perfected a good model for evolving game playing code that teaches itself as far as I know. Since it's a neural network or something like that, exactly how it plays so well is basically a mystery to any human who might peek under the hood at the weights in the connections or whatever the neuromorphic approach is.

    To wit, I believe any instance of the AI that has trained itself sufficiently and played a person would stump them on both the WHY it makes certain moves (why did it sacrifice the queen??) and HOW it came to make such moves (how did the weights in the neural net lead to this decision?).

    By analogy/equivalence, this is precisely how a much smarter person would seem to a much dumber person and in terms of GO and CHESS, we are now the dumber people. Even our software that plays these games and relies on our insights baked into the programming is dumber than this self taught evolving approach.

    --

    I don't know about what it means for a consumer except perhaps a continuation and deepening of the fog about why we buy what we buy as larger forces, smarter than the old advertising execs, fool us in ways we can't even fathom.

  7. #77
    Pull the strings! Architect's Avatar
    Type
    INTP
    Join Date
    Feb 2014
    Location
    West Coast
    Posts
    545
    Quote Originally Posted by Starjots View Post
    To wit, I believe any instance of the AI that has trained itself sufficiently and played a person would stump them on both the WHY it makes certain moves (why did it sacrifice the queen??) and HOW it came to make such moves (how did the weights in the neural net lead to this decision?).
    Well, not quite. For example take the historical AlphaGO games a few years ago. There was a move in match 3 IIRC which made no sense. I was watching the game live at the time, the American Go master commenting kind of got that quirky grin he gets and put the move where he thought it would be, and then did a few double takes to what the actual move was. Then they thought it was a misplacement by the human handler, but no that was the move, and AlphaGo won the game. Later analysis was clear it was no move that had ever been made before, and no human likely ever would make, but it was decisive and we can therefore see the kind of strategies that machine go makes. The training goal is to win the game, and while humans like to do that with a large win margin, turns out the machine prefers low margin wins. That is, nothing spectacular, just play that continually has a > 50% probability of winning. Second is again emotional/psychological, the machine is fine with making large embarrassing sacrifices, if the probability of win is maintained.

    Ultimately this all derives from what we call the 'cost function' - all the system cares about is optimizing for cost. Humans optimism for other reasons - we like large obvious wins and no bad looking moves. So in this way we know exactly why the network is making the decisions it makes. Indeed we must, we were the ones that set the training goal (cost function) up in the first place.

    Secondly, if you care about the specific weightings of the network there are many ways to see why it does what it does. For example, here's one from a network I did years ago to detect cats

    unnamed-2.jpg

    What you are seeing are the internal weights of a conv net cat detector, rendered as a picture. Think of it as a 'heat map'. Red means low probability of cat detection features (or regions), cooler colors is higher. This picture was created by the training - I didn't make it but the 'network made it' (roughly). Anyhow take a step back, notice anything? Do you see cats? My wife is an artist and can see lots of cats. If you spend some time at it you'll realize that the network recognizes a cat as something with two triangle ears, two eyes and a nose thing, usually centered in the middle of the picture (people take portraits like that).

    So it's all pretty clear actually, in a probabilistic kind of way.

  8. #78
    your cheapest wine Johnny's Avatar
    Type
    INTp
    Join Date
    Mar 2014
    Location
    Kamino
    Posts
    440
    I think Mary Shelley's brilliant Frankenstein influences the average AI debate likely too much.
    What's the difference? It's just soda, bro.

  9. #79
    Senior Member Starjots's Avatar
    Type
    INTP
    Join Date
    Dec 2013
    Posts
    2,174
    I think it's pretty human to see human-ness in all sorts of things, sort of a modern animism belief system. Maybe I'm more guilty of that.

    But - hook a machine mind to an artificial or virtual body and I see, at a conceptual level, no difference with a human mind except one is much more adaptable.

  10. #80
    your cheapest wine Johnny's Avatar
    Type
    INTp
    Join Date
    Mar 2014
    Location
    Kamino
    Posts
    440
    I'm not clear that we can't be guilty of that, no matter our precautions.

    I've been watching the performance of AIEQ with respect to SPY (both are stock market investment ETFs). I'm not impressed with AIEQ, but not because I think Watson sucks. I'm really just trying to reframe the purpose.of AI in some way that neuters any fear of it (e.g., Kubrick's 2001).

    Surprisingly, I'm having great difficulty...because Frankenstein you know...
    What's the difference? It's just soda, bro.

Similar Threads

  1. Tea strategies
    By Blorg in forum The Pub
    Replies: 46
    Last Post: 05-05-2017, 02:54 PM
  2. People's Rubic's Cube strategies
    By BarIII in forum The Pub
    Replies: 20
    Last Post: 08-09-2016, 04:28 PM
  3. Foolishly reveal your password creation strategies
    By notdavidlynch in forum The Pub
    Replies: 41
    Last Post: 09-15-2014, 09:34 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •