Why humans learn faster than AI—for now

Discussion in 'Computing and Networks' started by nsaspook, Mar 8, 2018.

  1. nsaspook

    Thread Starter AAC Fanatic!

    Aug 27, 2009
    4,814
    5,207
    https://www.technologyreview.com/s/610434/why-humans-learn-faster-than-ai-for-now/
    It's not surprising that these brute force machine learning or "deep" learning systems have problems when there is little to signal a 'good' path from a 'bad' path. There is very little intelligence in current AI that people didn't already program into it.

    [​IMG]
    Leisure Suit Larry in the Land of the Lounge Lizards
     
    Last edited: Mar 8, 2018
  2. spinnaker

    AAC Fanatic!

    Oct 29, 2009
    6,847
    3,002
    I never really understood how AI works. Does it simply look at all the possibilities of an action in some advanced search, weigh them, then pick from the top of the list? Or is there something far more advanced going on?

    And how is the data feed in? We hear about AI computers "reading" medical journals. Is it actually understanding the text in the files? Or is that data simply converted to some kind of database then loaded in to the AI computer?
     
  3. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    3,971
    5,374
    A cosmic shift will occur when someone invents the digital equivalent of dopamine.
     
  4. nsaspook

    Thread Starter AAC Fanatic!

    Aug 27, 2009
    4,814
    5,207
    I wish I understood how exactly how AI works too.

    With 'Deep learning' there is really no learning or understanding in the way we think of in a classical AI human mimic brain. It works because we now have massive computing power able to hash huge databases generated by the machines when information is loaded into them. For deep learning the machine creates the database from the information input, we don't generate it for them.
    http://karpathy.github.io/2016/05/31/rl/
    A 'learned' machine doesn't understand the physical world difference between these two things but it can recognize them from database created from their images.
    [​IMG]
     
    Tony Hall likes this.
  5. nsaspook

    Thread Starter AAC Fanatic!

    Aug 27, 2009
    4,814
    5,207
    We already have the digital equivalent of LSD-25.
    [​IMG]

    https://www.ibtimes.co.uk/google-de...duced-by-ai-inceptionism-users-online-1509518
     
    joeyd999 likes this.
  6. Raymond Genovese

    Active Member

    Mar 5, 2016
    997
    555
    Blueberry puppies are really tasty, but you can't find them anymore.
     
  7. Raymond Genovese

    Active Member

    Mar 5, 2016
    997
    555
    Not sure that's it. Lots of species have dopamine, even the "lowly" Aplysia
    [​IMG]
    ....and they can't even program a PIC ;)
     
    nsaspook likes this.
  8. spinnaker

    AAC Fanatic!

    Oct 29, 2009
    6,847
    3,002
    So basically exactly as I said? No real "thinking" going on? Just a very sophisticated search and probability algorithm?

    It is my understanding , this is the way AI chess games work. They "simply" run through all of the possible moves and calculate the outcome. I am by no means a good chess player but it is my understanding, this is pretty much what the human players do. The computer works so much faster.

    The photo above illustrates very well how amazing the human brain works. We really aren't doing any kind of search of images. Or at least I don't think so. For some reason, we can easily tell the difference between a puppy and a muffin even when most of the data is hidden from us.
     
  9. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    3,971
    5,374
    And there are things that Aplysia can do that Google could only dream of.
     
    atferrari likes this.
  10. takao21203

    AAC Fanatic!

    Apr 28, 2012
    3,783
    502
    There is some fundamental things absent, among such as

    1) Instinct of survival
    2) Memory of incidents threatening the own survival
    3) Autonomous reproduction
    4) Gaining access and ownership of resources + territory
    5) Rewards of all kinds

    One of the basic determinators of life of course is survival. Its fundamental.
     
  11. Raymond Genovese

    Active Member

    Mar 5, 2016
    997
    555
    Yes, joeyd999, you are quite right. In fact, Eric Kandel received a Nobel prize for telling us about what they can do and how they can do it. But, the key is not the neurotransmiter (any of them) so much as all those neurons.
     
  12. nsaspook

    Thread Starter AAC Fanatic!

    Aug 27, 2009
    4,814
    5,207
    There is something happening but it's unrelated to 'thinking'.

    Because the massive amount of data for learning generates a totally abstract representation of the original data some of the current AI methods can be easily fooled if you understand how they work by making a completely different input data produce a pattern similar to the learned computer response. No human would think these images are what the computer 'thinks' they are.
    [​IMG]
    http://www.evolvingai.org/files/DNNsEasilyFooled_cvpr15.pdf
     
  13. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    3,971
    5,374
    The key to learning is the neurotransmitter, IMHO. It is the reward our brain seeks for successful execution of actions that achieve a goal. It is the reason we repeat learned behavior -- to again experience the reward.
     
    nsaspook likes this.
  14. spinnaker

    AAC Fanatic!

    Oct 29, 2009
    6,847
    3,002

    So how does that explain why people get married a second time? :eek:
     
  15. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    3,971
    5,374
    What a stupid computer. The pictures are obviously blond, brunette, redhead...
     
    xox likes this.
  16. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    3,971
    5,374
    They're promised sex on their wedding night. Major source of dopamine overload.
     
    atferrari, panic mode and nsaspook like this.
  17. Raymond Genovese

    Active Member

    Mar 5, 2016
    997
    555
    No. the transmitters activate or modulate the activity of the neuron and it is much more complicated than that. The same neurotransmitters can act very differently in different species. In fact even in the same species during development. Some neurotransmitters are found all over the CNS and PNS doing their thing on neurons that are doing very different things in the system.

    You always hear things like you are saying that are associated with drug abuse and reward centers and the like, but it is a huge oversimplification.

    But look, I don't want to start an argument about this, it is very complicated, not completely understood and it is more than difficult to discuss in short posts. My "opinion" is based on being a neuroscientist for more than 30 years. I don't claim to know everything about neurons, learning and memory, but I do know some things about them.
     
    nsaspook likes this.
  18. nsaspook

    Thread Starter AAC Fanatic!

    Aug 27, 2009
    4,814
    5,207
    You are affected by
    Universal adversarial perturbations :D


    https://arxiv.org/pdf/1610.08401.pdf

    The deep networks AI ability to see images where no human recognizable image exists can also be exploited to make systems misidentify slightly changed images that look normal to humans.



    Imagine using this on self-driving cars by slightly modifying traffic signs in a way that humans can't detect easily.

     
    Last edited: Mar 8, 2018
  19. spinnaker

    AAC Fanatic!

    Oct 29, 2009
    6,847
    3,002
    But they should have learned what follows thereafter. ;)
     
    Last edited: Mar 9, 2018
  20. joeyd999

    AAC Fanatic!

    Jun 6, 2011
    3,971
    5,374
    You'd think the same about heroin and crack addicts.
     
    nsaspook likes this.
Loading...