Thursday, December 29, 2016

Learning assembly: the cure for AI delusions?

I am searching for an assembly language simulator for I can teach my Operating Systems students how processes work at the CPU level. In the course of doing so, I came across this site, and found:
10110000 01100001

The first few bits (10110) are an instruction to copy a value into a register. The next three digits (000) identify the register which the value will be copied into. The rest of it (01100001) is the value which is to be copied.

Of course 10110 is meaningless, and the computer doesn't "know" that it means "copy the value." The processor is designed so that the series of electrical impulses represented by 10110 (on-off-on-on-off) causes the desired result. This is part of what is meant by "mechanical."
Yesiree. Maybe if all of the AI true believers had to program in assembly for a month, they'd all realize, "Oh yeah, it's just a bunch of circuits performing that exact mechanical operations I set up for them to perform."


21 comments:

  1. Is that assembly code or machine code ?

    In any case at some point someone must have programmed for a month or more to write a programming language like lisp that is (I think) written in assembler and widely used in AI.

    ReplyDelete
    Replies
    1. "Is that assembly code or machine code ?"

      It is machine code in an introduction to assembly language. Which just is machine code with easier-to-remember mnemonics.


      "In any case at some point someone must have programmed for a month or more to write a programming language like lisp that is (I think) written in assembler and widely used in AI."

      So what? I love Lisp! I wrote a Lisp interpreter. (In Java, by the way. Lisp is not necessarily written in anything -- you could use Basic to write a Lisp interpreter -- although early implementations probably were written in assembly. Once C was available, most implementors probably switched to that.) In any case, what does this have to do with believing overblown AI claims? For this to be meaningful, you'd have to dig up (literally, in some cases) the people who wrote the assembly code and see what they thought of these claims.

      Delete
    2. I just meant that i think many (if not most) AI enthusiasts are very well aware of the fundamentals of how computers work and wouldn't disagree that programs written in a higher level language suitable for AI are at one level still just 'a bunch of circuits performing that exact mechanical operations I set up for them to perform'.

      My guess is however that some anti-AI people would not accept that human intelligence might itself be just a bunch of circuits performing exact mechanical operations programmed by evolution.

      Delete
    3. In fact its almost a defining feature of AI that intelligence can (in principal) be produced from 'a bunch of circuits performing that exact mechanical operations I set up for them to perform', and it is its opponents who would deny such a thing is possible.

      Delete
    4. Yes, as you rightly point out, AI people are in a terrible muddle here.

      Delete
    5. "My guess is however that some anti-AI people would not accept that human intelligence might itself be just a bunch of circuits performing exact mechanical operations programmed by evolution."

      Well, yes, it would be dumb to accept that, because something mechanically carrying out an operation is BY DEFINITION not doing it intelligently!

      Delete
    6. "I just meant that i think many (if not most) AI enthusiasts are very well aware of the fundamentals of how computers work..."

      And just think suddenly something magic happens that makes them miraculously intelligent! I know, rob, I know: what superstition!

      Delete
    7. rob, you're like someone who loses money on every sale, but thinks if he can just sell enough, he'll be fine.

      Delete
    8. 'Well, yes, it would be dumb to accept that, because something mechanically carrying out an operation is BY DEFINITION not doing it intelligently!'

      Aren't you somewhat assuming your own conclusions with that definition ? AI is impossible because you have defined it as impossible.

      Delete
    9. 'And just think suddenly something magic happens that makes them miraculously intelligent!'

      There are a lot of things once attributed to miracles and magic that are now explained by modern science and most of them are ultimately explainable as complex sets of 'exact mechanical operations'. It seems odd (and a little bit superstitious) to deny the same could not be true of intelligence.

      Delete
    10. "There are a lot of things once attributed to miracles and magic that are now explained by modern science and most of them are ultimately explainable as complex sets of 'exact mechanical operations'."

      God, rob, you just want to continue to be thick as a brick, don't you? You are just spouting off boilerplate scientism. My point is that YOU and your friends are relying on a miracle. You have NO explanation as to why something purely mechanical should suddenly by a miracle "give rise" to consciousness. YOU are the siperstitious one: once some mechanism gets too complex for you too grasp, like a cargo-cult islander, you bow down and gasp, "Oooh, AI!" The philosophical explanations behind my view have been stated repeatedly for centuries: YOU just ignore them. Seems a little odd (and superstitious) that you just keep doing so!

      Delete
    11. "Aren't you somewhat assuming your own conclusions with that definition ? AI is impossible because you have defined it as impossible."

      rob, why do you continually ignore the fact, which I have stated AGAIN AND AGAIN in this blog, that I do not hold that "AI is impossible"? It interferes with your cult beliefs, perhaps?

      I point to that definition because it IS the standard meaning of those words. If the doctor taps my knee, and it jerks, we say, "That is not an intelligent repsonse: it is just mechanical." But if the doctor asks me, "Do you smoke?" and I am worried about losing my life insurance, and I lie and say "No," we say, "Gene THOUGHT it would be better to say no."

      YOU want to redefine the term intelligence. But you can't offer any justification for your redefinition. So you resort to this sophistry.

      Delete
    12. You made that statement in response to my statement:

      'My guess is however that some anti-AI people would not accept that human intelligence might itself be just a bunch of circuits performing exact mechanical operations programmed by evolution.'.

      I leave it to any fair minded reader to judge if your response is consistent with someone who believes AI is possible.

      I agree you have stated elsewhere you don't believe AI is impossible - but I think you made a statement here inconsistent with that view.

      Delete
    13. Of course it is consistent: *I* believe AI is possible. But if it ever arises, it will be because the AIs *stop behaving mechanically*, and start thinking for themselves. (Behaving mechanically == not thinking, as shown by my knee reflex example.)

      Delete
  2. Gene suppose it was announced tomorrow that scientists have completely analyzed the human brain, and that they are now able to predict human behavior with 100% certainty by analyzing neuron firings, detecting the sensory input the human body is going to receive, and anticipating how neurons will fire in response.

    If that really happened, how would your beliefs change in response to it? I think most people's reaction would be "I am still a conscious being making intelligent decisions, even if it is the case that at some level that I am not aware of, my behavior is reducible to a series of mechanical operations."

    ReplyDelete
    Replies
    1. "If that really happened, how would your beliefs change in response to it? "

      I would belive scientists are stupider than I thought they were.

      Delete
    2. Hey, and what about my $50?

      Delete
    3. 'I think most people's reaction would be "I am still a conscious being making intelligent decisions,"

      And so these scientists are full of it. THAT's what most people's reaction would be. And they would be absolutely correct.

      Science is an abstraction. It cannot inform us about the world of concrete reality.

      Delete
    4. Gene, you asked me the same thing in Bob's blog and I gave you an answer: "And concerning our bet, you never got back to me on what odds you wanted, so as far as I’m concerned we never actually agreed on a bet. Tell me if you feel differently.

      And in any case it would have been ten dollars, not 50, and it would have been payable to a charity, not you. Are you confusing me with someone else?"

      Delete
    5. Don't worry: I don't remember the details, and I forgave the debt of everyone else I bet with. So we are fine!

      Delete
    6. But $10?! You really weren't very confident, were you? I was willing to take bets up to $1000!

      Delete

An orgy

“The advancement of science and the rationality of politics are interwoven in a social process that, in the perspective of a more distant f...