Re: How much intelligence?

Curt Welch wrote:

...

If we can't find the machine, and the learning
algorithm to shape it, then we can't do it.
But that's not why people reject the idea that
Language can be learned by reinforcement.
They reject it because they simply don't
understand the true power of reinforcement
learning to shape complex behavior. They just
don't understand reinforcement learning.

Reinforcement learning is well understood.

The processes involved in real reinforcement learning (perhaps that's not
what you are talking about?), as currently understood are so vast that no
one person is an expert in every area. Further, no one involved in the
science will say that these processes are understood in any kind of
complete
sense - that's why there is still a lot of ongoing research. Perhaps you
meant that the silly toy misrepresentations of these processes are well
understood by those that misrepresented the processes in the first place.

JC: Perhaps you had better explain all this to Curt? Is his version
a silly toy misrepresentation? Implied in his statement that they
just don't understand reinforcement learning is that he does.
What area of RL is he an expert in?



GS: You are the one that said "reinforcement learning was well understood."
The implication is, at least as I read it, that it is well understood and
obviously insufficient to explain behavior and, thus, insufficient to
produce "intelligence." I pointed out the ways in which your statement was
false, on the one and, and trivial on the other. There are still big
questions about the basic processes (but even so, it is clear that they can
account for complex behavior). Again, maybe you meant that the AI field
called RL is well understood (with the implication that it is well
understood that this well-understood endeavor cannot explain
"intelligence"). The AI field AI may be well understood, but its
practitioners don't have much of a clue about what is actually known about
the basic processes. They create a caricature of reinforcement processes and
then when they fail to get "intelligence" we are told that reinforcement
processes are insufficient.



As to Curt, his strength is that he intuitively grasps the power of
conditioning and he sees through some of the nonsense puked out by its
critics. But he is not an expert in the science, and much of what he says is
frankly mentalistic or just plain silly ("rocks are conscious," "the animal
reinforces itself," "we respond to the contents of our STM" etc. etc. etc.
etc. etc. etc. etc. etc.). I like Curt, but my opinion is that he needs to
take a step back, read some books and start again in a few years.



"JGCASEY" <jgkjcasey@xxxxxxxxxxxx> wrote in message
news:1141912902.328515.291540@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx



.

Relevant Pages

  • Re: Reinforce learn this
    ... controller, which works as well for the robot, that our brain works for us. ... give us a few years of exposure to an environment with ... Only some basic learning hardware was there at ... This is where reinforcement learning comes in. ...
    (comp.ai.philosophy)
  • Re: Ben G on reinforcement-learning and the wirehead problem
    ... And keep in mind that reinforcement learning is not ex nihilo (as you ... which actions yield the most reward by trying them. ... such simple conditioning is not the same as human knowledge. ...
    (comp.ai.philosophy)
  • Re: How much intelligence?
    ... If we can't find the machine, and the learning ... Reinforcement learning is well understood. ... Perhaps you had better explain all this to Curt? ... obviously insufficient to explain behavior and, thus, insufficient to ...
    (comp.ai.philosophy)
  • Re: What is innate and how can we determine it?
    ... learning", and "Learning is all that is important", ... It doesn't confuse me, now, but why would you want to confuse ... of reinforcement learning. ... network already answered all those questions. ...
    (comp.ai.philosophy)
  • Re: Energy constraints was Re: Androids & Strong AI in the Future
    ... of "situated" from nothing more than undefined streams ... and attempts have been made to build such machines. ... And to the extent that "learning" can be seen as "learning ... I'm talking about generic reinforcement learning algorithms that have 3 ...
    (comp.ai.philosophy)