Re: poly: Software for superintelligence

From: Anders Sandberg <asa@nada.kth.se>
Date: Fri Jan 23 1998 - 06:37:28 PST

"Nick Bostrom" <bostrom@ndirect.co.uk> writes:

> Ok, now I see what you mean. Yes, it is not that the time it takes
> for one cycle of weight updating scales differently for Backprop vs.
> Hebb, the important difference is rather that Backprop is supervised
> and Hebb's rule isn't. So backprop requires an evaluation function of
> behavioural output on the neuronal level, which gives signed error
> values for the activation level of all our output (motor?) neurons.
> How do we write down such a function?

Exactly. One could use reinforcement learning, where inputs determine
if it was a good or bad response, but there is still the "credit
allocation problem" - which synapses to update, and how to spread the
correction across the brain.

> The Hebb rule doesn't require us to define an error function. The
> difficulty is that it is not clear that it is sufficient to make the
> network behave as an intelligent human adult.

We'll see about that. Besides, global input can be useful to store
desirable behaviors more strongly or create inhibition against bad
behaviors (cf. aversion learning).

> I have a scheme for how the brain might store complex representations
> in long-term memory, after a one-shot presentation, using only
> Hebbian learning. I'm not sure this is the right place to discuss it,
> but perhaps someone could recommend an appropriate forum for such
> issues?

At least I would love to hear it, since my field of research is memory
consolidation using Hebbian learning right now.

-- 
-----------------------------------------------------------------------
Anders Sandberg                                      Towards Ascension!
asa@nada.kth.se                            http://www.nada.kth.se/~asa/
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y
Received on Fri Jan 23 14:40:57 1998

This archive was generated by hypermail 2.1.8 : Tue Mar 07 2006 - 14:45:29 PST