Is the brain a reasonable AGI design?

Is the brain a reasonable AGI design?

Shane Legg seems to think so:  Tick, tock, tick, tock… BING.

Having dealt with computation, now we get to the algorithm side of things. One of the big things influencing me this year has been learning about how much we understand about how the brain works, in particular, how much we know that should be of interest to AGI designers. I won’t get into it all here, but suffice to say that just a brief outline of all this information would be a 20 page journal paper (there is currently a suggestion that I write such a paper next year with some Gatsby Unit neuroscientists, but for the time being I’ve got too many other things to attend to). At a high level what we are seeing in the brain is a fairly sensible looking AGI design. You’ve got hierarchical temporal abstraction formed for perception and action combined with more precise timing motor control, with an underlying system for reinforcement learning. The reinforcement learning system is essentially a type of temporal difference learning though unfortunately at the moment there is evidence in favour of actor-critic, Q-learning and also Sarsa type mechanisms — this picture should clear up in the next year or so. The system contains a long list of features that you might expect to see in a sophisticated reinforcement learner such as pseudo rewards for informative queues, inverse reward computations, uncertainty and environmental change modelling, dual model based and model free modes of operation, things to monitor context, it even seems to have mechanisms that reward the development of conceptual knowledge. When I ask leading experts in the field whether we will understand reinforcement learning in the human brain within ten years, the answer I get back is “yes, in fact we already have a pretty good idea how it works and our knowledge is developing rapidly.”

(emphasis added.) Shane is one of the leading AGI researchers out there. I tend, in general, to agree with his analysis and predictions.

By | 2017-06-01T14:16:47+00:00 December 25th, 2009|Machine Intelligence, Nanodot|3 Comments

About the Author:


  1. Kevembuangga December 25, 2009 at 7:00 am - Reply

    Is the brain a reasonable AGI design?

    Is a bird a reasonable plane design?

  2. Al Fin December 26, 2009 at 9:18 am - Reply

    Very interesting article. Thanks for pointing it out.

  3. J. Storrs Hall December 27, 2009 at 7:39 am - Reply

    Kev: You’re definitely better off understanding how a bird works when designing an airplane. The basic planform — wings and a tail — works well. The bird’s propulsion system, flapping, would lead to a very bumpy ride in a machine and thus propellors are a better choice for a vehicle. There are several other capabilities birds have that airplanes only approach and which would (will) be useful once our underlying technology becomes capable of providing them: taking off from a standing start without runways; flying through spaces crowded with limbs and other birds; landing on a perch. Reconfigurable wings for different flight regimes.
    Note that the brothers Wright spent a lot of time watching birds, which gave them the idea for their great innovation, wing-warping, which made workable planes possible.
    Given its “design goals”, a bird is an excellent design for a flying machine. Given different goals and a different implementation technology, we get airplanes. But we do get them.

Leave A Comment