TED Conversations

Dan McNe

This conversation is closed. Start a new conversation
or join one »

One of the important points is that the feedback from our sensory system is noisy, I haven't seen this implemented in many A-Life programs.

Of the popular A-Life programs I've seen (avidia, Polyworld, Framsticks and breve), the focus of the program is often on developing some sort of adaptive movement (depending on the fitness function). Polyworld is only concerned with gross movement, but (more naturally) uses reproduction as the fitness function. Framsticks and breve allow multipart bodies and finer movement, but, to my knowledge, have "non-reproductive" fitness functions. Is it possible to combine the two (programs or neural nets) into one.

On a somewhat unrelated note, it may be that we have brains partly because we need to move, but brains also seem to be very good at pattern matching (both spatial and temporal -- and in ours abstract). Since there are very simple creatures which move but have no brains, I would suggest that "patterned/adaptive movement" is the reason for brains.

Related Talks:
Topics: artificial intelligence artificial life brain evolution technology
0
Share:
progress indicator

  • Steve C

    • 0
    Dec 7 2011: Don't Mark Tilden's BEAM robots basically use analog "noise"? That seems to work wonderfully on the physical level. I don't see why it wouldn't help in AI. Wouldn't that introduce a ground-level "reason" for AI to germinate or foliate?
    Great idea! If you wouldn't have said that, I probably would've continued to assume that noise was just "in the way."

  • Zachary Allinson

    • 0
    Nov 14 2011: Well it may not directly account for the noise but I would assume they are generally developed with an assumed amount of variance somewhat replicating the issue of noise. Just an Idea
  • thumb

    Tim Thornton

    • 0
    Nov 11 2011: Nature is noisy and its creatures must interpret quickly to survive. So there is filtering, parsing, and patterning of the inputs into a response. This response can be movement, chemical, or electrical. Brain is a clump of connections that have a fairly simple set of mechanisms that facilitate interpreting inputs quickly to outputs.
    Current neural nets are too simplistic to achieve the richness of our brains. Different sections of our brain have different tactical ambiances causing the necessary distinctive responses from that section. Our brains will be completely capable to comprehend its function.
Showing all 3 comments