I've been thinking about why it is relatively easy to evolve creatures that
can move pretty fast but so much more difficult to evolve a creature that
can find food. I've heard the idea put forward in this forum ( I think by
MacKo himself) that the smell sensors are too strong, but I'm not sure that
this is the key problem.
I'm wondering if instead the problem would be solved if the food balls could
be replaced with food gradients. A food gradient would have a center where
the energy was at its highest concentration, with diminishing concentration
farther away from the center. This would be a diffuse field at lest ten
times the radius of the current food balls.
The advantage of the graidients is that any move up the gradient would
bestow an advantge and any move down the gradient would bestow a detriment.
There can thus be many more increments of fitness than in the current
situation. Right now, a creature that moves half of the way toward a
foodball is no more fit than a creature that moves only a quarter of the
way.
Currently, selecting for speed instead of foodfinding is easier because in
selecting for velocity, there are many more oppurtunities for incremental
progress. A creature that moves a little bit faster is a little bit fitter.
But with food balls instead of the food gradients, a creature that moves a
little bit toward the food is not necessarily a little bit fitter for it.
The question might arise of how biologically realistic the food gradients
might be. I think they are quite realistic if we compare them to the
nutrients in solution that bacteria, euglena, or small animals depend on. It
is my understanding that studies of chemotaxis in, say, the nemotode worm c
elegans invovle nutrient gradients, not discrete pieces of nutrient.
--
___________________________________________
P E T E M A N D I K
Assistant Professor and
Associate Director, Cognitive Science Laboratory
Department of Philosophy
William Paterson University of New Jersey
265 Atrium Building
300 Pompton Road
Wayne, NJ 07470
(973)-720-2173
mandikp@wpunj.edu
http://www.wpunj.edu/cohss/philosophy/faculty/mandik
Re: Food gradients instead of food balls?
> I've been thinking about why it is relatively easy to evolve creatures that
> can move pretty fast but so much more difficult to evolve a creature that
> can find food. I've heard the idea put forward in this forum ( I think by
> MacKo himself) that the smell sensors are too strong, but I'm not sure that
> this is the key problem.
>
> I'm wondering if instead the problem would be solved if the food balls could
> be replaced with food gradients. (...)
>
> The question might arise of how biologically realistic the food gradients
> might be. I think they are quite realistic if we compare them to the
> nutrients in solution that bacteria, euglena, or small animals depend on. It
> is my understanding that studies of chemotaxis in, say, the nemotode worm c
> elegans invovle nutrient gradients, not discrete pieces of nutrient.
We are considering improving S receptor in v2 (not much time left!)
From my experience, the problem is not with the gradient
(S receptor already works in a 'fuzzy' way,
excitation=energy/distance^2), but with the fact
that a single S receptor cannot detect direction (only intensity).
It is really hard for evolution to construct more S receptors and
connect them in a meaningful way to derive direction information.
(There are some successful smellers handcrafted/evolved by advanced
Framsticks users though).
I think gradients would even make things worse. It is really hard
to find a good solution and be still biologically plausible.
Otherwise we would be talking about sight, which is directional.
MacKo
Re: Food gradients instead of food balls?
Hi MacKo,
thank you for your response.
> From my experience, the problem is not with the gradient
>(S receptor already works in a 'fuzzy' way,
>excitation=energy/distance^2),
Yes, but my suggestion would involve not changing S reception but instead
changing how ingestion works, for instance, making ingestion amount
=energy/distance^2
>. . .but with the fact
>that a single S receptor cannot detect direction (only intensity).
>It is really hard for evolution to construct more S receptors and
>connect them in a meaningful way to derive direction information.
that is a good point, it is hard to do. but i wonder how best to compare it
to making a fast creature. The minimal network to detect smell direction
seems to involve 2 sensors and one muscle. That is much less complex than
many of the networks that have evolved from scratch by selecting for
velocity. Note too that with a single smell sensor, that can at least slow
down a creature to absorb more food. A single smell sensor encodes
proximity information even though it takes two sensors to encode direction
information.
Cheers,
Pete
Re: Food gradients instead of food balls?
> A single smell sensor encodes
> proximity information even though it takes two sensors to encode direction
> information.
Could you have a single smell sensor on a long-ish limb, and waggle it
about? This single receptor could thus build up a picture of the suroundings
if the way it tracked through them was significantly shorter period than
that of the motion. A bit like a radar sweep.
Not sure how/if this translates into FS, though. I'll still be fiddling
about with blind sticks for a while yet.
Phil
Re: Food gradients instead of food balls?
>Could you have a single smell sensor on a long-ish limb, and waggle it
>about? This single receptor could thus build up a picture of the
suroundings
>if the way it tracked through them was significantly shorter period than
>that of the motion. A bit like a radar sweep.
>
I have thought of this one before and tried it. No success yet. This is
theoretically possible, but I haven't got a Fram to do it yet. You need a
network that can compute direction from the single smell sensor plus
feedback from the waggle-muscle. The basic gist of how directional
information is encoded will be something like:
Smell high + muscle right = food right,
Smell low + muscle right = food left,
Smell high + muslce left = food left,
Smell low + muscle right = food right
The basic archtiecture of the creature's nervous system would be: (1) a
central patter generator (CPG) for foward locomoton; (2) a seperate CPG for
right-left sensor waggle; (3) an orientation network that had as output a
steering muscle and had as input the smell sensor and feedback from the
waggle muscle.
Cheers,
Pete