[LOCKSUIT]

Me and goaty are internet  pals, I was just being joke-y.

[ivan.moony]

somewhere within the connectome is the basic structure of AGI. If we map the connectome of other creatures an do an invariable analysis, we should be able to isolate it.

I wouldn't be certain of that if I were you. We don't even know for sure that there *is* a "basic structure of AGI." Maybe each of the various skills or functions that go into human-level intelligence needs its own algorithm for best results. A worm would have some of those, but not all.

If you did an invariate analysis across species, I suspect that what you'd isolate is the part of each species' brain that implements worm-level faculties. Some of our [human] faculties might be so different on a conceptual level that it wouldn't matter how much you scaled up a worm brain, you'd never, ever get to them.

An algorithm for thinking, or a structure that implements that algorithm, is a form of information. Think of the worm connectome as having fewer "bits" to store information. You won't spontaneously generate the additional information in the human connectome just by making the worm connectome bigger. (You might say "but all that additional room in the human connectome is just extra memory that's filled in by learning, or the same thing replicated for additional processing power!" And you *could* be right, but again, I wouldn't bank on it.)

Maybe if we virtually breed a bunch of these worms, and mate them over X millions of generations, maybe then it would end with some intelligent species, who knows?

[HS]

There's a thought! Applying a genetic algorithm to a natural structure.

[goaty]

Ooof!  Locksuit, that was rude.

Goaty, I'm not sure what you're trying to say in the first post. A computer with some memory, that is connected to sensors and actuators -- that's a vague enough description that it would apply to most robot brains out there. Do you mean that this is the minimum set of components, i.e. you can't have an effective brain if you are missing one of those four things?

I don't know what definition of "computer" you're using here, though since you mention the word "program" maybe you mean at least a microcontroller.  It's possible to go simpler than that.  I have a robot whose brain is a 5-state synchronous state machine made out of three flip-flops and some gates.  You don't get a whole lot out of five states ... it can back up and turn when it hits an obstacle, and that's the extent of the behavior.  But it still counts as a brain.

I've heard of probably even simpler robots that only use analog circuits ... https://en.wikipedia.org/wiki/BEAM_robotics

I guess your right,  that it applies to all robots whether they play basketball (whilst dribbling the ball properly in a human form)- or just hone in toward a light (with a pair of bidirectional wheels)- one is a lot more successful than the other, but they both fit into the sensor -> program -> motors thing.

So if you ever bring it up,  its actually useless to,  because it doesn't teach the people wasting their time reading it anything.

[WriterOfMinds]

Maybe if we virtually breed a bunch of these worms, and mate them over X millions of generations, maybe then it would end with some intelligent species, who knows?

Genetic algorithms don't just give you things for free. There'd be no pressure to evolve intelligence unless intelligence made the virtual creatures better at surviving. So, you'd need to put effort into designing the fitness function such that it would optimize for intelligence. And you might find, when you were done, that you'd put as much work into your fitness function as you would have put into just writing the intelligence directly. In this case, you'd be injecting information into the system via the fitness function.

[goaty]

Maybe if we virtually breed a bunch of these worms, and mate them over X millions of generations, maybe then it would end with some intelligent species, who knows?

Genetic algorithms don't just give you things for free. There'd be no pressure to evolve intelligence unless intelligence made the virtual creatures better at surviving. So, you'd need to put effort into designing the fitness function such that it would optimize for intelligence. And you might find, when you were done, that you'd put as much work into your fitness function as you would have put into just writing the intelligence directly. In this case, you'd be injecting information into the system via the fitness function.

Yes.  writing the intelligence directly.   That's what I think is easier than making it develop it itself.  But I actually recommend 1/2 and 1/2 to be exact.

But one more thing,  if you solve the application at hand in a simple monotonous way, the problem is still solved,  and the model u built to the solve the problem is simpler and has less parts, but the job is done the same.
..

[ivan.moony]

Maybe if we virtually breed a bunch of these worms, and mate them over X millions of generations, maybe then it would end with some intelligent species, who knows?

Genetic algorithms don't just give you things for free. There'd be no pressure to evolve intelligence unless intelligence made the virtual creatures better at surviving. So, you'd need to put effort into designing the fitness function such that it would optimize for intelligence. And you might find, when you were done, that you'd put as much work into your fitness function as you would have put into just writing the intelligence directly. In this case, you'd be injecting information into the system via the fitness function.

This entire planet is a fitness function that leads to... I'm still having problems about this one.

[goaty]

Maybe if we virtually breed a bunch of these worms, and mate them over X millions of generations, maybe then it would end with some intelligent species, who knows?

Genetic algorithms don't just give you things for free. There'd be no pressure to evolve intelligence unless intelligence made the virtual creatures better at surviving. So, you'd need to put effort into designing the fitness function such that it would optimize for intelligence. And you might find, when you were done, that you'd put as much work into your fitness function as you would have put into just writing the intelligence directly. In this case, you'd be injecting information into the system via the fitness function.

This entire planet is a fitness function that leads to... I'm still having problems about this one.

the fitness function could be a Boolean model of an application for your robot.

FOR EXAMPLE.

 * washing dishes
 * sweeping the back porch
 * cooking dinner
 * doing the clothes washing
 * putting its head through windows

these can be "fitness functions" too.

[HS]

Maybe if we virtually breed a bunch of these worms, and mate them over X millions of generations, maybe then it would end with some intelligent species, who knows?

Genetic algorithms don't just give you things for free. There'd be no pressure to evolve intelligence unless intelligence made the virtual creatures better at surviving. So, you'd need to put effort into designing the fitness function such that it would optimize for intelligence. And you might find, when you were done, that you'd put as much work into your fitness function as you would have put into just writing the intelligence directly. In this case, you'd be injecting information into the system via the fitness function.

This entire planet is a fitness function that leads to... I'm still having problems about this one.

the fitness function could be a Boolean model of an application for your robot.

FOR EXAMPLE.

 * washing dishes
 * sweeping the back porch
 * cooking dinner
 * doing the clothes washing
 * putting its head through windows

these can be "fitness functions" too.

I think just reward tool use. Give lots of opportunities to get started on this path. It should lead to our type of intelligence.

[goaty]

Yeh. tool use.  but to what end?  to just pick it up?

[HS]

As a feasible fitness function of a genetic algorithm which could lead to the development of recognizable AGI, when jump-starting the process with a naturally occurring nervous system.