Man observes birds, concludes flight is possible. Builds various machines, eventually gets into the air.

Do we know the inner workings of the bird? - I don't think so. Do we need to? - to fly, no.

This is an interesting statement. Many many different interpretations:

1. Flight, as a category, is theoretically diverse. There are many ways to achieve flight.
2. The type of energy harnessed to drive a flying artifice is independent of flight.
3. Observation is either the only (or at the very least an) important motivator of invention.
4. Without an existing example, the invention of flight would have been impossible or unlikely.

All of these are valuable issues to be explored. As applied to the program where by one intelligent thing goes about attempting to build another intelligent thing, that goes double.

But Kevin, your logic is seriously out to lunch if you were trying to argue against Lex's point that the building of a synthetic intelligence requires much more than a familiarity with the external behaviors of an intelligent system. In Kevin's example he says it would be hard to build an internal combustion engine if all you knew was what you could observe as a car (or plane) drove by.

Lex is positing the argument that intelligence emerges only from entities that have certain internal mechanisms (in this case, some mechanism that constantly self-adjusts to some internally determined mean (homeostasis). To lex, this is to intelligence what neutral buoyancy is to flight – has to be there or you don't have intelligence.

Kevin, your argument seems to be that flight itself is sufficient proof of concept – that flight implies neutral buoyancy and all other essential elements of flight. You argue that it is not necessary to understand intelligence to build it – just make enough guesses and try enough configurations and wait until something starts thinking. When it does, just declare it intelligent and that is that.

Let's ignore the fact that Kevin's usual approach is far more deliberate than this. Lets just assume that an intelligent thing has arisen from blind combinatorial iterations. In front of us is a thinking thing that wasn't thinking moments before. We know it is intelligent. What we don't know is how or why. What have we accomplished? How is this newly intelligent thing any more informative as example than any of the intelligences that exist in the animal kingdom? Now what?

Is the goal to understand intelligence or is the goal simply to extend it beyond human limits? Evolutionary processes seem well capable of creating things beyond the imagination of the things that built them. But the locus of control shifts forward always to the most complex or most contemporary system. Are we OK with that? Is there any way to avoid the passing of the baton? Can we maintain authority in a system in which more complex intelligences co-exist with us?

What I think Lex is arguing, is that the arrival of an intelligence edifice is only possible of some part of that edifice acts the part of homeostasis engine.

I can never tell what Kevin is trying to say.

What is indisputably true of any system is that it can not be a system at all unless its form results in behavior that repeats at some interval. A cow has to eat periodically. A cow has to sleep periodically. A cow has to procreate periodically. Any system capable programmed into a computer has to operate by some set of looped conditional logic.

Though I am sure Lex has a narrower definition of the term, all of these examples, are analogues of what Lex labels "homeostasis".

What interests me is the complete list of all of the essential elements or subsystems that must be present in any and all intelligent systems, of which homeostasis might be but a single example.

Randall Reetz

But Kevin, your logic is seriously out to lunch if you were trying to argue against Lex's point that the building of a synthetic intelligence requires much more than a familiarity with the external behaviors of an intelligent system. In Kevin's example he says it would be hard to build an internal combustion engine if all you knew was what you could observe as a car (or plane) drove by.

I'm merely pointing out that humans are relatively good at coming up with ways of doing things (better than nature) on just the knowledge that the result is possible.

Kevin, your argument seems to be that flight itself is sufficient proof of concept – that flight implies neutral buoyancy and all other essential elements of flight. You argue that it is not necessary to understand intelligence to build it – just make enough guesses and try enough configurations and wait until something starts thinking. When it does, just declare it intelligent and that is that.

That would be an evolutionary approach - as used by nature (QED).

Let's ignore the fact that Kevin's usual approach is far more deliberate than this. Lets just assume that an intelligent thing has arisen from blind combinatorial iterations. In front of us is a thinking thing that wasn't thinking moments before. We know it is intelligent. What we don't know is how or why. What have we accomplished? How is this newly intelligent thing any more informative as example than any of the intelligences that exist in the animal kingdom? Now what?

If it's smarter, it might be able to answer that question for you. "why" doesn't really matter, "how" might matter if you need repeatability in the process. The important thing (IMO) is understanding the "evolutionary" process that will create smart software - the nuts & bolts level.

...
Though I am sure Lex has a narrower definition of the term, all of these examples, are analogues of what Lex labels "homeostasis".

What interests me is the complete list of all of the essential elements or subsystems that must be present in any and all intelligent systems, of which homeostasis might be but a single example.

Randall Reetz

Homeostasis has to do with maintaining equilibrium, intelligence (to me) seems to have more to do with reaching longer term goals (that were never part of the equilibrium), but most systems like to reduce their costs by reuse i.e. if you can use your intelligence to aid the existing homeostatic processes you are going to do that. So biological systems tend to have intertwined processes that engineers prefer to avoid if possible when designing systems deliberately.

I think it's a mistake to think that one has to understand the processes that are entwined with intelligence in the human brain to understand the basics of intelligence.

I would also say that intelligence is to a large extent just "enhanced decision making", and as such is just observation, prediction and reaction.