Ask HN: Is Linus creating AI Life?
15 comments
There is no artificial intelligence, just a lot of humans doing experiments and keeping the best results in the kernel. I think that's plain-old "intelligence".
Maybe the kernel is just a basis for some sort of massive ai-life simulation?
The kernel isn't the emergent life, the community is. The kernel is just the/(a) side effect.
The kernel isn't the emergent life, the community is. The kernel is just the/(a) side effect.
Right. The kernel is an extended phenotype.
He said he believes in guided evolution, which I agree is the best way to do things. But really it's the only way, because if you're not doing that then at some point you must have lied to yourself when your vision misaligned with results.
I think what he's going at is trying to design things logically and rationally, just as evolution does. That is, seeing how the world designs itself and modeling his own methods to it with a little human guidance to speed the what would be random process up.
I think what he's going at is trying to design things logically and rationally, just as evolution does. That is, seeing how the world designs itself and modeling his own methods to it with a little human guidance to speed the what would be random process up.
Evolution doesn't design anything. It just sometimes seems that way from our (builder-mind) perspective.
Well I guess it's how you define design. It designs in the sense that it produces something that simply works. Evolution/survival of the fittest is absolute and all human design is simply to imitate it's efficacy (its efficacy is perfect) in less time.
I think most people would agree that "designing" something means that its construction/development is planned. There's no planning in evolution.
Well our argument is about semantics. If "designing" by your definition requires planning, then so be it.
The point is that when we design we try to create something that works. Evolution is basically the products produced by reality that work, and this is by definition. So in essence we try to speed up evolution by designing. We try to predict what works before it gets there by random luck. Obviously the simplest algorithm is to try random iterations that are completely independent of past ones, but this is essentially evolution. So what Linus is in favor of is guided evolution, that is instead of trying random iterations, we try to predict the next working iteration based on past knowledge, and identifying any patterns. Of course everyone probably does this. Unfortunately most people break out of this if they don't understand this, and may be blinded by ego. So when you can't admit you're doing something wrong, you're no longer designing through guided evolution, you're designing by complete random chance at best.
In any case evolution is the absolute process that design tries to mimick because its results are by definition what we are trying to achieve.
The point is that when we design we try to create something that works. Evolution is basically the products produced by reality that work, and this is by definition. So in essence we try to speed up evolution by designing. We try to predict what works before it gets there by random luck. Obviously the simplest algorithm is to try random iterations that are completely independent of past ones, but this is essentially evolution. So what Linus is in favor of is guided evolution, that is instead of trying random iterations, we try to predict the next working iteration based on past knowledge, and identifying any patterns. Of course everyone probably does this. Unfortunately most people break out of this if they don't understand this, and may be blinded by ego. So when you can't admit you're doing something wrong, you're no longer designing through guided evolution, you're designing by complete random chance at best.
In any case evolution is the absolute process that design tries to mimick because its results are by definition what we are trying to achieve.
Does your definition of "designing" not include the concept of planning? I can see how "building" could forego planning (and result in poor products), but "design" implies planning I would argue.
Some comments about evolution:
1) Evolution very often has results that, if it were the product of design, would be considered very poor design
2) Evolution is not random luck. It is anything but random. Mutation is random, but evolution by natural selection is not.
3) Things do not evolve unless:
Some comments about evolution:
1) Evolution very often has results that, if it were the product of design, would be considered very poor design
2) Evolution is not random luck. It is anything but random. Mutation is random, but evolution by natural selection is not.
3) Things do not evolve unless:
- there's replication
- there are mutations that natural selection can "see"
- there is pressure for certain kinds of mutations to be favoredIn response to your bullet points:
1) Evolution always has results that will be considered poor design. Every time an evolved "design" becomes obsolete with respect to its surroundings, then it becomes a poor adaptation, at which point it must evolve again. But because this process is continuous, we consider this whole continuous process as evolution. As time increases, evolution will always force something to adjust to its environment and become perfect with it, therefore evolution as a whole can be considered absolute and perfect.
2) Evolution as a whole is absolute, so obviously it's not luck. The adaptations and changes before the perfect "design" arrives are determined by luck as you say.
When I say design I just mean the product of some systematic process. Planning is systematic, but so is evolution.
1) Evolution always has results that will be considered poor design. Every time an evolved "design" becomes obsolete with respect to its surroundings, then it becomes a poor adaptation, at which point it must evolve again. But because this process is continuous, we consider this whole continuous process as evolution. As time increases, evolution will always force something to adjust to its environment and become perfect with it, therefore evolution as a whole can be considered absolute and perfect.
2) Evolution as a whole is absolute, so obviously it's not luck. The adaptations and changes before the perfect "design" arrives are determined by luck as you say.
When I say design I just mean the product of some systematic process. Planning is systematic, but so is evolution.
I disagree. Evolution designed all life forms. Design and implementation are not separate for evolution, but that doesn't mean that design did not occur. Humans did not do (Edit: ..most of) the designing, but that also does not mean that design did not occur.
Living organisms are the product of evolution by natural selection. BUT, there is no forethought in the evolution process; i.e. no design. The way living organisms turn out is not according to a plan, i.e. it is not designed (blindly or otherwise).
sounds like he's also been reading wolfram's nks. the whole book is about how complexity is created. and wolfram suggests that humans wouldn't be able to purposefully create something more complicated than ourselves (though leaving open the possibility of doing it by semi-accident).
is _____ creating amazing technological breakthrough?
I'd prefer not to see question mark headlines. they're really easily abused.
I'd prefer not to see question mark headlines. they're really easily abused.
Perhaps "Hive Mind" is a better term than AI.
"I'm deadly serious: we humans have _never_ been able to replicate something more complicated than what we ourselves are, yet natural selection did it without even thinking.
Don't underestimate the power of survival of the fittest."
This intrigued me so I looked at his blog http://torvalds-family.blogspot.com where he comments a few times on what he is reading. I will list some of the books he mentions there:
Phantoms in the Brain, The Brain that Changes Itself, Why Evolution is True, Genesis: The Scientific Quest for Life's Origins
This quote is interesting:
"Usually, I tend to read about genetics or similar (that is, when I read anything serious to begin with, which tends to be less than 10% of the time). This one is obviously related, but about the processes that came before it all began. And it also gives more of a look into the issues faced by <BOLD> somebody </BOLD> trying to do experiments in the area."
From the above comments his viewpoint possibly falls into the artificial life camp that tries to imitate traditional biology by recreating biological phenomena (http://en.wikipedia.org/wiki/Artificial_life).
What do you think of this idea? He is well-noted for his programming skills and general ability. Maybe the kernel is just a basis for some sort of massive ai-life simulation?