Dynamic, lifelike AI

[DXnobodyX]

Enieublis McShlaggrmasker

DXnobodyX

Im not advocating the idea or anything, just pointed out that its already been done.

But... they're completely different. My machine isn't made out of human brains or whatever that thing was. I think the way I summarized my AI just happened to be similar in some way to the story of biblical magi, which that computer was based off of.

Curiosity isn't a human trait?
Self preservation isn't an animal trait (which humans are)?
Communication/empathy isn't a human trait?

They aren't made of human brains they are computers modelled around different traits of a humans brain specifically a women. The computer only draws its name of MAGI because of its intelligence being likened to the wise men, its AI function is not similar to the biblical magi.

[Shokushu]

Enieublis McShlaggrmasker

Shokushu

Enieublis McShlaggrmasker

Shokushu

To get to the point you can call it an AI in the first place you've got to work out consciousness which implies that it is doing more than just giving a canned response to environmental stimulus like is suggested here.

Actually I don't believe that consciousness is necessary for Artificial Intelligence. AI is defined as "the theory and development of computer systems able to perform tasks normally requiring human intelligence"(The New Oxford Dictionary of English, 1998.). Those computer systems do not have to perform all such tasks including consciousness. (if conscious is a task requiring intelligence)

I didn't say that AI's need it, just that you need to know how consciousness works to make an AI. Sort of set it as a milestone in understanding how humans think, as it would be difficult to grasp enough about our minds without having delved into that particular question. It's rather central to the issue so to speak, though not necessarily something we'd want to take on the burden of being creators of.

Plus it's an easy goal to understand sort of what the endpoint looks like and what hurdles we've got to clear to get closer to it.

OK, but you don't need to know how consciousness works to make an AI, you just need to know how people solve problems and create an intelligent agent that mimics that problem solving procedure. I'm just adding a little to the definition I found originally.

Consciousness may be related to the ethics issue, but I'm really not sure how related it is to Artificial Intelligence. I'm also not sure whether or not we would need to mimic conscious entities to find out what consciousness is, or if that would even help.

In fact, since consciousness is hard to define and since I have no idea how close we are to understanding it, I think a better basis for determining ethics is whether something is alive. Oxford dictionary defines life as: "the condition that distinguishes animals and plants from inorganic matter, including the capacity for growth, reproduction, functional activity, and continual change preceding death". The AI I described would have all of those capacities.

Teasing apart the various ways that we find strange connections between ideas is tantamount to understanding consciousness.

Life is a poor place to draw the line though as nobody bats an eyelash when you stab a cockroach in the brain or grind up algae into a paste. Those are both alive but fall short of the threshold that we actually care about.

Ryu Kei Shou Kawazu

If I put a fridge magnet on my head thatd be a little ridiculous lol.

Well fridge magnets are basically a bunch of bar magnets that are facing each other in sheet form. They're even weaker than the magnets you'd get from normally using those materials in that their magnetic field is extremely shallow. No way any number of those would even penetrate past the skin on your head.

Enieublis McShlaggrmasker

DXnobodyX

Im not advocating the idea or anything, just pointed out that its already been done.

But... they're completely different. My machine isn't made out of human brains or whatever that thing was. I think the way I summarized my AI just happened to be similar in some way to the story of biblical magi, which that computer was based off of.

Well both of them came from a very rudimentary decomposition of what it takes to make a thinking creature. You went with what is well known among lay people in the public that have any interest in these things so of course it's going to be fairly similar.

Getting at the latest peer review articles for neurology would both be difficult because those things aren't free and because they tend to be written at a technical level that is impenetrable to people that haven't already gone through a great deal of textbook study- but if you want any chance of coming up with ideas that haven't been put out there decades ago then that's probably the only way to do it.


For something a little more manageable you could look up news stories and try to become an autodidact from that. Science reporting tends to be really bad though so you'll accumulate lots of misinformation from when either reporters didn't care to understand the studies that were done or the scientists didn't care to explain it in an effective way. Still, it tends to give you at least a little perspective that the old school robotics thinkers lacked.

[Enieublis McShlaggrmasker]

DXnobodyX

Enieublis McShlaggrmasker

DXnobodyX

Im not advocating the idea or anything, just pointed out that its already been done.

But... they're completely different. My machine isn't made out of human brains or whatever that thing was. I think the way I summarized my AI just happened to be similar in some way to the story of biblical magi, which that computer was based off of.

Curiosity isn't a human trait?
Self preservation isn't an animal trait (which humans are)?
Communication/empathy isn't a human trait?

They aren't made of human brains they are computers modelled around different traits of a humans brain specifically a women. The computer only draws its name of MAGI because of its intelligence being likened to the wise men, its AI function is not similar to the biblical magi.

I never said that those weren't human traits, I meant that was where the similarity ended. I'm also pretty sure I saw a brain with wires being connected to it in that anime episode you linked.

Shokushu

Teasing apart the various ways that we find strange connections between ideas is tantamount to understanding consciousness.

Life is a poor place to draw the line though as nobody bats an eyelash when you stab a cockroach in the brain or grind up algae into a paste. Those are both alive but fall short of the threshold that we actually care about.

Actually I think the reason some people don't care about some forms of life is because they don't view them as conscious or similar to them. I think what we care about is the worst place to draw the line because it depends completely on whatever random people care about. I wouldn't be surprised if some people believed that humans were the only conscious beings on the planet. Bringing Nazi's into the argument, I think many Nazi's believed Jews weren't conscious either.

It might be that what "we actually care about" is just what is most convenient for us to care about at the time.

Shokushu

Enieublis McShlaggrmasker

DXnobodyX

Im not advocating the idea or anything, just pointed out that its already been done.

But... they're completely different. My machine isn't made out of human brains or whatever that thing was. I think the way I summarized my AI just happened to be similar in some way to the story of biblical magi, which that computer was based off of.

Well both of them came from a very rudimentary decomposition of what it takes to make a thinking creature. You went with what is well known among lay people in the public that have any interest in these things so of course it's going to be fairly similar.

Getting at the latest peer review articles for neurology would both be difficult because those things aren't free and because they tend to be written at a technical level that is impenetrable to people that haven't already gone through a great deal of textbook study- but if you want any chance of coming up with ideas that haven't been put out there decades ago then that's probably the only way to do it.


For something a little more manageable you could look up news stories and try to become an autodidact from that. Science reporting tends to be really bad though so you'll accumulate lots of misinformation from when either reporters didn't care to understand the studies that were done or the scientists didn't care to explain it in an effective way. Still, it tends to give you at least a little perspective that the old school robotics thinkers lacked.

My AI wasn't designed from a rudimentary decomposition of psychology, it was summarized with psychological terms because they seamed somewhat close and could imply about what each part did.


My thought process:

Creating an AI based completely around self defense would leave me with a boring program that does almost nothing unless it is provoked constantly. So I get it to dislike similar situations, this means having a database of the environment around the AI and of how that environment works, and designing the AI to constantly expand its database, that is a bit more than "curiosity", but it seamed like a good word for summarizing that system. However, those AIs would constantly kill themselves, so I would need to add back in the self defense system and give it access to the database(this isn't always necessary for a game AI though, when you can have multiple lives).

Now I have a pretty good AI with that, but if I have multiple of these around, then they wouldn't be as efficient as possible since each AI would have to go through the full problem solving process to do anything. So I thought of a way to get them to learn from each other. Having an object defining the current AI (based on previous outputs) in the database and comparing that object with any object being observed seamed like a good way to improve efficiency. Empathy seamed like a good label for that one. Communication would arise with the combination of this and one of the other systems.

Then, from this forum, I heard the term collective unconscious, which I think is nonsense, but it gave me an idea. When there are a massive amount of these things together, it is not as efficient as possible again, since some groups are doing the same things as other groups. So one AI would be designed to regroup the other AIs to improve efficiency. That part isn't really like anything I know of. (I'm not sure if I could call it a president)

Then those president AIs could be applied on top of each other in a hierarchical manner.


For this, each AI would need: Databases, Compiling, De-compiling, functions, objects, contextual data comparison, and performance testing.


It would work with just the curiosity system, but it would be slower in most cases.

[Ryu Kei Shou Kawazu]

DXnobodyX

Ryu Kei Shou Kawazu

DXnobodyX

Enieublis McShlaggrmasker

DXnobodyX

The point was its already been thought of before.

Then the point is wrong. Just because my AI and another AI have three parts doesn't mean they're exactly the same. If you can link to a page that describes it in more detail then I might agree with you. But so far the only information I found on that fictional computer that actually means anything is that it has three parts.

curiosity - scientist
mother - communication/empathy
self preservation - women

Better that you just watch the episode
http://www.dubbedepisodes.com/neon-genesis-evangelion-episode-13-english-dub/

Dont show him that lolll.
Its messed up.
But yeah I understand the three parts, its in alot of psychological theories and such. Id ego superego, concious unconcious subconcious collective-unconcious lol.

Im not advocating the idea or anything, just pointed out that its already been done.

Not that I meant the idea was messed, but the episode lol.

[DXnobodyX]

Ryu Kei Shou Kawazu

DXnobodyX

Ryu Kei Shou Kawazu

DXnobodyX

Enieublis McShlaggrmasker

DXnobodyX

The point was its already been thought of before.

Then the point is wrong. Just because my AI and another AI have three parts doesn't mean they're exactly the same. If you can link to a page that describes it in more detail then I might agree with you. But so far the only information I found on that fictional computer that actually means anything is that it has three parts.

curiosity - scientist
mother - communication/empathy
self preservation - women

Better that you just watch the episode
http://www.dubbedepisodes.com/neon-genesis-evangelion-episode-13-english-dub/

Dont show him that lolll.
Its messed up.
But yeah I understand the three parts, its in alot of psychological theories and such. Id ego superego, concious unconcious subconcious collective-unconcious lol.

Im not advocating the idea or anything, just pointed out that its already been done.

Not that I meant the idea was messed, but the episode lol.

Both are to me

[Ryu Kei Shou Kawazu]

Shokushu

Enieublis McShlaggrmasker

Shokushu

Enieublis McShlaggrmasker

Shokushu

To get to the point you can call it an AI in the first place you've got to work out consciousness which implies that it is doing more than just giving a canned response to environmental stimulus like is suggested here.

Actually I don't believe that consciousness is necessary for Artificial Intelligence. AI is defined as "the theory and development of computer systems able to perform tasks normally requiring human intelligence"(The New Oxford Dictionary of English, 1998.). Those computer systems do not have to perform all such tasks including consciousness. (if conscious is a task requiring intelligence)

I didn't say that AI's need it, just that you need to know how consciousness works to make an AI. Sort of set it as a milestone in understanding how humans think, as it would be difficult to grasp enough about our minds without having delved into that particular question. It's rather central to the issue so to speak, though not necessarily something we'd want to take on the burden of being creators of.

Plus it's an easy goal to understand sort of what the endpoint looks like and what hurdles we've got to clear to get closer to it.

OK, but you don't need to know how consciousness works to make an AI, you just need to know how people solve problems and create an intelligent agent that mimics that problem solving procedure. I'm just adding a little to the definition I found originally.

Consciousness may be related to the ethics issue, but I'm really not sure how related it is to Artificial Intelligence. I'm also not sure whether or not we would need to mimic conscious entities to find out what consciousness is, or if that would even help.

In fact, since consciousness is hard to define and since I have no idea how close we are to understanding it, I think a better basis for determining ethics is whether something is alive. Oxford dictionary defines life as: "the condition that distinguishes animals and plants from inorganic matter, including the capacity for growth, reproduction, functional activity, and continual change preceding death". The AI I described would have all of those capacities.

Teasing apart the various ways that we find strange connections between ideas is tantamount to understanding consciousness.

Life is a poor place to draw the line though as nobody bats an eyelash when you stab a cockroach in the brain or grind up algae into a paste. Those are both alive but fall short of the threshold that we actually care about.

Ryu Kei Shou Kawazu

If I put a fridge magnet on my head thatd be a little ridiculous lol.

Well fridge magnets are basically a bunch of bar magnets that are facing each other in sheet form. They're even weaker than the magnets you'd get from normally using those materials in that their magnetic field is extremely shallow. No way any number of those would even penetrate past the skin on your head.

Enieublis McShlaggrmasker

DXnobodyX

Im not advocating the idea or anything, just pointed out that its already been done.

But... they're completely different. My machine isn't made out of human brains or whatever that thing was. I think the way I summarized my AI just happened to be similar in some way to the story of biblical magi, which that computer was based off of.

Well both of them came from a very rudimentary decomposition of what it takes to make a thinking creature. You went with what is well known among lay people in the public that have any interest in these things so of course it's going to be fairly similar.

Getting at the latest peer review articles for neurology would both be difficult because those things aren't free and because they tend to be written at a technical level that is impenetrable to people that haven't already gone through a great deal of textbook study- but if you want any chance of coming up with ideas that haven't been put out there decades ago then that's probably the only way to do it.


For something a little more manageable you could look up news stories and try to become an autodidact from that. Science reporting tends to be really bad though so you'll accumulate lots of misinformation from when either reporters didn't care to understand the studies that were done or the scientists didn't care to explain it in an effective way. Still, it tends to give you at least a little perspective that the old school robotics thinkers lacked.

hence why I called it ridiculous....
...

[Shokushu]

Enieublis McShlaggrmasker

Shokushu

Teasing apart the various ways that we find strange connections between ideas is tantamount to understanding consciousness.

Life is a poor place to draw the line though as nobody bats an eyelash when you stab a cockroach in the brain or grind up algae into a paste. Those are both alive but fall short of the threshold that we actually care about.

Actually I think the reason some people don't care about some forms of life is because they don't view them as conscious or similar to them.

We could make AIs that fit the definition for life (as you stated) yet were less like us than cockroaches. I actually think that would be simpler since cockroaches don't do things like problem solving.

Quote:

I think what we care about is the worst place to draw the line because it depends completely on whatever random people care about.

Yes, human metrics do tend to stink but in this case it's more of a short cut for ethics in general, which is to say that people accept and propagate the ethical conclusion without needing to work through the steps that lead up to it. This issue was properly thought out ages ago and we haven't come to a point where the underlying assumptions of it need to be reevaluated.

Quote:

I wouldn't be surprised if some people believed that humans were the only conscious beings on the planet. Bringing Nazi's into the argument, I think many Nazi's believed Jews weren't conscious either.

The general propaganda was just that Jews had less of whatever miraculous qualities people thought of as human, like they were halfway between people and animals or something. They were really eager to point to numbers and say this or that was smaller for Jews (zeir brains are only two thirds the size of ours! achtung!) but they didn't seem to say that Jews were entirely devoid of the various human characteristics.

Just my impression though so maybe I missed the worst of it.

Quote:

It might be that what "we actually care about" is just what is most convenient for us to care about at the time.

Shokushu

Enieublis McShlaggrmasker

DXnobodyX

Im not advocating the idea or anything, just pointed out that its already been done.

But... they're completely different. My machine isn't made out of human brains or whatever that thing was. I think the way I summarized my AI just happened to be similar in some way to the story of biblical magi, which that computer was based off of.

Well both of them came from a very rudimentary decomposition of what it takes to make a thinking creature. You went with what is well known among lay people in the public that have any interest in these things so of course it's going to be fairly similar.

Getting at the latest peer review articles for neurology would both be difficult because those things aren't free and because they tend to be written at a technical level that is impenetrable to people that haven't already gone through a great deal of textbook study- but if you want any chance of coming up with ideas that haven't been put out there decades ago then that's probably the only way to do it.


For something a little more manageable you could look up news stories and try to become an autodidact from that. Science reporting tends to be really bad though so you'll accumulate lots of misinformation from when either reporters didn't care to understand the studies that were done or the scientists didn't care to explain it in an effective way. Still, it tends to give you at least a little perspective that the old school robotics thinkers lacked.

My AI wasn't designed from a rudimentary decomposition of psychology, it was summarized with psychological terms because they seamed somewhat close and could imply about what each part did.

I didn't mean what you actively tried to do. Like it or not in this day and age you're bombarded with these ideas from too many angles to count. Most of the thinking we do is along the well trodden path, as well it should be, so your ideas actually make sense instead of that cliched science genius character in so much media that uses some impenetrable process where he pulls the answers to questions out of nothing having anything to do with the task at hand.

You, like all real people, mostly follow ideas to their logical conclusion. Creativity is where you get bumped off the main road for a bit but then willingly explore the new path you've found yourself on- but the thing is that for decades you've been taking in these ideas so even off the main road there are still a lot of paths that have already been trodded.

Maybe you actually didn't come across a few pieces for the idea here or there but the framework is already there for you to rediscover them as you come to little stopping points in the road. This is why people say things like "there is nothing new under the sun."

So instead imagine if you yourself had spent a lifetime thinking about this question. Now stop imagining because that's basically already happened, except that it's just other people. They are really quite a ways ahead of you in terms of the time they've put in, but you've been passively taking in the easy bits all this time. If you just spend awhile seeking out the more complicated stuff and really looking into the topic in depth then you can catch up really fast and extend your reach out to where the trail blazers haven't already been and sent back postcards from.

That make the idea any clearer?

Quote:

My thought process:

Creating an AI based completely around self defense would leave me with a boring program that does almost nothing unless it is provoked constantly. So I get it to dislike similar situations, this means having a database of the environment around the AI and of how that environment works, and designing the AI to constantly expand its database, that is a bit more than "curiosity", but it seamed like a good word for summarizing that system. However, those AIs would constantly kill themselves, so I would need to add back in the self defense system and give it access to the database(this isn't always necessary for a game AI though, when you can have multiple lives).

Now I have a pretty good AI with that, but if I have multiple of these around, then they wouldn't be as efficient as possible since each AI would have to go through the full problem solving process to do anything. So I thought of a way to get them to learn from each other. Having an object defining the current AI (based on previous outputs) in the database and comparing that object with any object being observed seamed like a good way to improve efficiency. Empathy seamed like a good label for that one. Communication would arise with the combination of this and one of the other systems.

Then, from this forum, I heard the term collective unconscious, which I think is nonsense, but it gave me an idea. When there are a massive amount of these things together, it is not as efficient as possible again, since some groups are doing the same things as other groups. So one AI would be designed to regroup the other AIs to improve efficiency. That part isn't really like anything I know of. (I'm not sure if I could call it a president)

Then those president AIs could be applied on top of each other in a hierarchical manner.


For this, each AI would need: Databases, Compiling, De-compiling, functions, objects, contextual data comparison, and performance testing.


It would work with just the curiosity system, but it would be slower in most cases.

A general question worth asking with any technical idea is "what's the mechanism for that?" With curiosity you've sort of answered this via "make it dislike similar situations." The language leaves room for a lot of potential complaints but I can easily fill in the gaps to get at what you generally want the program to do.
This does seem to fall a bit short of useful though as you've basically declared that you want to give it ADD instead of making it actually figure things out- you just gave it a desire for different stimulus. Plan to wait until it exhausts the easily accessible stuff and has to invent ways to get at the stuff we actually cared to have it work on?

Self defense? What's the mechanism for that? I get that you don't want them destroying themselves left and right but how does it determine what is dangerous?

And you can keep asking that question down the list.

*And having some programming experience myself I don't think you really need so many coordinator roles to work between these things. Computers are fast so you could easily have every AI ask 1000 other AIs if they had looked into some issue and about how thoroughly they worked it out before it started to look at the issue. Then for anything above 0 but less than 100% it could devote some more time to accessing what the other machine did. Even if the things had ADD and went at problems in small increments before moving on to something else you could get progress like this- provided that you could afford to make so many of them to hook together. Before that you'd have to work out systems of recognizing what queries were really about the same subject but I don't think you're so much interested in that angle of AI development.

[Enieublis McShlaggrmasker]

OK, I can accept most of your points.

Shokushu

A general question worth asking with any technical idea is "what's the mechanism for that?" With curiosity you've sort of answered this via "make it dislike similar situations." The language leaves room for a lot of potential complaints but I can easily fill in the gaps to get at what you generally want the program to do.
This does seem to fall a bit short of useful though as you've basically declared that you want to give it ADD instead of making it actually figure things out- you just gave it a desire for different stimulus. Plan to wait until it exhausts the easily accessible stuff and has to invent ways to get at the stuff we actually cared to have it work on?

I also agree with the "what's the mechanism for that?" reasoning.

However, I think I might disagree somewhat with the part about the program not being useful because of ADD. If it was pure problem solving then it could get stuck forever on problems that are either impossible or that the program doesn't have the functionality to solve yet. If the program was given some "ADD", then it could sometimes give up on really hard problems temporarily, which is a drawback. But it would make up for that with the ability to gain new functionality and solve those problems more quickly.

Shokushu

Self defense? What's the mechanism for that? I get that you don't want them destroying themselves left and right but how does it determine what is dangerous?

I think I decided on measuring performance to determine this. And to have the machine prefer actions that lead to higher performance. It's not perfect, you might still have these AIs all jump off of a cliff because they don't know yet that they don't function very well after impact.

Shokushu

And you can keep asking that question down the list.

*And having some programming experience myself I don't think you really need so many coordinator roles to work between these things. Computers are fast so you could easily have every AI ask 1000 other AIs if they had looked into some issue and about how thoroughly they worked it out before it started to look at the issue. Then for anything above 0 but less than 100% it could devote some more time to accessing what the other machine did. Even if the things had ADD and went at problems in small increments before moving on to something else you could get progress like this- provided that you could afford to make so many of them to hook together. Before that you'd have to work out systems of recognizing what queries were really about the same subject but I don't think you're so much interested in that angle of AI development.

That depends on how much data is being sent, and how it is being sent. If it was all on the same computer and the AIs were sending references to databases, then I might have everything send data to everything else at once. It might not be the most efficient process, but it would still happen quickly. If the AIs were on multiple computers throughout the internet, and information from the databases had to be sent between them, then it would probably slow down a lot if everyone was asking everyone about everything.


I think I might create diagrams for ideas like this so they don't become walls of text. I might create a UML for this... eventually. I have some other things to do first.

[Shokushu]

Enieublis McShlaggrmasker

OK, I can accept most of your points.

Shokushu

A general question worth asking with any technical idea is "what's the mechanism for that?" With curiosity you've sort of answered this via "make it dislike similar situations." The language leaves room for a lot of potential complaints but I can easily fill in the gaps to get at what you generally want the program to do.
This does seem to fall a bit short of useful though as you've basically declared that you want to give it ADD instead of making it actually figure things out- you just gave it a desire for different stimulus. Plan to wait until it exhausts the easily accessible stuff and has to invent ways to get at the stuff we actually cared to have it work on?

I also agree with the "what's the mechanism for that?" reasoning.

However, I think I might disagree somewhat with the part about the program not being useful because of ADD. If it was pure problem solving then it could get stuck forever on problems that are either impossible or that the program doesn't have the functionality to solve yet. If the program was given some "ADD", then it could sometimes give up on really hard problems temporarily, which is a drawback. But it would make up for that with the ability to gain new functionality and solve those problems more quickly.

Come on now, I didn't say that it should devote infinite time to solving the same program. Even today programmers have lots of ways of detecting infinite loops so as soon as an AI wasted some specific amount of time without making any progress it could easily have some routine for switching to more fruitful ventures.

Quote:

Shokushu

Self defense? What's the mechanism for that? I get that you don't want them destroying themselves left and right but how does it determine what is dangerous?

I think I decided on measuring performance to determine this. And to have the machine prefer actions that lead to higher performance. It's not perfect, you might still have these AIs all jump off of a cliff because they don't know yet that they don't function very well after impact.

I'm not at all satisfied with this.

Quote:

Shokushu

And you can keep asking that question down the list.

*And having some programming experience myself I don't think you really need so many coordinator roles to work between these things. Computers are fast so you could easily have every AI ask 1000 other AIs if they had looked into some issue and about how thoroughly they worked it out before it started to look at the issue. Then for anything above 0 but less than 100% it could devote some more time to accessing what the other machine did. Even if the things had ADD and went at problems in small increments before moving on to something else you could get progress like this- provided that you could afford to make so many of them to hook together. Before that you'd have to work out systems of recognizing what queries were really about the same subject but I don't think you're so much interested in that angle of AI development.

That depends on how much data is being sent, and how it is being sent. If it was all on the same computer and the AIs were sending references to databases, then I might have everything send data to everything else at once. It might not be the most efficient process, but it would still happen quickly. If the AIs were on multiple computers throughout the internet, and information from the databases had to be sent between them, then it would probably slow down a lot if everyone was asking everyone about everything.

Just the sentence "have you looked into genetic factors contributing to dwarfism?" takes up a negligible amount of memory and you could easily ping a million other computers with it even with our current day infrastructure.

There's something worse here though because your coordinator role doesn't really help with this problem though as it has to handle the full volume of inquiries in addition to recording all of the projects that have been worked on. With the ADD issue above this would basically mean there was double the traffic load constantly flooding it.

Of course with computers you've always got to check how much memory/transfer rate things actually take before you can tell if some system can actually handle it, which is sort of my (sub)point here. You've just decided that this seems like a computer-y way to organize things when you don't really have anything tangible to evaluate.

Quote:

I think I might create diagrams for ideas like this so they don't become walls of text. I might create a UML for this... eventually. I have some other things to do first.

Might want to see what leaders in the field of computing (within the last few decades) have had to say about setting up this kind of thing.

[Enieublis McShlaggrmasker]

Shokushu

Come on now, I didn't say that it should devote infinite time to solving the same program. Even today programmers have lots of ways of detecting infinite loops so as soon as an AI wasted some specific amount of time without making any progress it could easily have some routine for switching to more fruitful ventures.

OK, I guess it would be another infinite loop detector then.

Shokushu

Enieublis_Mcshlaggrmasker

I think I decided on measuring performance to determine this. And to have the machine prefer actions that lead to higher performance. It's not perfect, you might still have these AIs all jump off of a cliff because they don't know yet that they don't function very well after impact.

I'm not at all satisfied with this.

I'm not really satisfied either, but I can't think of anything better than that short of adding a whole bunch of special cases to the database before starting the program.

Shokushu

Just the sentence "have you looked into genetic factors contributing to dwarfism?" takes up a negligible amount of memory and you could easily ping a million other computers with it even with our current day infrastructure.

There's something worse here though because your coordinator role doesn't really help with this problem though as it has to handle the full volume of inquiries in addition to recording all of the projects that have been worked on. With the ADD issue above this would basically mean there was double the traffic load constantly flooding it.

Of course with computers you've always got to check how much memory/transfer rate things actually take before you can tell if some system can actually handle it, which is sort of my (sub)point here. You've just decided that this seems like a computer-y way to organize things when you don't really have anything tangible to evaluate.

Actually I decided on this because it seamed similar to certain fractal that is usually useful for these things.

If the coordinator had to take in data from everyone, then send data to everyone else, then it would make communication worse. However, if the coordinator took in data from a specific region of AIs that are looking into x and asked another coordinator about y because that coordinator talks to a region that looks into y, then it would improve communication. It would be like having addresses based around what you are studying.

The coordinators would still have to deal with a lot of requests though, so they would probably require better computers, or servers.

Shokushu

Quote:

I think I might create diagrams for ideas like this so they don't become walls of text. I might create a UML for this... eventually. I have some other things to do first.

Might want to see what leaders in the field of computing (within the last few decades) have had to say about setting up this kind of thing.

I looked into it. Commonsense reasoning seamed similar to what I was doing. I found some of the leading researchers in the field here, and I found a website showing some problems in commonsense reasoning here. Those will probably be useful.

[Shokushu]

Enieublis McShlaggrmasker

Actually I decided on this because it seamed similar to certain fractal that is usually useful for these things.

If the coordinator had to take in data from everyone, then send data to everyone else, then it would make communication worse. However, if the coordinator took in data from a specific region of AIs that are looking into x and asked another coordinator about y because that coordinator talks to a region that looks into y, then it would improve communication. It would be like having addresses based around what you are studying.

The coordinators would still have to deal with a lot of requests though, so they would probably require better computers, or servers.

-Really only have anything to say about this section-

So you're not describing it thoroughly because it's an idea somebody else specifically told you and this is all you remember?

[Enieublis McShlaggrmasker]

Shokushu

Enieublis McShlaggrmasker

Actually I decided on this because it seamed similar to certain fractal that is usually useful for these things.

If the coordinator had to take in data from everyone, then send data to everyone else, then it would make communication worse. However, if the coordinator took in data from a specific region of AIs that are looking into x and asked another coordinator about y because that coordinator talks to a region that looks into y, then it would improve communication. It would be like having addresses based around what you are studying.

The coordinators would still have to deal with a lot of requests though, so they would probably require better computers, or servers.

-Really only have anything to say about this section-

So you're not describing it thoroughly because it's an idea somebody else specifically told you and this is all you remember?

No.

I don't know if the fractal even has a name. I was thinking about road-maps around cities, the skeleton of a leaf, or other similar structures. I think it could more easily be described as an index though... I don't know why I didn't just say index.

If an AI wanted to know about biology, and it was in a psychology group, then its request would only have to go through a few other systems instead of all of them. Or, if an AI did not know to request anything, then a coordinating AI could suggest something, because it searched actions x, y, and z, because they seamed significantly different from the rest of the group. If some coordinators were constantly getting requests from each other, then they could move closer together to improve communication.

It's almost exactly the same as an index, but it changes all the time and has AIs linking addresses instead of partial words. This is useful since many non-coordinating AIs might not know that what they're doing is significantly different, and since the search terms could change constantly as new rules are discovered.

[Shokushu]

Enieublis McShlaggrmasker

Shokushu

Enieublis McShlaggrmasker

Actually I decided on this because it seamed similar to certain fractal that is usually useful for these things.

If the coordinator had to take in data from everyone, then send data to everyone else, then it would make communication worse. However, if the coordinator took in data from a specific region of AIs that are looking into x and asked another coordinator about y because that coordinator talks to a region that looks into y, then it would improve communication. It would be like having addresses based around what you are studying.

The coordinators would still have to deal with a lot of requests though, so they would probably require better computers, or servers.

-Really only have anything to say about this section-

So you're not describing it thoroughly because it's an idea somebody else specifically told you and this is all you remember?

No.

I don't know if the fractal even has a name. I was thinking about road-maps around cities, the skeleton of a leaf, or other similar structures. I think it could more easily be described as an index though... I don't know why I didn't just say index.

If an AI wanted to know about biology, and it was in a psychology group, then its request would only have to go through a few other systems instead of all of them. Or, if an AI did not know to request anything, then a coordinating AI could suggest something, because it searched actions x, y, and z, because they seamed significantly different from the rest of the group. If some coordinators were constantly getting requests from each other, then they could move closer together to improve communication.

It's almost exactly the same as an index, but it changes all the time and has AIs linking addresses instead of partial words. This is useful since many non-coordinating AIs might not know that what they're doing is significantly different, and since the search terms could change constantly as new rules are discovered.

Sounds like any branching fractal but if I had to guess then I'd say the major keyword you're missing is nodes.

[Enieublis McShlaggrmasker]

Shokushu

Enieublis McShlaggrmasker

Shokushu

Enieublis McShlaggrmasker

Actually I decided on this because it seamed similar to certain fractal that is usually useful for these things.

If the coordinator had to take in data from everyone, then send data to everyone else, then it would make communication worse. However, if the coordinator took in data from a specific region of AIs that are looking into x and asked another coordinator about y because that coordinator talks to a region that looks into y, then it would improve communication. It would be like having addresses based around what you are studying.

The coordinators would still have to deal with a lot of requests though, so they would probably require better computers, or servers.

-Really only have anything to say about this section-

So you're not describing it thoroughly because it's an idea somebody else specifically told you and this is all you remember?

No.

I don't know if the fractal even has a name. I was thinking about road-maps around cities, the skeleton of a leaf, or other similar structures. I think it could more easily be described as an index though... I don't know why I didn't just say index.

If an AI wanted to know about biology, and it was in a psychology group, then its request would only have to go through a few other systems instead of all of them. Or, if an AI did not know to request anything, then a coordinating AI could suggest something, because it searched actions x, y, and z, because they seamed significantly different from the rest of the group. If some coordinators were constantly getting requests from each other, then they could move closer together to improve communication.

It's almost exactly the same as an index, but it changes all the time and has AIs linking addresses instead of partial words. This is useful since many non-coordinating AIs might not know that what they're doing is significantly different, and since the search terms could change constantly as new rules are discovered.

Sounds like any branching fractal but if I had to guess then I'd say the major keyword you're missing is nodes.

Yup, that's the word.