True AI When?

SirNitram · Post by **SirNitram** » 2003-05-18 12:07am

You know, I really have to interject. Why do we HAVE to do it top down? We have no proof that such is even possible, IE, that an intelligence can be designed. Why don't we use the tools that made us? We've proven that can work, at least in part, in the digital realms.

kojikun · Post by **kojikun** » 2003-05-18 01:22am

Why make AI/AS computers? Because they can be made on demand, faster, and in larger numbers, without the same demands as humans.

We could send AI/AS on interplanetary colonisation missions more cheaply then humans, because theres no food reqs, living space reqs, life support reqs, etc. All the limitations of an organic body would be nigh removed with AI/AS.

BlkbrryTheGreat · Post by **BlkbrryTheGreat** » 2003-05-18 02:33am

The Turing Test is a good way to discover if something is sentient or not. Not perfect, of course, but it's suitable

Actually the Turing Test is a horrible test to determine if an "AI" is indeed sentient. Think about it, no one (or least almost no one) on this board actually knows if they're corresponding to a man or woman 100% of the time. Knowing that we should ask ourselves why would we use a test to determine sentince when the members of humanity cannot pass it 100% of the time.

kojikun · Post by **kojikun** » 2003-05-18 02:49am

a problem with many AI that presently come close to passing the turing test, and wouldnt do so if the test was done properly, is that theyre repetetive and dont respond uniquely.

find a chatbot that talks realistically and say something very odd and random to it. try to trip it up. its easy.

UltraViolence83 · Post by **UltraViolence83** » 2003-05-18 04:33am

kojikun wrote:I know I won't treat them second rate. Then again, when they're possible, I'll have moved over into a computer brain so I'll be fighting right along side them.

The thing is, when we do have true AI in ANDROID bodies, we'll also be so integrated with computers as intelligence inhancers and such that we won't see terribly much distinction.

*Shoots Koji with an EMP gun* Ooops...

Post by **Darth Wong** » 2003-05-18 04:38am

So the question is "when will AI's achieve sentience, eh?"

Please define "sentience". What does it mean to be "self-aware?" Need we only mimic a human?

UltraViolence83 · Post by **UltraViolence83** » 2003-05-18 07:40am

kojikun wrote:Why make AI/AS computers? Because they can be made on demand, faster, and in larger numbers, without the same demands as humans.

We could send AI/AS on interplanetary colonisation missions more cheaply then humans, because theres no food reqs, living space reqs, life support reqs, etc. All the limitations of an organic body would be nigh removed with AI/AS.

Then it's not true colonisation. You need people to live somewhere to colonize it. That's like saying we've been to Jupiter because we sent probes to it.

Rye · Post by **Rye** » 2003-05-18 08:53am

Darth Wong wrote:So the question is "when will AI's achieve sentience, eh?"

Please define "sentience". What does it mean to be "self-aware?" Need we only mimic a human?

Sentience...not overly sure

self awareness is when you can make a map of your surroundings and put yourself in the map, (according to 2nd terminator dvd) which is what skynet does.

The neural net that became skynet was designed to learn, which it did at e "Geometric rate". It learned the most efficient way to do what it was programmed to do (preserve itself, and defend itself against aggressors) which is precisely what it did when the humans tried to pull the plug.

According to the section on NNs, because james cameron wanted extensive research into everything he was doing (well not when the truck's diesel was ignited by a spark

) an the processor was designed from that research with technology we dont have yet.

Thanks to the cpu being from the future, they thought of a revolutionary way to design new cpus, massive amounts of multi processing neurones, "molecular memory", and a room temperature superconductor.

That's all i can remember for now...but i think they purposefully left a lot of it open ended, being as theoretical as anything the science behind it...

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-05-18 09:21am

I'd just like to point out that most scientists have thrown out the idea of top-down structuring of AI now as it is really too complicated to be done. The best bet is to get basic learning AIs that can, like we did, figure out things themselves but at increased speeds via heuristic neural nets etc. A good example would be the Seven Dwarves who are arguably the best learning AIs on the planet despite being tiny vehicles that avoid bumping into things as a collective hive mind.

kojikun · Post by **kojikun** » 2003-05-18 12:57pm

UltraViolence83 wrote:Then it's not true colonisation. You need people to live somewhere to colonize it. That's like saying we've been to Jupiter because we sent probes to it.

Ain't true colonisation, but it is a way to put a preperation team on the planet without the cost of putting humans there. Putting an AI/AS team there to start building and such is easier then putting humans. Their requirements are lower, so they can do the same things we would do in their place minus the extra stuff. They would be the first teams sent.

Darth Wong wrote:So the question is "when will AI's achieve sentience, eh?"

Please define "sentience". What does it mean to be "self-aware?" Need we only mimic a human?

I'd say the ability to control ones own thoughts. Thoughts control action, but controlling ones own action is already done with computers. I think deciding what one is going to think about is more important. And that requires the ability to know what you're currently thinking about as well as some for of imagination which would give you litle ideas of what else to think about.

Admiral Valdemar wrote:I'd just like to point out that most scientists have thrown out the idea of top-down structuring of AI now as it is really too complicated to be done. The best bet is to get basic learning AIs that can, like we did, figure out things themselves but at increased speeds via heuristic neural nets etc. A good example would be the Seven Dwarves who are arguably the best learning AIs on the planet despite being tiny vehicles that avoid bumping into things as a collective hive mind.

What do you mean top down, exactly?

NapoleonGH · Post by **NapoleonGH** » 2003-05-18 03:08pm

The people in Terminator made the biggest mistake ever, they programed a computer where self preservation was the most important programming then tried to kill it. Bad move.

If we used Isaac Asimov's 3 (or maybe 4 if we include the Zeroth law) laws of robotics as basic fundamental programing aspects that are hard wired into each AS brain, we will have no problems.

Drooling Iguana · Post by **Drooling Iguana** » 2003-05-18 03:59pm

NapoleonGH wrote:If we used Isaac Asimov's 3 (or maybe 4 if we include the Zeroth law) laws of robotics as basic fundamental programing aspects that are hard wired into each AS brain, we will have no problems.

No problems? Just about every story in I, robot revolved around robots screwing up because of their three laws, often in such a way that it puts human lives in danger.

kojikun · Post by **kojikun** » 2003-05-18 04:08pm

iguana, howd that happen??

Admiral Valdemar · Post by **Admiral Valdemar** » 2003-05-18 04:18pm

kojikun wrote: What do you mean top down, exactly?

Top-down AI architecture means to create a sentient or extremely smart program from scratch, making it so it has a command line and protocol for every possible scenario. Obviously, this is foolish now, years was spent by programmers trying to make something have a contingency for any problem, it's just too diverse a task.

Instead, bottom-up means you get a program to have the basics on how to learn intelligently and let it teach itself without human interaction or someone programming the machine to respond a certain way to everything.

Rye · Post by **Rye** » 2003-05-18 04:49pm

Admiral Valdemar wrote:
kojikun wrote: What do you mean top down, exactly?
Top-down AI architecture means to create a sentient or extremely smart program from scratch, making it so it has a command line and protocol for every possible scenario. Obviously, this is foolish now, years was spent by programmers trying to make something have a contingency for any problem, it's just too diverse a task.

Instead, bottom-up means you get a program to have the basics on how to learn intelligently and let it teach itself without human interaction or someone programming the machine to respond a certain way to everything.

That's what skynet does, according to t2 2nd dvd. The guy who was the tech consultant for the neural net actually appears in the scene where we see Dyson for the first time (i think) and he says at a just audible level "perhaps it's the behavioural inbitors, which is what you need in an NN apparently, so it doesn't keep sending the same signals thorughout itself and getting a massive feedback loop.

IT's something to do with multiprocessing, with each of the miniprocessors equalling a neurone, and getting them to "fire" simultaneously is a big part of the learning process. I don't remember it entirely, but it goes on for aaaaaaaaages.

Baron Mordo · Post by **Baron Mordo** » 2003-05-18 04:58pm

Drooling Iguana wrote:
NapoleonGH wrote:If we used Isaac Asimov's 3 (or maybe 4 if we include the Zeroth law) laws of robotics as basic fundamental programing aspects that are hard wired into each AS brain, we will have no problems.
No problems? Just about every story in I, robot revolved around robots screwing up because of their three laws, often in such a way that it puts human lives in danger.

What are those laws, again?

kojikun · Post by **kojikun** » 2003-05-18 05:04pm

Admiral Valdemar wrote:Top-down AI architecture means to create a sentient or extremely smart program from scratch, making it so it has a command line and protocol for every possible scenario. Obviously, this is foolish now, years was spent by programmers trying to make something have a contingency for any problem, it's just too diverse a task.

Instead, bottom-up means you get a program to have the basics on how to learn intelligently and let it teach itself without human interaction or someone programming the machine to respond a certain way to everything.

Ahh ok. Thats what theyre doing at MITs AI Lab with COG and Gizmo.

kojikun · Post by **kojikun** » 2003-05-18 05:06pm

Baron Mordo wrote:What are those laws, again?

Code: Select all

A robot may not injure a human being or, through inaction, allow a human being to come to harm.

A robot must obey orders given it by human beings except where such orders would conflict with the First Law.

A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

I'm not sure the Zeroth law tho

NapoleonGH · Post by **NapoleonGH** » 2003-05-18 08:29pm

Zeroith law came from the later robot books and the later foundation books (the 2 prequils writen by Asimov go into this alot)

It basically states that
"A robot must not harm or through inaction allow to come to harm the human race" or somesuch. It overrides all other laws, and basically says that it is ok to disobey/hurt a human singularily if it will benefit the human race as a whole.

And as far as I, Robot having all those problems associated with the laws, yes you are right, but most of those problems end up getting worked out by the time the robot books come into play.

kojikun · Post by **kojikun** » 2003-05-18 08:33pm

what are those problems tho? what problems arise from the 3 laws?

NapoleonGH · Post by **NapoleonGH** » 2003-05-18 08:43pm

basically they can arise when conflicts in their programing occur.

like when law 2 and 3 get in the way of each other. In one case one of the robots was brand new and really expensive so they strenghtened the 3rd law so that it preserves itself more, then someone ordered it into a dangerous situation and it couldnt resolve the problems

in another case a robot brain was asked to design and build a hyperspace ship, but the problem was that the inhabitants "died" instantly every time you went into hyperspace, so the brain couldn't figure out how to deal with being ordered to do it and it causing people to "die" temporarily so it shut down, luckily this was a consolidated computer rather than a US Robotics one.

kojikun · Post by **kojikun** » 2003-05-18 09:02pm

They died temporarilly or permanently?

NapoleonGH · Post by **NapoleonGH** » 2003-05-18 10:39pm

well the people were alive before going into hyperspace, and they were alive after coming out of hyperspace, but during the instant they were in hyperspace, they weren't alive

kojikun · Post by **kojikun** » 2003-05-18 10:48pm

hows that work now? lol

Queeb Salaron · Post by **Queeb Salaron** » 2003-05-18 11:37pm

Darth Wong wrote:So the question is "when will AI's achieve sentience, eh?"

Please define "sentience". What does it mean to be "self-aware?" Need we only mimic a human?

Only? Are you saying that there's a higher form of sentience, Wong?

Sounds like an exerpt from a conspiracy theory.