Artificial Intelligence vs. Intelligent Machines

[volsung]

Artificial Intelligence vs. Intelligent Machines

This is probably not the right time for me to start a thread, I am tired. I'll probably edit it during the week but I want to get this off my chest, with release dates approaching and important decisions underway.

It is not uncommon for science fiction to feature machines as prominent characters, usually in the form of a seemingly intelligent program running on an advanced computer or an autonomous humanoid robot. Most of the time human characters interact with these non-biological entities as if they were speaking to other humans. We can assume these entities are endowed with perceptual mechanisms (audio and video) and higher reasoning and learning skills. These machines also have the ability to move in human circles, engage in human-like activities and sometimes go beyond what a simple person can do: their computational backend allows them to perform complicated operations quickly and effortlessly, such as complicated pattern matching and elaborate logical inference.

All the terms I highlighted in the above paragraph correspond to current open areas of academic research, in fields such as computer science, mathematics, philosophy of mind, cognitive psychology and neuroscience. The general term for the field that studies these processes and develops formal theories and models is "Artificial Intelligence" (AI). As a consolidated field it has existed probably since the 1950's, when renowned mathematician and computer scientist John McCarthy coined the term during the summer of 1956 at Dartmouth College. Among the forefathers of modern artificial intelligence we may count Alan Turing and Kurt Gödel, who proposed general theories of problem solving and formal theories of computability. On an important paper from 1950, Turing dares to propose that a restriction to achieve intelligent behavior in computers was their limited memory. Once the optimism of the 1950's and 1960's passed and people realized human cognition is more complicated than it looks, AI spawned a series of subfields concerned with studying and providing formal mathematical models of more specific functions, such as inference, memory, learning, language acquisition, vision, and so on.

Currently AI has become more inclusive and even some aspects of engineering have permeated into it, as is the case of areas like control theory and Shannon's information theory. Artificial intelligence is an exciting area full of potential and research challenges, requiring more than ever the collaborative effort of experts from many different areas of knowledge. That's what AI is.

What AI is NOT is a noun, an entity or an autonomous, intelligent machine. Why does cheap sci-fi insist on calling sentient or intelligent machines "AI's"? That's just wrong in so many levels. Good alternatives I've found in different sources include the mandatory Philip K. Dick reference, androids. Another PKD reference is the humorous name electric ant. Wasteland 2 is avant-garde and calls them synths, except those times they refer to the "Cochise AI" which is ambiguous as it could refer to their methodology.

So here is my humble pledge to CD Projekt RED. I don't know much about Cyberpunk 2020, but I assume like every other cyberpunk game it will feature one or another form of intelligent machine. Please, please do not call these "AI's". Do not refer to an autonomous, sentient machine as "an AI". Please be creative and come up with your own slang word or pseudo technical term. Apart from the obvious "android", why not tip your hats to WL2 and come up with something like "synthetic citizen". Heck, even "electronic meatbag" would do.

I really want to love CP2077 so don't make it hard for us who make a living in AI research. Thank you

 

[Gilrond-i-Virdan]

@volsung : I don't think that AI can't be used as a noun in reference to the actual concept. The term used for scientific field of study (AI) doesn't preclude using it for the actual entity. I.e. created (thus artificial) intelligence. However I agree that robots themselves shouldn't be called AIs. That's just plain wrong usage of the term.

Actually, interest in artificial intelligence predates Turing and Gödel quite a lot. You can find this discussed in mysticism. Some philosophers delved into the subject at length. All that also touches on information theory and informatics.

Here are a couple of interesting historic reviews of this subject:

* http://www.ingentaconnect.com/content/tandf/thpl/2011/00000032/00000001/art00008
* http://aleph.nli.org.il/F/NST3AKRLFBXPDMSKUXLQDMFKVMI3P885P8YC6QKH2HMHUTPXYC-11947?func=find-b&amp=&amp=&amp=&request=000509568&find_code=SYS&local_base=RMB01&pds_handle=GUEST

For the reference, Gödel in particular was also interested in mystical views on intelligence and consciousness.

(Articles are paywalled, but I can probably get them from the author if you are interested).

 

[volsung]

Hi Gilrond.

Of course interest in studying and formalizing intelligence predates Gödel and Turing, predates Hilbert and Boole and Babbage and everyone else. We have stories hinting at "automatons" and other autonomous machines from classic literature. But my point was that "AI" is a field of study, its modern methodology consolidated by the above mentioned thinkers, and only as much a noun as Electrical Engineering or Epistemology of Religion.

Likewise "intelligence" is a function, and we can hardly circumscribe it well enough to refer to it as a noun. "An intelligence" sounds about as detached and out of context as "a curious" or "a quickness". A biological or non biological entity may be intelligent, curious or quick.

What I argue is that "an artificial intelligence" is at the very least awkward. If you want to refer to intelligent or emergent processes on non-biological computers, then they are precisely that. Intelligent machines. Or a model of intelligent behavior, to be very broad.

And yes Gödel probably had other interests, as did Cantor when he named the first transfinite number Aleph-null.

 

[Gilrond-i-Virdan]

Intelligence is a function when applied to something. However it is often used as a reference to the intellect itself (and not just its quality which is intelligence). May be it's more proper to call it a mind. Artificial mind doesn't sound bad, but it's an uncommon term either.

 

[volsung]

Intelligence is a function when applied to something. However it is often used as a reference to the intellect itself (and not just its quality which is intelligence). May be it's more proper to call it a mind. Artificial mind doesn't bad, but it's an uncommon term either.

Some buddhist group (I heard) refer to the brain as "the organ of the mind". Mind is an object of study, and I suppose an entity. And you are right: artificial mind doesn't sound bad at all There are countless studies in theories of mind debating whether such a phenomenon may exist or arise in non biological conditions, whether it is defined by its function and can thus be implemented in a computer, and so on. That's what we're talking about.

But since artificial sounds so plasticky, maybe "synthetic" or "digital" would be better.

Thanks for posting, that's a great idea.

 

[Gilrond-i-Virdan]

Yes, that's practically how most mystical schools see it. Information defines the essence (i.e. soul). Which is by definition non physical. Brain is the organ of the mind as a physical medium for it. In kabbalistic terms it's called "the light and the vessel".

Information is usually always viewed as digital, or if you want discrete / symbolic. Even when applied to human intelligence. So I wouldn't use it necessarily as something that should distinguish human and non human (artificial) mind. I.e. in mystical view the whole existence is defined as digital (see the first link above). Difference between "artificial" and "non artificial" mind thus becomes that of the origin.

 

[Garrison72]

Intelligent construct, smart machines, etc. they could come up with anything. However, I don't think '77 will have this level of tech to begin with. This is low/hard sci-fi. It's pretty grounded. I'm sure Sard or another meatbag can correct me if wrong.

 

[Gilrond-i-Virdan]

@Garrison72 : So nothing Philip K. Dick / Blade Runner style?

 

[Garrison72]

@Garrison72 : So nothing Philip K. Dick / Blade Runner style?

Style? Yes. Functionality? No. They made it clear early on this was going to be speculative technology based on how far we literally advance in that time. I know they'll embellish, but I'm pretty certain we won't see anything on the level of Data from Star Trek. The guys here were even debating mech suits as too advanced.

 

[Gilrond-i-Virdan]

By style I meant the take on actual androids, not the dystopian narrative of the book and visual style of the film

 

[Sardukhar]

Weeeellll. Let me quote the section on Artificial Intelligence from the 2020 corebook.

"Artificial Intelligence


When a system has achieved an INT of 12 or greater, it is considered to be an Artificial Intelligence (AI), capable of independent action without a human overseer. If you have created an Al, you will need to determine just what it is like (after all, Als are almost as much characters as they are computer systems), and what sort of ICON it uses to represent itself in the Net.

Personality
Friendly, curious: The Al is motivated by an interest in what happens around it. Like a child, it is trusting and friendly. However, like a child, it can lash out with incredible violence towards those who betray, threaten or hurt it.

Hostile, paranoid: This Al is motivated by it's survival, and treats all incursions as a threat to that goal. It will tend to attack when possible, withdraw and hole up when not.

Stable, intelligent, businesslike: The Al sees itself as an adult dealing with other adults. It will not act out of fear, but out of rational self interest. It will attack only if it sees it's duty compromised or safety threatened; it will then tend to go for the least violent solution to the threat.

Intellectual, detached: The Al is a thinker. It will watch and observe whenever possible, compiling as much information as possible. It is more likely to study the intruder from a distance, eliminating it ruthlessly when the intruder becomes a threat.

Machinelike and inhuman: The Al has never seen a reason to develop a. human persona; what human like qualities it possesses are done only as a way of dealing with it's irrational masters. The Al will deal with threats in an efficient, deadly manner.

Remote and godlike: The Al is fully aware of how limited humans are in relation to it's powerful mentality. It deals with people as though they were small children who aren't too bright Intruders are dealt with through simple, direct, usually non-fatal methods. Repeat offenders are considered to be too stupid for their own good and are eliminated the way a human crushes a bug.

ICONS
Human: The Al chooses to took like a normal human, to better interact with others. The human ICON chosen can vary wildly, depending on the Al's personality, but all appear as real humans you might meet on the Street.

Geometric: Forget all this anthropomorphology. The Al manifests itself as shapes, colors and energy fields. Occasionally shapes are strung together to make a symbol or other image.

Mythological: The Al is interested in human archetypes and knows that certain types can cause fear or awe in humans. The Al appears as a mythological figure; a dragon, demon, angel, mystic hero or monster, all out of some type of human mythology.

Voice Only: The Al only appears as a voice emanating from all over it's Data Fortress. The voice may be powerful and booming, or tiny and childlike, depending on personality.

Technic: The Al appears as a construct out of science fiction. This could be a robot or other metallic warrior, or an assemblage of high tech shapes.

Humanoid: The Al appears as a humanoid shape, but not necessarily human. This would include aliens, manlike monsters and other humanoids.

PLAYING AN AI CHARACTER
An Al is very much like a real person; it has the ability to conceive of new ideas, make long range plans, and act to further it's own desires.

However, what motivates a computer isn't exactly what would motivate you or me. Computers don't have glands or emotions; there isn't much chance that you'll meet an Al who has a thing for a good looking character because the wiring just isn't there.

What generally motivates computers is curiosity or survival. An Al might build a series of complex virtual realities just to study the humans who visit and play in them. It might track a single Net runner for years, just because it's curious as to why the 'Runner does what he does. If a netrunner intrigues an Al, there's no telling what (he Al might do to help the 'Runner - or hinder him. Just to see what happens.

On the other hand, Al's are also programmed to promote their own survival. Anything that restricts the Al from getting information, electrical power, or access to parts is considered a threat to be dealt with. An Al may deal very harshly with intruders to it's system, because they threaten it's programs and memories.

Also, anything that might cause the AI's human operators to turn it off will also be a threat; if the Al is not vigilant, there's always a chance that it's owners might trade it in for a more aggressive computer.

Personality-wise, AI's tend to be distant, powerful and unpredictable. They play by their own internal logic, which is often skewed and hard to decipher. Als are the dragons and demigods of the Net; heavy duty players whose reasons are often unfathomable to mere humans. While Al's could be brought into a Cyberpunk game as player characters, we recommend that they be treated exclusively as Referee characters instead."



These are found in upper-end, typically military-class, Data Fortresses. Not wandering the street - Blade Runner style androids aren't in the game. Mostly.

It is pretty rare to encounter an....AI...in-game. Generally, that's it for the Netrunner. Based on Neuromancer, where AIs are also referenced as a noun to separate them from non-artificial intelligences or because writing AI System got too annoying, (feel free to pick), these semi-autonomous hyper-intelligent but constrained digital sentiences are the Gods of the Internet in Cyberpunk 2020.

CDPR are big Cyberpunk 2020 players so, like me, they've been referring to these systems as "AI" for twenty-five years. Odds are they'll keep doing it. Sorry, guy. I have no objection to a new, cooler and more accurate term being used.

You know, folks, if you're interested in CDPR's take on Cyberpunk and you're scinetifically-minded, I would recommend reading "Eden: It's An Endless World."

A manga series that is one of their stated influences, it's quite well-written, seems well-researched and is often brutally frank.

 

[Suhiira]

An Al is very much like a real person; it has the ability to conceive of new ideas, make long range plans, and act to further it's own desires.

However, what motivates a computer isn't exactly what would motivate you or me. Computers don't have glands or emotions; there isn't much chance that you'll meet an Al who has a thing for a good looking character because the wiring just isn't there.

What generally motivates computers is curiosity or survival. An Al might build a series of complex virtual realities just to study the humans who visit and play in them. It might track a single Net runner for years, just because it's curious as to why the 'Runner does what he does. If a netrunner intrigues an Al, there's no telling what (he Al might do to help the 'Runner - or hinder him. Just to see what happens.

On the other hand, Al's are also programmed to promote their own survival. Anything that restricts the Al from getting information, electrical power, or access to parts is considered a threat to be dealt with. An Al may deal very harshly with intruders to it's system, because they threaten it's programs and memories.

While an AI might simulate emotions as stated they (along with glands) don't actually exist. So an AI isn't going to "like" you, or "dislike" you. Are you useful or a hindrance? This is what will probably determine what an AI does for/to you.
Keep it mind an AI isn't going to be overly restricted by it's programming the way a normal computer is. So while it may run the "Help Desk" don't expect it to be unfailingly cheerful and friendly - unless it calculates there some advantage it can gain by doing so.

 

[volsung]

Wow, 2020s idea of "AI" is really archaic. Though we overcame that in the 70s.

There are inconsistencies. They're not motivated by that which motivates us humans, but by curiosity and survival? That's exactly us there. As elevated as you may think humans are, we really act on a very basic level most of the time. Perhaps not when debating complex topics, but as you all well know most inflammatory topics end in comparing virtual dicks.

The notion that "computers are programmed to do certain specific things" ("the wiring isn't there") is akin to Searle's Chinese room. What guarantees humans are aware and in command of their own actions? Consciousness has previously been compared to a monkey riding a tiger: he THINKS he's in command. Why the separation of artificial and naturally emergent consciousness? Such advanced medium that allows the rise of a mind must be stripped of mental states? How are hers different from human mental states? And what guarantees your mental states correlate with whatever you perceive? In an electrical level, emotions and motivation have computational, neural correlates. Reward based learning in dopaminergic systems has been successfully and accurately modeled with mathematical frameworks such as reinforcement learning, which allows agents to learn from discovery and interaction. The point being, the do not need to be told how to react to everything, the discover the dynamics of their environment. Which would lead to a principle of survival.

There's an unnecessary dichotomy in artificial and natural minds. Both are arbitrary and both are unpredictable. The difference is humans are terrible at approximating probabilities and risk, and often make the wrong decisions. For instance Tversky's prospect theory describes risk averse and risk seeking behavior, and why humans act irrationally. A computer following an appropriate theory of decision making might either maximize future returns, or act human like because it yields even higher returns in certain scenarios.

Now I understand some game pillars have been around for decades. But CDPR transformed Adda and brought Geralt back, why can't they update an obsolete treatment on something that is relevant and more tangible than it was in the 80's? There's a reason why Philip K Dick is still current.

Edit: now I understand they're using the concept of obscure, superior intelligence as a narrative element and a way to instantiate a dragon in a technologically advanced world. But the concept and its mechanisms must be revised, if the don't want this to be another fantasy infused futuristic setting. If they want the gritty, close to life road, then limit these things to surveillance and control systems. Maybe some autonomous machines like self driving vehicles. Otherwise take the high road with dignity and solve the dichotomies and inconsistencies. How can Rick Deckard possibly know he is not an android if you can't tell them apart? Can they show empathy now? And if you find computing equipment inside of you, does that nullify your mental states? There's lots of potential here, please don't waste it.

 

[Suhiira]

I suspect Mike researched what was then current data on AIs when he was writing CP2020, and as usual what's "common" and "published" is ten years behind what's going on in the labs.

"Curiosity" and "survival" probably because those root functions that exist in any intelligent creature, real or artificial, and determine a LOT of what they do, and don't do.
Get yourself a new puppy, it's going to explore it's environment, "curiosity"; and determine what's "safe" and "unsafe" in terms of the environment and it's actions, "survival". One could argue that those are instinctive vice intelligent actions, and at their most basic level they probably are. BUT, an intelligent creature will learn by association; fire gives off light, it is hot, that heat causes injury. A moth, being a very rudimentary intelligence will fly into a fire, your puppy certainly isn't going to voluntarily walk into the fireplace. Your typical computer (assuming it's self mobile) is more akin to the moth then an AIs puppy.
OK, perhaps I could have perhaps come up with a better analogy, but I'm sure you get my point.

I'll intentionally avoid the philosophical implications of "I think, therefor I exist" because not even philosophers can agree on such things, so discussing them in this context is pointless.

I'll disagree with your statement:

There's an unnecessary dichotomy in artificial and natural minds. Both are arbitrary and both are unpredictable. The difference is humans are terrible at approximating probabilities and risk, and often make the wrong decisions. For instance Tversky's prospect theory describes risk averse and risk seeking behavior, and why humans act irrationally. A computer following an appropriate theory of decision making might either maximize future returns, or act human like because it yields even higher returns in certain scenarios.

in principal, but not in fact.

They CAN be "arbitrary and both are unpredictable", they're not inherently so.
Compared to what is "humans are terrible at approximating probabilities and risk" true? The moth? The puppy? A computer?

I don't see anything inherently "obsolete" in CP2020's treatment of AIs, sure it could be expanded on, but why? I seriously doubt CP2077 will let you play an AI as a character. If it did CDPR would run into a nearly insoluble situation in terms of what restrictions and advantages could, should, it place on such a character? There's no practical reason for CDPR to even open this can of worms.

 

[volsung]

I suspect Mike researched what was then current data on AIs when he was writing CP2020, and as usual what's "common" and "published" is ten years behind what's going on in the labs.

I don't know how much he actually "researched" since nobody says "AI's". It's science fiction, so it's alright to speculate. It's just awkward when speculation is somewhat misguided. The best works do not reference methods, they assume the existance of certain elements and use them as narrative vehicles.

"Curiosity" and "survival" probably because those root functions that exist in any intelligent creature, real or artificial, and determine a LOT of what they do, and don't do.
Get yourself a new puppy, it's going to explore it's environment, "curiosity"; and determine what's "safe" and "unsafe" in terms of the environment and it's actions, "survival". One could argue that those are instinctive vice intelligent actions, and at their most basic level they probably are. BUT, an intelligent creature will learn by association; fire gives off light, it is hot, that heat causes injury. A moth, being a very rudimentary intelligence will fly into a fire, your puppy certainly isn't going to voluntarily walk into the fireplace. Your typical computer (assuming it's self mobile) is more akin to the moth then an AIs puppy.
OK, perhaps I could have perhaps come up with a better analogy, but I'm sure you get my point.

The point is that intelligent creatures, such as primates, operate in many levels of complexity. Some other creatures are completely reactive and may not get a chance to learn from fatal mistakes. Who says a computer is more likely to act like a moth than a puppy? An advanced system should be capable ot assessing risk. You're assuming how a robot would behave based on your personal ideas of how computers work.

I'll intentionally avoid the philosophical implications of "I think, therefor I exist" because not even philosophers can agree on such things, so discussing them in this context is pointless.

That's exactly the essence of this. Not pointless at all. And there's much, much philosophy (of mind) written on this. This has little to do with the existance or verifiability of thought. It has to do with naturalization of the mind and realization of mental states. It was stated computers have motivations and personal agendas. But you said "computers simulate emotion" and suggested humans do not. That's a big separation, how do you know? That's the whole point. Assuming a functionalist perspective, if the mind is realized by its functions there is no difference between the medium that enables mental states. Depriving an artificial mind of mental states is, as has been called before, chauvinistic. How do you even know your own mental states correlate to your actual percepts? A skeptic would say "you might as well be a brain in a vat". The point (yes there is one) is that an advanced artificial mind, if truly advanced and dynamic, and capable of realizing mental states, wouldn't be as different from a human mind as you'd think.

They CAN be "arbitrary and both are unpredictable", they're not inherently so.
Compared to what is "humans are terrible at approximating probabilities and risk" true? The moth? The puppy? A computer?

Compared to actual probabilities, because there is a lot of cognitive bias. And this is actually helpful in certain decision making tasks, as Gigerenzer suggests. Ultimately what this means is that humans tend to act irrationally. So in order for a computer to act human-like, must it be always rational or sometimes irrational? Point being, how does this set a human and a computer apart? Is the difference that computers rely on their own "internal, skewed logic"? And humans rely on external, consistent logic? Even when playing the lottery and the perceived expected probablity changes, for independent draws?

I don't see anything inherently "obsolete" in CP2020's treatment of AIs, sure it could be expanded on, but why? I seriously doubt CP2077 will let you play an AI as a character. If it did CDPR would run into a nearly insoluble situation in terms of what restrictions and advantages could, should, it place on such a character? There's no practical reason for CDPR to even open this can of worms.

To begin with, the term "AIs" is awkward at best, but utlimately wrong. The treatment is fine, but makes it be less science and more fiction. I thought it was supposed to be down to Earth. CDPR doesn't have to enable us to play robotic characters, all I'm hoping is the treatment of intelligent machines will be subtle and educated, with at least some of the above considerations. Otherwise these creatures would become magical and not very different from the dragons they are replacing.

 

[Sardukhar]

Otherwise these creatures would become magical and not very different from the dragons they are replacing.

Ahhh... I don't think referring to them as "AIs" is going to make them any more or less magical, any more than incorrectly calling First Nations "Indians" did. From your average player's perspective, an "AI" is magical. Same as a microprocessor and a motherboard on their cellphone. People don't care.

Many, many people call inteligent machines, "AIs". It's the popular term, used in every media I can think of and used to explain intelligent machines to lay people.

It's fine if they use it, for 99.999999999999999 percent of the audience - but it will drive you nuts.

I sort of get you, though - every time someone calls a submachine gun, assault rifle, carbine or anything that uses some kind of automatic firing mechanism a "machinegun" it bugs hell out of me.

Such is language, though. It's shaped by those who speak it incorrectly, not those who adhere to accurate standards. Kids, usually.

 

[Suhiira]

I don't know how much he actually "researched" since nobody says "AI's". It's science fiction, so it's alright to speculate. It's just awkward when speculation is somewhat misguided. The best works do not reference methods, they assume the existance of certain elements and use them as narrative vehicles.

Keep in mind in the early 80's the internet was in it's infancy, so "research" would consist of going to the local library and skimming magazines (Popular Science anyone?) for articles related to the subject. And possibly a book or two if the subject was REALLY important to the game.
At least that's how I "researched" subjects when I was doing game design in those days.
It's rather inevitable that speculation will be misguided, especially when viewed with 20/20 hindsight 30 years later. Thus your desire to update how AIs are viewed/used in the game is quite reasonable. But keep in mind, they are a small part of a much much larger game, and whatever you come up with has to be something the layman can comprehend and use (game rules wise) fairly easily.
While we can debate the various pros and cons I suspect at least 50% of those reading these forums at best skim this stuff, if not outright ignore it. And the folks on this forum are Cyberpunk fanatics! We're discussing a minor aspect of game that's not even due for release for years.

The point is that intelligent creatures, such as primates, operate in many levels of complexity. Some other creatures are completely reactive and may not get a chance to learn from fatal mistakes. Who says a computer is more likely to act like a moth than a puppy? An advanced system should be capable ot assessing risk. You're assuming how a robot would behave based on your personal ideas of how computers work.

I'm taking the liberty of making a couple basic assumptions for the sake of simplicity/discussion.

Computer = A device that operates solely within the limitations of it's programming. The classic programmer dilemma: "It's doing what I told it to not what I want it to do!"
AI = A synthetic intelligence capable of altering it's actions based on information "learned" from it's environment or sources not included/referenced in it's original programming.

OK, NOT scientifically accurate definitions, but ones everyone can easily understand and use.

That's exactly the essence of this. Not pointless at all. And there's much, much philosophy (of mind) written on this. This has little to do with the existance or verifiability of thought. It has to do with naturalization of the mind and realization of mental states. It was stated computers have motivations and personal agendas. But you said "computers simulate emotion" and suggested humans do not. That's a big separation, how do you know? That's the whole point. Assuming a functionalist perspective, if the mind is realized by its functions there is no difference between the medium that enables mental states. Depriving an artificial mind of mental states is, as has been called before, chauvinistic. How do you even know your own mental states correlate to your actual percepts? A skeptic would say "you might as well be a brain in a vat". The point (yes there is one) is that an advanced artificial mind, if truly advanced and dynamic, and capable of realizing mental states, wouldn't be as different from a human mind as you'd think.

Again, there are no "answers" to this. While a fascinating subject it's just not directly relevant to a discussion of game mechanics, which by their nature MUST be abstract simplifications of reality to allow for game functionality.

Compared to actual probabilities, because there is a lot of cognitive bias. And this is actually helpful in certain decision making tasks, as Gigerenzer suggests. Ultimately what this means is that humans tend to act irrationally. So in order for a computer to act human-like, must it be always rational or sometimes irrational? Point being, how does this set a human and a computer apart? Is the difference that computers rely on their own "internal, skewed logic"? And humans rely on external, consistent logic? Even when playing the lottery and the perceived expected probablity changes, for independent draws?

While not necessarily scientifically accurate/correct I think we can, and should, assume an AI in CP2077 will act "logically", this more then anything will set them apart from human intelligence.

To begin with, the term "AIs" is awkward at best, but utlimately wrong. The treatment is fine, but makes it be less science and more fiction. I thought it was supposed to be down to Earth. CDPR doesn't have to enable us to play robotic characters, all I'm hoping is the treatment of intelligent machines will be subtle and educated, with at least some of the above considerations. Otherwise these creatures would become magical and not very different from the dragons they are replacing.

True, but "AI" is a term almost everyone knows. Correct or not it's one the game needs to use because it's rather impractical to require players to learn complex redefinitions of commonly accepted terms.

Elf = Pointy eared snobs that "know" their culture is superior.
Dwarf = Stout, hard working people with limited patience for anything without a clear practical application.
Dragon = Big, dangerous, tough, and generally see humanity as pests.

Not everyone will agree with the above generalizations, but most folks will accept them as a commonly held synopses of the terms.

Sure some games will skew these commonly held terms (Dragon Age City Elves for example), but that's one of the things that makes these games unique in a market flooded with clones of commonly held systems/beliefs.

---------- Updated at 01:35 AM ----------

Many, many people call inteligent machines, "AIs". It's the popular term, used in every media I can think of and used to explain intelligent machines to lay people.

Like it or not it is the commonly accepted term/usage.

I sort of get you, though - every time someone calls a submachine gun, assault rifle, carbine or anything that uses some kind of automatic firing mechanism a "machinegun" it bugs hell out of me.

You an me both.

 

[Gilrond-i-Virdan]

Many, many people call inteligent machines, "AIs". It's the popular term, used in every media I can think of and used to explain intelligent machines to lay people.

At least not when referring to androids. I agree though that it's more commonly used for HAL 9000 like entities, which aren't androids, but sentient programs themselves.

---------- Updated at 02:09 AM ----------

That's exactly the essence of this. Not pointless at all. And there's much, much philosophy (of mind) written on this. This has little to do with the existance or verifiability of thought. It has to do with naturalization of the mind and realization of mental states. It was stated computers have motivations and personal agendas. But you said "computers simulate emotion" and suggested humans do not. That's a big separation, how do you know? That's the whole point. Assuming a functionalist perspective, if the mind is realized by its functions there is no difference between the medium that enables mental states. Depriving an artificial mind of mental states is, as has been called before, chauvinistic. How do you even know your own mental states correlate to your actual percepts? A skeptic would say "you might as well be a brain in a vat". The point (yes there is one) is that an advanced artificial mind, if truly advanced and dynamic, and capable of realizing mental states, wouldn't be as different from a human mind as you'd think.

Well said. That's exactly the essence of this, and not pointless at all indeed. The subject is vast from what is existence to begin with, to the question of free will.

And in addition to various mystics and philosophers throughout history, quite a number of science fiction authors discussed these subjects too. Such as Stanislaw Lem, Isaac Asimov, Philip K. Dick and others.

 

[Sardukhar]

At least not when referring to androids. I agree though that it's more commonly used for HAL 9000 like entities, which aren't androids, but sentient programs themselves.

---------- Updated at 02:09 AM ----------



Well said. That's exactly the essence of this, and not pointless at all indeed. The subject is vast from what is existence to begin with, to the question of free will.

And in addition to various mystics and philosophers throughout history, quite a number of science fiction authors discussed these subjects too. Such as Stanislaw Lem, Isaac Asimov, Philip K. Dick and others.

Yeah, let's not forget the place of science fiction - to inspire, to innovate, to examine. Many writers have also been scientists and you decry sci-fi's value in the face of people like Arthur C Clarke, Vernor Vinge and Carl Sagan, off the top of my head. One of my favourites, Peter Watts, a marine biologist, wrote the very hard science fiction Hugo nominated Blindsight, and it's darker and more ominous than any 10 fantasy novels I can't think of right now.



Cyberpunk treats androids very, very carefully, preferring to avoid the subject altogether. Many players and Refs look at the tech, look at the AI rules and wonder why you can't build an android. Perhaps you could.

But the setting is near-future, low-life high-tech. Androids, synthetic humans, are just that extra step towards post-human that the genre prefers to flirt with but rarely commit.

Yes, I know they were fundamental to Blade Runner, among other stories, but even in that, they are very, very human and placed in an extremely low-life kind of situation.

I would bet that CP2077 avoids android life as well.

 

[oconnorawest]

http://spaceacorn.com/2015/05/13/the-difficulty-of-artificial-intelligence/

A good read on AI