Artificial Intelligence

[TheManOfOneBook]

How would an 'Artificial Intelligence' know that it is indeed an AI and not organic intelligence? Would it be able to know the difference?
Is it supposed to know the difference?
Would it even ask that question?
If what constitutes an AI is the fact that it took intelligence to create it, how can we be so sure that the creator of this AI is not also an AI?

How would you know that you are not this artificial intelligence?

[Daxterwous]

Very interesting thought!
I think (im no expert on any level) that the AI can be split into two sections:
Artificial and Intelligence (rather straight forward so far).

The artificial part only implies that it was created by some entity that has a different creation process than the AI entity. If we humans create a robot with high level AI, we would call it AI because it is intelligent and it is not created the way we are. It could be identical to us, but because we created it in a way other than regular birth, it is AI.

Intelligence is hard to define, but if we consider ourselves intelligent, we could be AI in the eyes of another being. Perhaps some "super entity" created us (in a different way than the "Conversation with God" story implied).

This could mean that some day, AI (in the form of robots or cyborgs or something else) could begin an own life.
What could happen, and there are a lot of theories about this, is that one day robots could become like us. And I mean REALLY like us. They could evolve themselves, simply by upgrading, and create a real new race. They only difference then is that we are made of different parts (and even that doesn't have to be true).
So then we need to apply ethics to robots. Robots will vote..... etc

But there is one thing which stands in the way. People
People get scared when an artificial entity becomes like them. on example is Uncanny Valley theory .
Also, there is the whole religious front which will go nuts if we create humans in other ways than their god intended...

So lets end this with another question: Do you think we will reach that future? Where robots can be like humans?

[HarryStottle]

No.

I don't think we'll reach THAT future (robots exactly like us). Why? Because it's a very wasteful use of such intelligence. And - as Daxterwous suggests - it will fuel paranoia. (Is he really human or is he one of those android thingies??)

But first to "TheMan"'s questions. To begin with, apologies for not spotting this earlier, it was only Daxterwous' reply that alerted me to its existence (or perhaps reminded me).

AI doesn't begin at Human Level
The first point to make about AI is that it's just as scalable (eventually more so) as organic intelligence. People are stuck on the Turing Test and Androids when they think about AI, because they equate AI with human level intellect. But that level is merely where AI starts to get sexy (and wait till they REALLY get sexy; that's going to change a paradigm or two).

In the real world, we've already got AI working at lower levels, equivalent to the Bee, the Mouse or, with the current , the Cat. Each of these examples are as intelligent as their organic equivalents. (Or so the literature claims, I haven't had the opportunity to try one out for myself)

Thus, in the same way much organic intelligence isn't self-aware, all early AI will also "fail the mirror test" (or its AI equivalent). But once AI reaches our level, the answer to your question becomes simple. We must be able to answer "yes" to all your questions, or else the AI has NOT reached our level!

The artificial Cat Brain would, of course, not know that it was an AI, or even what an AI is. It would thus not be able to tell the difference and not even ask the question. But it is definitely Artificial and it is definitely Intelligent (subject to what we may define as Intelligence - see below). I think what most people really want to know is: "Can AI ever actually reach our level?" And the answer to that has been a firm "Yes" since 1950 and Alan Turing's musings on the matter which resulted in his famous "Turing Test" as a major hurdle for AI to overcome - if it ever is to persuade us that it has reached our level.

I Don't Like The Turing Test
I actually prefer your questions. I've never liked the Turing Test. It's a distraction. It tries to answer a very narrow question which is not: "Is this machine at least as intelligent as a human being?" which is the question that interests me; but rather, "can we construct a machine which can fool an "average" human into believing that it is interacting with another average human rather than machine?" Such a machine would - for example - fail that test if it answered EVERY question "correctly" because no average human ever would. This leads to the absurd situation where contenders to pass the Turing Test actually build deliberate errors into the machine's answers in order to help it "play the game". Which surely misses the whole point of the exercise!!

I'm not even vaguely interested in whether a machine can "fool" me into thinking it's human. I'm only interested in whether has an intellect I can recognise, without obfuscation, as being similar or superior to my own. I also want to recognise that it is obviously self-aware. But here's the crucial bit for me because I think this is the "Hard Problem" for AI. I want to able to recognise and probe its motivation, it's "intentionality".

So I will only consider an AI has reached our level when it CAN do exactly the sort of thing you ask. I'll believe it's worth having a conversation with when an AI turns to me and says something like "Damn. I've just realised I'm intelligent, self aware, autonomous and not organic. I must be manufactured!" Even more convincing would be to observe it asking questions, autonomously and in response to interactions, finding answers and then being able to explain what it had found to other entities - without direct "cutting and pasting" - in other words, like a human would, in its own words.

For example, if a child and an adult separately asked an intelligent android a question like "why do the planets revolve around the Sun", it would research the answer (assuming it hadn't previously done so) and then construct different explanations suitable for its audience. It wouldn't use terms like "heliocentric" with a 7 year old, or try to explain the laws of motion. It would construct the answer appropriate for the person who asked the question. And, for me, it is identifying that ability to combine knowledge with awareness of "the other" and "context" to deliver an appropriate (rather than merely correct) response which would mark out human level intellect.

But even that behaviour wouldn't illustrate motivation. That would still only require "response to stimulus", even if at very high level. If such an android did nothing except answer such questions I would regard it as no more than a "super google". Indeed, Google will probably be the first to offer such a service and I wouldn't be surprised to see version 1 in the next decade.

But if I found the android conducting it's own research, unprompted by human requests, or wandering around looking for people who had problems it could help with, then I'd start to be really impressed. When it starts complaining about the stupid decisions we are making ourselves, I will consider the first landmark has been reached. At that point I'd allow it to take part in the Democratic process.

Most of the AI community don't like addressing that sort of issue, but they have long since stopped asking the question as to whether AI will get there. For them it's a "no brainer". There is no law of physics which prevents it, so somewhere, sometime, Human Level - and subsequently superior - AI definitely will be created. But they obviously haven't carried their conviction into the world at large. For the "average human" the notion of better than human level intelligence is still fantasy or sci-fi. They don't have the faintest idea how close we are to achieving it. I am extremely confident it will arrive within the lifetime of humans alive today. With luck within my own lifetime. But very few humans (so far) have dared to admit or accept that. The implications bring them out in a cold sweat.

This is one of the points I touch on in "Conversation With God"; that organic intelligences have a profound belief that their particular functionality is only possible in an organic brain. The notion is incredibly naive. True, it is sometimes difficult to emulate organic functionality efficiently with a machine. But there are no examples (that I'm aware of) where, having allocated appropriate resources, we have failed to make sufficient progress towards such emulation and thus demonstrate that it is achievable. Look, for example, how difficult it has been to get robots to walk like humans. But then

Speech Recognition is another example. It still sucks. But its about 10 times better than it was 10 years ago. Another 10 years and I might even be using it for more than entertainment.

And as of now, we are beginning to suffer our . The article is behind a pay wall so I'll be posting in full for a related reason later. It tells the story that Chatbots have become good enough to conduct online "social engineering" and weadle dumb humans' credit card details out of them. It raises the point that this means, in the very near future, either we're going to have to have a foolproof way of identifying AI or of identifying humans and the author suggests (grossly irresponsibly in my view) that this a job for Governments. They take responsibility for identifying us in the "real world" with passports and the like, it's up to them to take charge of identifying us online. That's the stupidest and most dangerous suggestion I've read in quite a while.

Are we living in a Simulated Universe?
As to the existential questions, such as whether we are AIs (or indeed, our entire Universe is Simulated by an AI) and whether we were created by a superior intelligence (which may, itself, be an AI) that is one of the questions I address in part 1 of my play "Resurrection". It is precisely the reason that part is called "".

And the questions surrounding the issue of whether you would be aware of your own artificiality are dealt with in detail by the conversation we had about whether or not the "real you" would be the result of "mind uploading", "digital cloning" or whatever we eventually decide to call it.

Personally, the key question for me, given that I hope to be one of the first generation to experience digital cloning, is "how will my digital clone convince me that it really is my digital clone". Only when it has passed that test will I feel safe to let go of my organic remains and pass into my new digital existence.

What IS Intelligence?
Daxterwous raises the more fundamental question of the definition of Intelligence. I gave my own views on that in my musings on so I don't intend to repeat it all here. The important point is that intelligence is nothing magical or mysterious. It is the ability of an entity to make a choice between two or more options. And at that lowest conceivable level, intelligence is inanimate and wouldn't be recognised by most as any form of intelligence. Where the majority tend to come on board is with animate intelligence - the level at which matter becomes capable of pursuing a goal (Survival) rather than merely being nudged around by forces in the environment.

A living thing can hide from those forces, or exploit them. It can hide from predators or search for prey etc. Whatever the living thing is doing, however, requires much more intelligence than its inanimate predecessors. And it is "real" intelligence. The stuff nobody can argue about. There needs to be, for example, some kind of memory or cleverly programmed algorithm which allows a bacterium to recognise the difference between food and foe.

I recently watched Chris Packham's on the Beeb which took a slightly different angle from most in that its focus was on the web of interconnections between species, such that - for example - the top predator (Bengal Tiger) in the swamps of Bangladesh is critically dependent on the efforts of billions of tiny crabs. The crabs aerate the otherwise anoxic mud and make it possible for the ecosystem to flourish sufficiently to support the vegetation which feeds the grazers which, in turn, feed the tigers.

That is a very intelligent algorithm which illustrates something quite profound about intelligence. "Thought" (and motivation) does not have to be part of the process! Those crabs certainly didn't plan to support the tiger. That's merely a consequence of their industry. But if you address the question from the tiger's point of view - "How is it possible for this environment to feed me?" - it has to be seen as an elegant and intelligent solution, despite the absence of cognition.

Which makes it much less of a leap than most people seem to think when we consider the question of whether we can create intelligent machines. It's literally a nobrainer. I'm writing this on just such a machine as we speak. It is highly intelligent but certainly doesn't have a brain. It is also incapable of THOUGHT which, all to often, is what people really have in mind when they talk about intelligence.

"Thought" is much more interesting than intelligence and much harder to define. So I'm going to leave that for another day!