[Discussion] What is AI?

I missed a good portion of that a few decades ago and rarely have an excuse to think about it so my knowledge is foggy at best.

 

If I rememebr correctly anything above viruses is life, and viruses are not life (they lack some properties).

An intelligent systems human-created system by default will not be a lifeform, so we might end up creating some sort of equivalent of a robo-lich or undead.

As for alternative lifeforms, if I remember correctly, some alternatives were using methane instead of water, plus there was a possibility of plasma emitting required properties in certain circumstances.

 

Ants are adaptable, they will change their 'jobs' in ant society based on their effectiveness at carrying out their responsibilities.

Intelligence is a complex subject and people tend to define it too narrowly.

An individual ant is not smart, an ant colony is shockingly intelligent.

 

Also - in general terms - people dramatically disrespect animals. Most animals are far far 'smarter' than we give them credit for.

The differences between us and a cat/dog are not nearly as wide as we would generally like to think.

 

I defined intelligence as adaptability at one end (the upper end) of the complexity scale, so the fact that ants are adaptable does not necessarily make them smart.

Without knowing too much about ants, I would be inclined to think that intelligent behaviour is dependent on the group as a whole - i.e. an ant needs feedback from many other members to act in a way that appears intelligent. In fact as far as I know, most 'intelligent' behaviour of an ant is not intelligent in isolation, but rather only in terms of how it fits with the direction of the group - that's to say that the ant's behaviour is not complex as such but rather it's unexpectedly appropriate for the situation given what the other ants are doing.

So with that I would say that a group of ants represents a degree of complexity, in somewhat the same way that a group of neurons represents complexity, while neither a neuron or an ant is particularly intelligent.

Also I don't believe that intelligence is something that animals do/don't share, rather it's a property of every living organism by a matter of degrees, with humans providing the best representation of it so far. Intelligence as a concept though, is represented IMO by adaptation toward the upper end of the complexity scale - by complexity I basically mean the ability to adapt through modelling/planning rather than a mechanical reaction or inbuilt hardiness.

 

You are a hive mind - a colony of cells that evolved to the point where they can't live without each other. But despite that you're a group and not that different from an ant colony.

What now?

 

Not sure what you mean? But I'm not saying that an ant colony is anywhere near as intelligent as a human being. I simply meant that whatever amount of intelligence is shown in ant behaviour, is a property of the group as a whole, and so an individual ant is probably not much more intelligent than other insects.

In terms of a human brain being a hive mind, I think it is absolutely the case. You don't event need to go to the level of neurons - there are many 'macro' systems in of the brain that act as fairly individual entities which (most of the time) work together to chew information, add/remove from it, and produce an effective result, the effects of which are somewhat deceptively stitched together into a singular and spontaneous experience of a universal 'reality'.

What I wanted to say in response to @frosted 's comment was that I don't see the ant colony as being different from the complexity of a human brain in principle.

 

Poked around the interenet, found an article on ant intelligence:
http://jonlieffmd.com/blog/ant-intelligence-update
And bee intelligence:
https://phys.org/news/2012-05-insects-master-abstract-concepts.html

Also, it turns out that an ant has 250000 neurons.

I agree with that. People appear to have the need to prove that they're superior to other species. That leads to idea that "other animals are robots". Then they notice that rodents can see dreams.

 

Like I said, I don't know too much about ants, which is why I described them simply as being probably around the same intelligence as any other insect.

Of course I'm not comparing an ant to a single neuron. I'm was just illustrating the point about the relatively high intelligence of an ant colony probably being a function of the group - which may not be correct, since I don't know much about them.

But anyway whether one creature is intelligent or not, is beside the point - my point is simply that intelligence as a concept is related IMO to adaptability with complexity, rather than inbuilt hardiness. The degree to which an AI can be said to be intelligent is related to it's ability to adapt to novel situations using generalized learning rather than a simple reactionary system such as "if raycast hits object, turn around".

 

Pretty much, yeah. A relevant term is "superorganism."

 

Intelligence and superiority don't necessarily coexist. I think it's pretty conclusive that humans are more intelligent, but whether or not it makes them superior than other species depends on your point of view. Nature still hasn't given a clear point of view, and probably never will.

 

Well, there are opinions that dolphins may be more intelligent (bigger brain with larger surface area and larger amount of "wrinkles"),

However, people usually go on defensive when hearing this idea.

 

I am a big fan of ants. As someone who was deeply interested in machine intelligence, I did a ton of reading about them (although I've forgotten most of it).

They're really fascinating.

I also think that our view of the "ant" model (bottom up) is a much better model for machines than our view of the human model (top down).

I also don't think that intelligence and CA is really fundamentally different. But that might be a bit more controversial.

 

So long, and thanks for all the fish...

 

Yeah, I read that book.

 

But what is cognition? the ability to react to our surrounding to achieve a goal (generally survival)? Then some cellular automata have this property. I would argue that life itself is the most primitive form of advance intelligence.

We do understand what intelligence mean, and it's only the mystic surrounding hi level intelligence that fools some people, the difference between human and monkey is tenuous, both in intelligence and dna, we only have access to a few more process that have a high pay off (like the ability to ask question).

If we see life as being intelligence, then the smallest process of intelligence is natural selection, which can be seen as a motivated random local search, motivated because it tries to optimize homeostasis, local, because it search only immediate neighbor "hypothesis" to his state.

If we accept this life as intelligence hypothesis, everything start to make sense!

Let's assume reality is a gradient of bad conditions to good conditions, for example water temperature for cells (you want to avoid destructive extreme and be into the median).

- Fear would be the first safeguard: avoiding place where the gradient destroys you
- Joy would be the second safegurad: moving to place that is safe
- Memory would be the third: identifying pattern in time that allow moving/avoiding place
- Curiosity would be the fourth: seeking knowledge to anticipate and maintain the database up to date.
- Social would be the fifth: mutualizing knowledge.
- Logic would be the sixth: generalizing on knowledge (ie optimizing and maintaining coherence).
- Consciousness would be the seventh: being able to reflect on self and knowledge, to go beyond the limit of time.
- Metaphysics would the eighth: being able to reboot the knowledge and logic of existence. Notice that metaphysics mostly happen in period of crisis, when logical dead end happen but goal is threaten. Crisis happen in extreme condition or extreme boredom or happiness, it's a hi level version of curiosity.

It also happen to mimic the evolution of the brain's structure, while it's not confirmation it's a strong indicator there might be some validity to this hypothesis.

 

As far as I'm aware we don't know, and don't understand. A good example of that is IQ test which was originally designed to measure intelligence "level", after which point a lot of flaws were discovered with it.

Also...

This is a misunderstanding of natural selection.
Natural selection is completely brainless. Things that don't live long enough and don't reproduce fast enough don't get to pass their genes on. That's all this is about - there is no intelligence involved in it.

One of the "perfect organisms", for example, is an antarctic sponge. This thing has no brain, and can live for 10000 years. There's also immortal jellyfish, or tardigrades and various microsocopic organisms that didn't change much because they were already perfectly suitable for their environment.

I'll reject it. It doesn't make sense.

This example is not how natural selection occurs. Natural selection means that over time some organism may slowly adapt to any conditions present. Too hot? Well, it is not too hot for something, and over time that "something" will move to the "too hot" area and starts getting better and better at living there.

Also:

Life require no brain, no intelligence and no emotions.

Speaking of which, smbc had a funny take on human intelligence:

Spoiler

http://www.smbc-comics.com/?id=3476

 

In all honesty I think words like "intelligence" and "cognition" are kind of like "AI". Once we start to understand how exactly systems work we stop calling it "AI", the same happens to words like cognition and intelligence in general.

 

I think there may be an underlying principle in "cognition", that all life with a brain follows. Meaning that perhaps at its core an ant, a cat, a hamster and a human may be operating in the same fashion, with increased processing power allowing for extra... bonus behavior. Would be nice to know what that underlying principle is, though.

On a side note, a human has 344000 times more neurons than an ant. However, the largest colony had 300 million ants in it.

 

@neoshaman, I think @neginfinity said everything I could or would.

As for the idea of cognition...I can definitely understand the idea that it's literally nothing more than firing electrical impulses across synapses...but the gap between our required knowledge to come to this conclusion and where we are now leaves me too skeptical to accept it completely. For right now at least.

 

I don't know about that lol natural selection is a process, it happen it process information, as long as you delimit the entity with an internal state and an external state, natural selection become effectively "intelligence"

In fact in AI we use montecarlo search that is barely different, except we search in an hypothetical space instead of a real one. In fact many outperforming process use random sampling like simulated annealing. natural selection is suing a sampling process call mutation, it map the immediate possibility space around the current state, survival effectively select the best hypothesis and the process start again at the next generation with the state moving to the next "fit" states. I mean how is that different than any other "intelligent process"? you measure your surrounding with hypothesis and select the best answer, that's best first search that's a*, that's backprop.

They are all basically advance version of the same process. The only things is that advance cognition can simulate the possibility space, so all move happen in that simulation space instead of the physical space, which allow for more abstract and faster compositions instead of processing that takes as much time as spectrum ZX Natural selection isn't brainless, natural selection IS the brain

Now defining consciousness is harder I swear.

 

Let's not just focus on higher intelligence here. It's easy to forget that biological intelligence systems are doing a bunch of other things too. That includes breathing, digestion, locomotion, reproduction and so on.

When an AI can do all of that, and carry out some higher level functions, and do it on a tiny amount of energy, then I will be impressed.

 

I would say it's different because there's no "measuring" taking place. Selection is a completely blind process.

This feels like the Chinese room, aka CA, again. We see higher order complexity from an incredibly simplistic system and perceive intelligence.

For fun (though far-fetched) reading along these lines, I recommend Michael Crichton's Prey.

 

This lol alphago won with 3000w vs Lee Sedol at 30w and a glass of water lol

 

That's also our brain and our brain chemistry system.

I don't see intelligence because it has complex behavior, I see intelligence because I defined intelligence from a process of information processing that use sampling hypothesis toward a goal, as such natural selection fit the bill.

 

That's not quite correct.

Natural selection does not search for anything. "Searching" implies goal. There is no goal.

In essence, natural selection is a process of murdering as many living things as possible.

Basically, as a species you spend millions of years adapting to the environment and fighting enemies, and then you eat the meteor. Result? You're dead, and rats took over the planet. That's how natural selection works. No logic, no purpose, no goal.

 

We don't understand it well enough to say that definitively. Like, at all.

Well this is probably part of the problem, like @neginfinity said in the other thread, dictionary wars. But even if we accept that definition, we don't know enough about the human brain to claim it's that simple in any way, shape or form - and the fact that we can't replicate it would imply otherwise (at least until we do).

 

Technically yes. We can't describe every single process in the brain well enough to identify a single thought.

But we don't have any reliable evidence to suggest that we are anything other then complex chemical reactions. In the absence of any evidence otherwise, it's fairly logical to say intelligence is just our brains and chemistry.

 

On what do you based that assumption we don't enough about our brain? We use the model of the brain to create rather successful artifact called neural network that is rocking AI right now.

While we don't understand everything (cough consciousness) we understand enough to emulate basic "cognition" happening, like being able to recognize cat and describe what a person is doing with actual sentence or driving a car in the real world. It's not the glamour, but that's rather good for something we supposedly don't understand lol. And we even figures out that these simple model "dream".

I would say we know "enough", not enough to replicate a human, true, but enough to understand how it does basic cognition.

 

We haven't made a brain yet. Or a brain-machine interface that could read brain contents. Our neural networks cannot think.

Which does not act like biological neurons.
Neural network is missing neuroplasticity, and response to chemical stimuli.
https://en.wikipedia.org/wiki/Biological_neuron_model

A programming model is a solution that describes the problem well enough for practical purposes. Neural network is an oversimplification that is still useful for specific practical purposes, but it does not behave like a biological network would.

-------

Now, in all honesty, all this stuff is being researched, but still isn't finished.

When someone will make a robotic dog that will act like a real thing, then we'll be talking. Or at least a robotic hamster.

 

sensory input processing is huge too. The stuff that different animals are capable of (racoons build sense of high detail topography using their hands for example) is pretty amazing.

 

I strongly disagree. The logical answer is "we don't know." To move in either direction is to make an assumption, which requires a leap of logic.

Sorry, that's mainly a response to "it's fairly logical." As for the idea itself, one can certainly use it and build up a framework around it. I can't, though. Both directions require a leap of logic, and the leap of logic in that direction seems too far to me.

 

It's a matter of perspective, and anyway I'm parroting other people who research AI so That's not me saying anything

Because if that's true, then it's also true of our existence, and that's one big "metaphysic" discussed in philosophy.

 

An idea I just had: could the placebo effect be seen as a sign of intelligence? It's all about us perceiving on a "higher level of complexity" some variable state and by that influencing lower levels, or at least things we don't have any direct control over. It REQUIRES some type of higher-order cognition (almost a sense of imagination, to comprehend the intended effect of medicine or something). Just read this article about how some animals (including mice) might show the placebo effect.

Just a very basic idea, not sure if it's really applicable or not.

 

That's the debate of "low" vs "high" (conscious cognition), vegetative system (autonomic nervous system) is an intelligent system that control very complex process happening in the body without the conscience having direct access to it, generally we control and assess it's state it by having indirect access and controlling environment variables (relative to the system). For example, controlling breath to reduce stress or trying to have happy thought to not trigger automatic responses. When the brain dies, the body actually continue to function until it dies from hi level inaction (like feeding ourselves).

So is it intelligence? we can't even access the type of data processing it does right now, it's possibly more complex that we do consciously because it integrate much more variable. Experiment (and brain lesion) show that consciousness is generally about saying "no" to impulse and letting go those it agree with (ie it's a control system mostly), so who know?

Consciousness seems to be directed toward the external environment and this system on internal state. Different responsibility.

Disclaimer Im' not an expert, that parroting and opinion. I guess that's implicit to the format of the discussion but let's put it here lol.

 

IMO, the best idea is to speak for yourself instead of repeating after others. Partoting is how we get arowx threads.

 

I believe in them after processing the facts, so I put that too, we can't just invent stuff, I think none of us are AI expert, and we have just differing implication in knowing the field. And even then, there is a video where many expert gave similar answers and it map relatively well to all opinions here, even expert disagree so I would say nobody is more valid than one another lol, just differing degree of opinionated. I did speak for myself, only pointing I'm not the origin.

But anyway it's all in good fun, these discussion help me maintain a certain mental alertness lol Irl don't provide me this kind of stimulation lol

 

I don't claim to know anything about brain death, but the first link i got contradicts you @neginfinity and agrees with @neoshaman

https://www.livescience.com/42301-brain-death-body-alive.html

again, i claim zero knowledge of this subject

 

Nah. Some assumptions are better then others.

To accept our brains as just chemistry, we need to make two key assumptions.

• Current science so far is true.
• Science we will discover in the future is more of the same.

Historically in 99% of cases, those assumptions have held correct. There are a few cases where we got things drastically wrong and had to rewrite most of science to accomodate. But these are rare.

To accept that our brains are more then simple chemistry requires the following assumptions.

• Current science is wrong
• Future science will go in a different direction to current science
• There is some other principle of intelligence that we have not discovered, despite thousands of years of searching.

As far as I can tell the only ones who hold out in that regard are the religiously orientated.

So while it's technically possible we will find something other then chemistry drives intelligence, it's an incredibly long shot. It would be as dramatic as when we discovered we were wrong about the speed of light, gravity, the atom or sub atomic particles.

 

I'm not fond of arguments from authority.
If expert says something, the right idea is to pick their opinion apart and attempt to understand why they think this way, not just repeat after them.

 

Key point - "on ventilator", meaning sustained by machines only.

Anyway this is not a topic I find enjoyable to discuss.

 

I would say that's not argument from authority when I use that to demonstrate there is no authority?

We have brain prosthetic that read muscle command neuron.
Neural network works now reasonably well.
"Think" well that's semantic anyway lol.

The argument is that we understand enough to replicate simple function. The problem is semantic about where do you draws the line, is enough or not enough? I would say if it wasn't enough we would have no result even with simplification. Else it's just loki's wager.

 

The only thing I know for sure is that my mind presents me with a believable reality. It's pretty difficult to take anything as solid evidence when, for all I know, I just made it up. Have we gone far enough down the wormhole to turn around and head back to talking about AI?

 

Well, smashing certain things together can be used to produce fire. However, being able to do so does not mean understaning of atomic structure of materials involved or reason of the fire.

When you can replicate behavior of a biological system. Or when you'll be able to read data from the brain, such as memories. Or we make a brain. Or when we simulate one.

BlueBrain projects lists scientific discoveries made few weeks ago. We're only scratching the surface of it.

 

Well we can retrieve some memory and some concept mapping, and we know what region produce what, Convolutional neural network mimic the visual neurons pathway and structure and produce similar artefact to what some studied cat visual neuron did when we were stimulating them ...

We know more than you let it believe, less that you want maybe.

 

Not really. This thread was a wormhole from the beginning.

 

We haven't even discussed what's the nature of information, since it's what the brain process, what does that mean? I won't go there because this is even more out of common sense than intelligence and consciousness (ie it's mass energy).

By the way what we see isn't real, the color magenta has no equivalent in reality for example, if light is en electromagnetic wave, and color different wavelength, then magenta isn't a wavelength at all. What other perception has no reality?

 

I think it's more important to ask the question of what is the purpose of AI.

Edit:

 

Ha.

NOPE WE CAN'T.

Convolutional neural networks brute force for solution without actually giving you any explanation regarding how this solution works. "It kinda sorta looks like the original" does not mean informational retrieval. Good luck finding any document describing data structures involved in vision processing within human brain.

There's a prosthetic eye being approved that acutally encodes received data in the way human eye transmits it. That is an example of actual understanding.

Honestly, this is getting depressing I've pretty much lost the interest at this point. I would prefer to get some new information out of this, and not more of semantic argument nonsense.

 

To kill all humans, probably.