It's more than just computation, I think. Today, compared to true AI, the algorithms we use are deterministic; there's no "leeway". It's practically impossible to make "polymorphic" programs work as AI with current technology. ML, GANs, NLPs and whatever are just toy models of AI, they don't actually learn - they just use huge amounts of training data for their generalized function. For example, how do you train an NLP AI to distinguish - in the same way humans do - between a "border collie" and "poodle" belongs in the same group as "dog"? We're basically just using short-handed tricks to solve these things. They work, until they don't. At least I think so. Memristors might potentially change all that significantly - although I don't know well enough about them. That or quantum computers.

(September 28, 2020 at 4:36 pm)The Grand Nudger Wrote:

(September 28, 2020 at 3:44 pm)Rhizomorph13 Wrote:  Why wouldn't this process occur within cloud computing hardware? or the internet at large?

Cloud computing hardware (or software) isn't subject to evolutionary pressures, and has no particular need of the control schema that biological entities possess.

Think bigger.  They'd come out different.  Fit for their environment, from their construction.  Capable of doings things we do in their own way (and better than we do them) - like so much else on earth.  A plant is an incredibly sophisticated analog computing device, for example - why can't that be what's happening in the cloud, if we have to draw the analogy?

Evolutionary pressures? Do you mean things limiting the success of some entities? If so, I know how evolution works and it is through successfully propagating that a thing is expressed at a higher frequency. If you are saying that there would be lower flora and fauna, of course, I agree. There are literal examples in the AI of virtual plants, animals, and enemies within games. They are of varying degrees complex and successful. I imagine that there could be more things going on that we interpret as errors that the software/hardware is hiding from us either with or without intention.

(September 28, 2020 at 4:45 pm)Sal Wrote: It's more than just computation, I think. Today, compared to true AI, the algorithms we use are deterministic; there's no "leeway". It's practically impossible to make "polymorphic" programs work as AI with current technology. ML, GANs, NLPs and whatever are just toy models of AI, they don't actually learn - they just use huge amounts of training data for their generalized function. For example, how do you train an NLP AI to distinguish - in the same way humans do - between a "border collie" and "poodle" belongs in the same group as "dog"? We're basically just using short-handed tricks to solve these things. They work, until they don't. At least I think so. Memristors might potentially change all that significantly - although I don't know well enough about them. That or quantum computers.

See that's the thing; AI won't, necessarily, be born from our efforts directly. I imagine an AI scaffolding itself together by sheer chance through collective errors that successfully evade our detection.

Hmm, maybe, if the environmental conditions are just right. It's probably comparable to all the conditions on the ancient Earth of the chemical soup going from simple organic chemical reactions between carbon and other stuff, and just by stochastic process producing more and more complex molecules ... you know the story.

(September 28, 2020 at 4:59 pm)Rhizomorph13 Wrote: Evolutionary pressures? Do you mean things limiting the success of some entities? If so, I know how evolution works and it is through successfully propagating that a thing is expressed at a higher frequency.

I mean evolutionary pressures, not an evolutionary analogy.  As in the process that lead to human beings has little to nothing in common with the process that leads to computers.

The question was why they wouldn't end up like us, they aren't subject to the same circumstances or process.  Differences in circumstance and differences in process are expected to create difference in results. 

Those examples you pointed to as lesser forms aren't really a different form of anything even in context.  They're all the one entity, multitasking, directing it's faculties.  This is what cog sci researchers are referring to when they describe attention schemas, and we program them into games when we want the contents of the game to remind us of life, but it's not clear why or how an attention schema would arise in a machine that doesn't require an attention schema to function, even if there were some process we could credibly call evolution at play in their development.

I think that if we wanted to have an ai that seems like us in any relevant and meaningful sense (even to the trivial point of hiding things), we'd have to explicitly create one.

Well, that's just like, your opinion man.

I see what your saying, but there isn't a reason for us either, we are just the thing that succeeded to replicate itself so reasonless, it is a point of fact that we... are. It stands to reason that something could birth itself either with intention from us or tangentially, from our totally unrelated efforts to produce AI.

There is, or at least attention schemas provide one reason for us™. At a granular scale, we exist and exist as we do because our faculties lack full access to the internal model - our experience is silent on the particular neurons firing or the order or the amount or even that neurons exist. That's how the brain is doing things - but it's not telling us any of that.

That is an incredibly dumb way to go about intelligence, and that's why we've had to construct so many machines to help us think. Machines that do have full access to their internal models. Machines that accurately report what they're actually doing. We may have found utility in surviving by approaching intelligence in this way, but our machines have found greater utility and greater ability, even, by having been constructed much differently than ourselves.

Machines don't have to follow the rules that govern biological life - but spitballing for fun... It would be strange to see a machine evolve, at least in this context..downwards...and then become sentient. The ones that provide reports like ours (and they demonstrably can be made to provide those reports) are malfunctioning, in a literal sense.

Yeah, I've done a bit of coding and some of the things I've gotten computers to do by accident has been interesting and that is why I wonder about what might be going on between the reported clock-ticks. I push buttons on a program and it takes a certain amount of time to open a file for example. Some random other time it will take ten times as long opening the same file. What happened?! Like was the computer in a bad mood? NO it was the Silicon Natives highjacking those clock ticks and I caught them!

I saw a program that connected an AI to haphazardly assembled servos and the parts had MEMs devices on them to track tilt and acceleration and the AI would create a model of itself and then do the thing it was told to do like move to a specific location. It would do this by moving its parts according to the model it built. What was funny about the sequence I watched was how the mechanical robot had four legs so I was expecting it to walk like a dog or something but instead it rolled three legs in and used the fourth to push off and roll towards the goal.

I've seen many examples of large programmable matter and I think they will eventually shrink it down to be useful at much higher resolution and that would create actual discreet units that in pair with a slightly damaged AI could rattle against itself and create its own goals and splinter into different actual robots that would develop purpose through an accidental process of trial and error.

That's a body model. We have one of those, it's muuuuuuuch better than our attention schema. OFC, machine body models are orders of magnitude more precise than ours. Ultimately, they do the same thing (in both cases), control.

Again for purposes of discussion we're going to assume that the scenario above is not only possible but inevitable, but here again I'd wonder why we would expect anything like us to come of it. We aren't the product of an accidental process of trial and error anymore than a digital thermostat is a product of biological evolution. We're what's left. Consider how different your personal experience could be if you had access to the amount and types of data that a machine has access to weith regards to it's internal models. Any machine that was aware, in that way, would be as different from you and I today as you would be from yourself a million years from now (or however long it took to breed or design people with that ability, if it's even possible).

Qualitatively equal reports of subjective experience may be incomprehensible between us. Rather than being the lone intelligence in a world full of inanimate objects, we are alone in a world surrounded by intelligence. Rather than being the lone experiencer, we are alone experiencing in a world full of conscious entities.

Natural selection and other related stuff, yeah.

(September 28, 2020 at 6:11 pm)The Grand Nudger Wrote: That's a body model.   We have one of those, it's muuuuuuuch better than our attention schema.  OFC, machine body models are orders of magnitude more precise than ours. Ultimately, they do the same thing (in both cases), control.  

Again for purposes of discussion we're going to assume that the scenario above is not only possible but inevitable, but here again I'd wonder why we would expect anything like us to come of it.  We aren't the product of an accidental process of trial and error anymore than a digital thermostat is a product of biological evolution.  We're what's left.  Consider how different your personal experience could be if you had access to the amount and types of data that a machine has access to weith regards to it's internal models.  Any machine that was aware, in that way, would be as different from you and I today as you would be from yourself a million years from now (or however long it took to breed or design people with that ability, if it's even possible).

Qualitatively equal reports of subjective experience may be incomprehensible between us.  Rather than being the lone intelligence in a world full of inanimate objects, we are alone in a world surrounded by intelligence.  Rather than being the lone experiencer, we are alone experiencing in a world full of conscious entities.

I don't necessarily think AI will look or act anything like us although we will be a great decider in which AI get to exist if it doesn't adapt to stealth or overpowering us. We will cull the litters of whatever comes out of the process until we either can't or are happy with what's happening much like we did with dogs or any other domestic animal.

I would put my kinesthetic model up against what I've seen in that video or any other VR model of hand placement with all its jitters and latency. ANY DAY!

I think things are going to get very interesting in the next 40 years if we don't burn the planet to ash.