patrick.net

EBGuy says

Tell that to Heisenberg....

You mean Walter White?

Peter P says

Then it is no different from another person perceiving the world.

And we go back to "other person perceiving the world" implying that there is a world.

Peter P says

Why? Your reality and mine are somewhat in sync. We all have our own biases, opinions, and narratives but I think the correspondence is there.

Your argument is crumbling under its own weight: yes I can communicate with other people and they see more or less the same as I do, which makes it very likely there is a common reality being observed. Maybe there is no bias in perception, or opinions, but that's irrelevant: the point is there is a common reality.

The mere fact that you are posting means you assume I exist.

Peter P says

And no, the machine does not have to form rigid models about the world. It will be a positivist/reductionist *only if* you program it so.

A photo in a computer is a model of an image. Every higher level notions will also be some kind of model. A computer thinking about high level concepts means it has a model of these concepts.
Not sure what you mean by rigid. If anchored in perception, then by definition the model is fluid and changing. But there has to be a model.

Peter P says

Practically, yes.

And that's all that matters. Sure, you could be living in the Matrix and it could be in your best interest to dive off the Empire State Building, but you aren't going to make any decisions based on that premise. Nor are you going to act on the possibility that the only way to save the universe from popping out of existence is to suck Ozzy Osbourne's dick. Sure, sucking Ozzy's dick makes perfect sense if there's even a remote chance that the universe will pop out of existence if you don't, but you act on the premise that there isn't even a remote possibility of that happening.

Peter P says

Not proof. But observations. The "fact" that "reality" fits into formulas further illustrates that it is a matter of linguistics.

If you get ridiculous, nothing is provable, not even a priori matters. You might as well argue that this proof that the square root of two is an irrational number is incorrect. After all, all proofs are based on assumptions, even this one is based on the assumption that there are no contradictions in mathematics.

If you're going to accept that nonsense then the word proof becomes meaningless and you end up not being able to do jack shit like formally verifying the correctness of the code running the nuclear reactor, so you don't, and the reactor goes critical killing millions. Best not to be ridiculous.

Heraclitusstudent says

A computer thinking about high level concepts means it has a model of these concepts.

Not necessarily. It can pick up nuances without structures. Concept is a matter of language. It does not require a model.

Mathematics is about taking certain things as granted.

Evidence still has a place in life. It alters the expected payoff of any bet or speculation.

Even formally verified systems can fail due to factors outside of that formal system. At some point, you have to draw a line and accept some unknowns.

Remember, the set of axioms is either incomplete or inconsistent.

Dan8267 says

Nor are you going to act on the possibility that the only way to save the universe from popping out of existence is to ...

To me, the universe going out of existence has the same negative payoff as me being struck by lightning. I still walk in the rain.

Peter P says

It can pick up nuances without structures. Concept is a matter of language. It does not require a model.

Vacuous pontification.
Explain how you can represent concepts, including language, in the memory of computer without structures.

Peter P says

Remember, the set of axioms is either incomplete or inconsistent.

Godel's theorem is restricted to mathematical first order logic in its common formulation.

All it shows is that there are paradoxes, which happen only for knowledge about knowledge itself. Paradoxes don't happen in the physical universe, nor in arithmetic.

Heraclitusstudent says

Peter P says

It can pick up nuances without structures. Concept is a matter of language. It does not require a model.

Vacuous pontification.

Explain how you can represent concepts, including language, in the memory of computer without structures.

Data structures are probably needed to implement various AI algorithms and supporting systems.

However, concepts are not necessarily modeled in formal constructs.

Heraclitusstudent says

Paradoxes don't happen in the physical universe, nor in arithmetic.

Only because the modernist/reductionist/positivist understanding of reality does not allow them.

Peter P says

However, concepts are not necessarily modeled in formal constructs.

You are aware that everything in the memory of a computer is a formal construct, right?

Heraclitusstudent says

Peter P says

However, concepts are not necessarily modeled in formal constructs.

You are aware that everything in the memory of a computer is a formal construct, right?

There are:

1. Constructs designed by human intelligence to implement AI, such as chips, memory, arrays, data structures

2. Constructs formed by AI to perceive/comprehend/reason/speculate

(2) are not necessarily formal structures.

For example, we need to use "formal constructs" to implement algorithms, e.g. a denoising autoencoder.

However, the machine does not need models about features to extract information regarding such features from an image.

Peter P says

2. Constructs formed by AI to perceive/comprehend/reason/speculate

(2) are not necessarily formal structures.

You don't know much about programming, do you?

Heraclitusstudent says

Peter P says

2. Constructs formed by AI to perceive/comprehend/reason/speculate

(2) are not necessarily formal structures.

You don't know much about programming, do you?

What kind of AI are you building?

Peter P says

What kind of AI are you building?

One based on programming structures, like other programs.

Peter P says

This is why we must live at the frontier of its development. He who controls AI controls the human destiny.

Not even sure if the frontier is needed. The current development of personal digital assistants, already have a lot of implementation arcs where in effect, an organization can add more work while also laying ppl off. In the past, that strategy usually failed because customers became aware that their support efforts went downhill, as well as the general quality of work. Thus, a loss of let's say 25% of a firm's staff, usually resulted in a loss of output. In only a few short years, it'll be more like a 25% layoff will not only add to the bottom line but also increased output for those who're left behind. Add a few more product generation cycles on this and soon, we'll have a very limited white collar workspace.

Heraclitusstudent says

Peter P says

What kind of AI are you building?

One based on programming structures, like other programs.

IMO, there are several types (stages) of Artificial Intelligence:

1. Programs (written in a programming language), you give the machine exact instructions to perform a task

2. Supervised Machine Learning, you teach a machine what things are

3. Unsupervised Machine Learning, the machine teaches itself with or without your guidance

Obviously, (3) is the most interesting because machines can move beyond its programming. This is where true emergence can occur.

Which one are you talking about?

There are other interesting views on intelligence. Some equate it with entropy maximization. Perhaps the universal will-to-power is all about maximizing future possibilities?

So exciting! :-)

This video is intriguing:

https://www.youtube.com/watch?v=PL0Xq0FFQZ4

Rin says

Add a few more product generation cycles on this and soon, we'll have a very limited white collar workspace.

Pretty much jobs will no longer exist. Soon enough, machines can do practically anything a human can do, only better and cheaper. Moreover, self-replicating robotic law enforcement can maintain peace in ANY environment, effectively and without moral confusion.

It is going to be interesting. :-)

Peter P says

Obviously, (3) is the most interesting because machines can move beyond its programming. This is where true emergence can occur.

Which one are you talking about?

(2) and (3) are always included in (1). They are programs like any other.
(2) and (3) are both necessary for intelligence. They just play different roles.

Heraclitusstudent says

Peter P says

Obviously, (3) is the most interesting because machines can move beyond its programming. This is where true emergence can occur.

Which one are you talking about?

(2) and (3) are always included in (1). They are programs like any other.

(2) and (3) are both necessary for intelligence. They just play different roles.

Yes, they build on one another. (2) and (3) are programs but not in the same sense as (1). You as the programmer further removed from the problem (as its solver) in (2) and (3) then in (1).