AI Progress // Ex-OpenAi Engineer Document

[CARDOZZO]

10 minutes ago, Rafael Thundercat said:

We can send Marçal as well hehe. Sorry, Side Joke. 

Lol

Are you 🇧🇷 ?

[Rafael Thundercat]

19 minutes ago, CARDOZZO said:

Lol

Are you 🇧🇷 ?

Yes, not living there but following the whole Circus online. There is any hope for the Country in General? I say in general because some hidden paradises there seems like another dimension inside de same country. 

[CARDOZZO]

2 minutes ago, Rafael Thundercat said:

Yes, not living there but following the whole Circus online. There is any hope for the Country in General? I say in general because some hidden paradises there seems like another dimension inside de same country. 

In terms of politics, we don't have hope anywhere on earth.

I live in the south. (Rio Grande do Sul)

We are the farm of the planet. We don't have wars. 

We have a lot to advance as a whole.

 

 

[OBEler]

On 24.6.2024 at 1:45 PM, nuwu said:

Is strong AGI worth researching? If it provides radical quality of life and longevity improvements, and if internal models of institutions (ie OpenAI) are weak with respect to philosophy of self-compoundable computational symbolism or something

Deterministic computational substrate may be inherently limited by inward constraint systems, such as in silica pseudo-general intelligence is relatively safe by induction, assuming humans inputs are reasonable. However if effective structures are discovered on traditional computing, they might be transposed in quantum computing with potential inherent uncertainty and exposition to universe intermediate chaos, or implemented in bio-engineered hardware which may effective outcompete human perspectives.

 I really don't know what you want to try to say. It's really some rambling, throwing buzzwords around.hard to read.

Did you read the paper at least?

[OBEler]

@nuwu No Problem, with the help of Chatgpt4 I can understand your thoughts.

I think government will not stop in developing further at least China will do it. AGI research itself and improving itself seems reasonable.

 

[Ero]

On 27/6/2024 at 5:05 AM, nuwu said:

I am simply pondering whether AGI is worth researching in itself, propositional to generality of metaphysics paralleling autodual fractal from computational determinism or equivalent, and safety of eventual translation into non-deterministic substrate abstraction.

"AGI" is not something being researched itself, since there is neither an agreed definition of what it actually means, nor are we actually close to it. Research in AI is currently conducted under the Deep Learning paradigm, so most of the work is done empirically by scaling existent models and seeing what comes out of it. 

Seconding what OBEler said, "metaphysics paralleling autodual fractal" is a word soup. "Autodual" is spanish for "self-dual", a mathematical term expressing an isomorphism between a mathematical object and its dual, something that does not apply to fractals.

 

[Ero]

@nuwu I believe "AGI" in the way you are referring to it is worth contemplating/ researching, due to its expected revolutionary impact. From an expected (E[X] = P(X) x C) opportunity cost point of view, even at a 20 year timeline, your effort will most likely be worth it.

What I was pointing to you is that without a consensus on what the various words and concepts mean, such as "AGI" itself, or without you clarifying what exactly the "implicit undisclosed contest" for your use of "autodual" is, it is very hard to communicate on such topics.

AI is fundamentally a scientific research discipline, currently residing at the intersection of Mathematics, Statistics and Computer Science. If you have studied any of them in a formal manner, you may realize that all three of the disciplines inherently work with “strong formal mathematical objects”. Every AI model created in existence is a “mathematical object”. When you read a paper, say "Attention is all you need" or "Categorical Deep Learning is an Algebraic Theory of All Architectures", all the terms used in them, such as "computation", "graph", "automata" are terms with a precise meaning, such that whoever reads them and has the necessary background can without a problem understand the gist of it. 

If you personally begin researching, which if you are serious, would entail reading papers, you will encounter many such concepts. And if you want to understand and possibly contribute, your current mode of communication and approach to definitions as "subjective" would be rather unfruitful.

P.S. Outside of the nerdy context, throwing around words like "mafia" without you actually having experienced what it means to lose a family member to organized crime is an example of a difference in the subjective meaning we have for that word. Just saying. 

[Ero]

3 minutes ago, nuwu said:

nitpicking subjective semantics with no regard to intellectual depth

I am sorry if that is how you felt. What I am trying to say is that it is hard to grasp your intellectual depth when this is the way you communicate.

You pick and choose what different words mean, rather than explaining them to us, the ants in the "idiot realm", or maybe your depth is beyond my "lame" human state.

Now, the sentence above was not meant to ironize, but rather show you that if we are to play the "gaslighting game", your use of such pejoratives far exceeds the "condescending tone" you presumed I had. 

7 minutes ago, nuwu said:

I am sorry, present epistemology, commonly denotationally referred by "science" is nothing but precise, and barely scratching the surface of metaphysical coherence.

I agree with your observation about science in its present form as epistemically limited due to its fundamentally materialistic ontology and metaphysics. But that doesn't change the fact that it is this "science" that will bring AI. And epistemically and metaphysically-revolutionary ideas will still have to be communicated through language.

So, I would be interested to hear your thoughts on where these innovations would come from? From what I understood of your earlier posts, you think we have to transition away from "computational determinism", which correct me if I am wrong, but I understand as the philosophical position of finitism, i.e only computable objects exist. Your suggestions was for a non-deterministic substrate, which is something I agree with.