News for hackers. Stuff that matters.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by PartiallyTyped on Wednesday March 15, 2023 @04:03PM (#35173959)
Why is it not an AGI?

I keep seeing naysayers complain that it is not, but nobody actually articulates exactly what makes it a non-AGI in a manner that is sufficiently robust to argumentation.

I am willing to play the GPT's advocate and have this argument here provided that I do not get rate limited.

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by wizofaus on Wednesday March 15, 2023 @04:28PM (#35174282)
I've posted a number of times why I don't think current GPT tech classifies as AGI and certainly not human-like intelligence, as have others. Ask it to solve novel mathematical and logic problems, for a start (even ones you'd expect high school kids to manage), or simple word puzzles (anagrams and the like - again, avoiding "known" examples that it's likely to have come across in its training data). Further, an AGI should logically attempt to "improve" its understanding of the world during a conversation, yet chat GPT never initiates a topic or asks you anything at all (even to clarify what you mean). It also repeatedly reminds you that it is only a language model.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by antibasilisk on Wednesday March 15, 2023 @04:33PM (#35174334)
Machines can never obtain human like intelligence because they are foundationally not human like, that doesn't mean they lack intelligence. That's like saying dolphins lack intelligence.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by wizofaus on Wednesday March 15, 2023 @04:53PM (#35174555)
I wouldn't claim ChatGPT lacks intelligence though. Clearly in terms of language comprehension and generation it does so.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by z3c0 on Wednesday March 15, 2023 @04:56PM (#35174594)
False equivalence. Dolphins have exhibited general problem-solving skills in a myriad of situations, and thus qualify as a general intelligence.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by wizofaus on Wednesday March 15, 2023 @05:16PM (#35174830)
Indeed, I had started to write something similar. And pretty clearly animals like Dolphins or Chimps when they exercise their general intelligence don't do it by manipulation of language tokens.

BTW I tried to get Chat-GPT to solve a problem chimpanzees have been observed being able to solve (using an inverted bowl to stand on in order to reach some bananas):

"I'm sorry, but it is not biologically plausible for a chimpanzee to use a bowl of water to access bananas on a tree"

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by dave84 on Wednesday March 15, 2023 @07:30PM (#35176125)
GPT-4 had no issues answering this.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by wizofaus on Thursday March 16, 2023 @02:18PM (#35186013)
Presumably because it's similar enough to descriptions of problems and how they were solved that are in its training data.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by antibasilisk on Wednesday March 15, 2023 @09:34PM (#35177068)
As has PaLM-E
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by pixl97 on Wednesday March 15, 2023 @05:37PM (#35175066)
I believe you're correct, but for possibly the wrong reasons. I believe that AI will be able to simulate a lot (if not most) humanlike action in the near future. The particular problem here is they will also simulate far more non-humanlike intelligence making the alignment issue an even larger problem than it already is.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by ekleraki on Wednesday March 15, 2023 @05:18PM (#35174866)
> Ask it to solve novel mathematical and logic problems

How exactly do you define novel tasks and problems?

My brother has had it do his programming homework (he's a teenager), written in Greek mind you and therefore novel task to the first-degree, and it succeeded. Thus it solved a logic task [1].

I have had ChatGPT parse math I had written in LaTeX and reach a correct result, do modifications and so on.

> Further, an AGI should logically attempt to "improve" its understanding of the world during a conversation, yet chat GPT never initiates a topic or asks you anything at all (even to clarify what you mean). It also repeatedly reminds you that it is only a language model.

Anecdata: I asked Bing (GPT4) to answer a few stuff and it asked for clarification twice in a row. The fact that it repeatedly clarifies its prompt is not a limitation of the underlying model as much as it is a limitation of the interface that injects a prompt. Dan and all the other jailbreaks out there bypass that.

[1] https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspon...

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by UncleEntity on Wednesday March 15, 2023 @06:39PM (#35175696)
> I have had ChatGPT parse math I had written in LaTeX and reach a correct result, do modifications and so on.

Interesting, didn’t know it could do that.

Wondering if it could also follow along with the lemmas in various papers and reach reasonable conclusions.

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by PartiallyTyped on Thursday March 16, 2023 @05:28AM (#35179829)
It depends, I have used Copilot to modify LaTeX and write segments of my thesis, and it works surprisingly well, though its arguments are not very precise.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by wizofaus on Thursday March 16, 2023 @06:35PM (#35189839)
https://medium.com/illumination/i-put-chatgpt-to-the-test-he... has some examples of simple maths/logic problems it can't really solve. It'd be good to know how much better chatgpt-4 is at them.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by Karunamon on Wednesday March 15, 2023 @04:41PM (#35174419)
* * *
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by PartiallyTyped on Wednesday March 15, 2023 @05:05PM (#35174712)
> One major reason it doesn't qualify is that it is impossible for it to learn from its input. It's right there in the name: generative pre-trained transformer

Both chatGPT and GPT models exhibit in-context learning, just as you do. The difference is that during your context updates modifications to your synapses occur as a consequence of your brain's activations.

Let's entertain a hypothetical scenario where I am a 5D creature, and you are (3 Space + 1 Time)D human.

I could go back to any point in your time and have any conversation that I wish with you. Each time I do that, I am effectively resetting your brain's state to that particular point in time.

The analog here is that for ChatGPT and the likes, time does not exist in the manner that it does for us, they are purely abstract in that they don't even need physical form to exist, it's a collection of ones and zeros that we store in such form because it is convenient.

You can very well store the parameters of the GPT models in any other medium capable of holding information and actually execute the model.

You can even compute gradient updates, or hook the models up to external data sources.

In the same vein, you "could" do the same for any human brain no matter how computationally prohibitive it may be.

You'd still classify the human brain as a general intelligence. Furthermore, we classify humans with long-term amnesia or inability to form new memories generally intelligent even though they only exhibit in-context learning.

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by Karunamon on Wednesday March 15, 2023 @05:29PM (#35174985)
Interesting analogy… Actually a damn good one because I'm having a hard time finding how it leaks :)

I have to ask what definition(s) of intelligence we are using here. There are multiple, and GPT satisfies some to some degree but not others. Emotion and self-awareness can probably be written off entirely. Some experimenting shows that it still sucks at logic when presented with a pristine problem (something that definitely is not in its training set); and I would say this is a requirement of a general intelligence, being able to take the knowledge it has and speculatively applying that to things it has no knowledge of using induction/deduction.

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by PartiallyTyped on Thursday March 16, 2023 @05:48AM (#35179947)
As I have said in a parallel chain, ChatGPT has been able to solve programming assignments for 15-17 year olds written in Greek. The P it has seen those tasks in Greek before is approaching zero imho.

> being able to take the knowledge it has and speculatively applying that to things it has no knowledge of using induction/deduction.

One could argue that its 'confidence' is an example of speculatively applying what it knows. The problem is that the inference machinery behind it is quite weak.

Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F... by catoc on Wednesday March 15, 2023 @04:11PM (#35174077)
Not trying to be a naysayer or downplay how cool GPT4 is, but the term 'AGI' has existed for decades and is suffering from general confusion and inflation (again like 'crypto' now being read as 'coins' or even 'NFT' by the general public).

See eg: https://en.wikipedia.org/wiki/Artificial_general_intelligenc... or better read Bostrom's book "Super Intelligence"