Mainly because LLMs have so far basically passed every formal test of ‘AGI’ including totally smashing the Turing test.
Now we are just reliant on ‘I’ll know it when I see it’.
LLMs as AGI isn’t about looking at the mechanics and trying to see if we think that could cause AGI - it’s looking at the tremendous results and success.
“You're in a desert, walking along in the sand when all of a sudden you look down and see a tortoise. You reach down and flip the tortoise over on its back. The tortoise lays on its back, its belly baking in the hot sun, beating its legs trying to turn itself over. But it can't. Not with out your help. But you're not helping. Why is that?”
got-3.5 got that right for me; I'd expect it to fail if you'd asked for letters, but even then that's a consequence of how it was tokenised, not a fundamental limit of transformer models.
This sort of test has been my go-to trip up for LLMs, and 3.5 fails quite often. 4 has been as bad as 3.5 in the past but recently has been doing better.
if this is the test you're going to then you literally do not understand how LLMs work. it's like asking your keyboard to tell you what colour the nth pixel on the top row of your computer monitor is.
An LLM could easily answer that question if it was trained to do it. Nothing in its architecture makes it hard to answer, the attention part could easily look up the previous parts of its answer and refer to the fourth word but it doesn't do that.
So it is a good example that the LLM doesn't generalize understanding, it can answer the question in theory but not in practice since it isn't smart enough. A human can easily answer it even though the human never saw such a question before.
> the model doesn't have a functionality to retrospectively analyse its own output; it doesn't track or count words as it generates text. it's always in the mode of 'what comes next?' rather than 'what have i written?'
Humans doesn't do that either. The reason humans can solve this problem is that humans can generate such strategies on the fly and thus solve general problems, that is the bar for AGI, as long as you say it is unfair to give such problems to the model we know that we aren't talking about an AGI.
Making a new AI that is specialized in solving this specific problem by changing the input representation still isn't an AGI, it will have many similar tasks that it will fail at.
> also, again, tired of explaining this to people: gpt models are token-based. they operate at the level of tokens - which can be whole words or parts of words - and not individual characters. this token-based approach means the model's primary concern is predicting the most probable next token, not keeping track of the position of each token in the sequence, and the smallest resolution available to it is not a character. this is why it can't tell you what the nth letter of a word is either.
And humans are a pixel based model, we operate on pixels and physical outputs. Yet we humans do generate all the necessary context, and adapts it to the task at hand to solve arbitrary problem. Such context and inputs manipulations are expected of an AGI. Maybe not the entire way from pixels and 3d mechanical movement, but there are many steps in between there that humans can easily adapt in. For example humans didn't evolve to read and write text, yet we do that easily even though we operate on a pixel level.
If you ask me to count letters my mind focuses on the letter representation I created in my head. If you talk about words I focus on the word representation. If you talk about holes I focus on the pixel representation and start to identify color parts. If you talk about sounds I focus on the vocal representation of the words since I can transform to that as well.
We would expect an AGI to make similar translations when needed, from the token space you talk about to the letter space or word space etc. That ChatGPT and similar can't do this just means they aren't even close to AGI currently.
I reckon an LLM with a second pass correction loop would manage it. (By that I mean that after every response it is instructed to, given the its previous response, produce a second better response, roughly analogous to a human that thinks before it speaks)
LLMs are not AIs, but they could be a core component for one.
Every token is already being generated with all previously generated tokens as inputs. There's nothing about the architecture that makes this hard. It just hasn't been trained on this kind of task.
The following are a part of my "custom instructions" to chatGPT -
"Please include a timestamp with current date and time at the end of each response.
After generating each answer, check it for internal consistency and accuracy. Revise your answer if it is inconsistent or inaccurate, and do this repeatedly till you have an accurate and consistent answer."
It manages to follow them very inconsistently, but it has gone into something approaching an infinite loop (for infinity ~= 10) on a few occasions - rechecking the last timestamp against current time, finding a mismatch, generating a new timestamp, and so on until (I think) it finally exits the loop by failing to follow instructions.
I don’t think the test of AGI should necessarily be an inability to trip it up with specifically crafted sentences, because we can definitely trip humans up with specifically crafted sentences.
I disagree about the claim that any LLM has beaten the Turing test. Do you have a source for this? Has there been an actual Turing test according to the standard interpretation of Turings paper? Making ChatGPT 4 respond in a non human way right now is trivial: "Write 'A', then wait one minute and then write 'B'".
Your test fails because the scaffolding around the LM in ChatGPT specifically does not implement this kind of thing. But you absolutely can run the LM in a continuous loop and e.g. feed it strings like "1 minute passed" or even just the current time in an internal monologue (that the user doesn't see). And then it would be able to do exactly what you describe. Or you could use all those API integrations that it has to let it schedule a timer to activate itself.
By completely smashes, my assertion would be that it has invalidated the Turing test, because GPT-4s answers are not indistinguishable from a human because they are, on the whole, noticeably better answers than an average human would be able to provide for the majority of questions.
I don’t think the original test probably accounted for the fact that you could distinguish the machine because it’s answers were better than an average human.
LLMs can't develop concepts in the way we think of them (i.e., you can't feed LLMs the scientific corpus and ask them to independently to tell you which papers are good or bad and for what reasons, and to build on these papers to develop novel ideas). True AGI—like any decent grad student—could do this.
that being said, it is highly intelligent, capable of reasoning as well as a human, and passes IQ tests like GMAT and GRE at levels like the 97th percentile.
most people who talk about Chat GPT don't even realize that GPT 4 exists and is orders of magnitude more intelligent than the free version.
That’s just showing the tests are measuring specific things that LLMs can game particularly well.
Computers have been able to smash high school algebra tests since the 1970’s, but that doesn’t make them as smart as a 16 year old (or even a three year old).
Answers in Progress had a great video[0] where one of their presenters tested against an LLM in five different types of intelligence. tl;dr, AI was worlds ahead on two of the five, and worlds behind on the other three. Interesting stuff -- and clear that we're not as close to AGI as some of us might have thought earlier this year, but probably closer than a lot of the naysayers think.
This isn't the same thing. This is a commanded recital of a lack of capability, not that its confidence in it's answer is low. For a type of question the GPT _could_ answer, most of the time it _will_ answer, regardless of accuracy
I just noticed that when I ask really difficult technical questions, but for which there is an exact answer, It often tries to answer plausibly, but incorrectly instead of answering "I don't know". But over time, It becomes smarter and there are fewer and fewer such questions...
I read an article where they did a proper Turing test and it seems people recognize it was a machine answering because it made no writing errors and wrote perfectly
I've not read that, but I do remember hearing that the first human to fail the Turing test did so because they seemed to know far too much minutiae about Star Trek.
Now we are just reliant on ‘I’ll know it when I see it’.
LLMs as AGI isn’t about looking at the mechanics and trying to see if we think that could cause AGI - it’s looking at the tremendous results and success.