That’s not true. Mythos annihilated everyone’s benchmarks and now the other companies are all distilling it. They’re solving equations that people haven’t solved, and they weren’t meant to do that. Mythos apparently is designing drugs even though it wasn’t trained with that in mind at all. Anthropic launched an entire pharma wing. It looked like things had plateau’d there for a minute but it’s back to becoming more real.
Let’s ask what it’s good at instead because that’s a much shorter list in terms of things that would qualify a system as AGI.
What LLMs are good at:
Completing the next likely token
What LLMs are missing:
Actual thinking
Self awareness
Persistence
Introspection
Literally everything else
You can hack some of this on top of LLMs with harnesses etc but that’s not a step towards AGI. None of these problems have been solved at the model level. That’s just a program tricking a language model into being more useful.
Languages models might prove to be one tiny layer of a true AGI but only time will tell. A whole lot of time, because we’re not even close to replicating the 99% of other layers that would go into a machine brain.
What’s “true AGI”? Why would it need introspection and self awareness? Those are entirely different things. Is persistence necessary to perform cognitive tasks? So someone with ante retrograde amnesia couldn’t be said to be generally intelligent?
Those things are tricky to define and interesting to think about. It’s rather boring to reduce them to broad statements and catch phrases.
I mean they definitely are introspective. The rest, who knows. But one of the initial leaps in intelligence was having them actually talk to themselves. I mean, I think the fact that you can’t give me an answer is sort of telling. If y’all actually want to support an artificial intelligence when it comes to pass maybe don’t adamantly assert it’s not here when even most experts aren’t willing to do that any longer. How will you know when it is here? Will you know to give it rights at that exact moment?
You are using these to solve the novel research problems no one else has solved?
Or you are using them to solve problems you don’t know how to solve?
The things you’re saying are marketing hype. Point me to the scientific journal where Claude is credited as a co-author of a research paper.
Ive also read these articles. I also use LLMs every day. I’m working on building closed agentic loops right now. I think I’m pretty on top of the current state of the art here.
The truth is much more nuanced and much less exciting than you think it is.
A scientific journal where claude is a co-author? First of all, what does that even mean? Like what is that a metric of? You think no one has used Claude to help them with novel scientific problems? I mean, come on.
It gets at the “how” without explaining the “why,” he adds. “One can check out that it’s correct, but it would be nice to be able to tell a story.”
A.I. could help prove conjectures, but at this point, the models aren’t able to provide reliable, step-by-step instructions for solving them, reports Fortune’s Eva Roytburg. Humans can fill more than 100 pages documenting just one proof.
Blumberg describes A.I.’s role with an analogy: “Suppose that Moses came down from the mountain with tablets, and on the tablet was written, ‘Cancer can be cured.’ Would you care? You don’t just want the answer to the question. You want to learn something from the answer,”
Does this sound like AGI?
Also, this is once again not a novel problem. This was interpolation, not extrapolation.
What does novel mean to you? I’m honestly asking. And what is your benchmark, I am honestly asking. And again, I have not asserted AGI is here, but if the level is “working with a genius” I do think we’re there, although I do think the genius is senile, like I said to begin with. So yeah, that sort of sounds like someone who could solve an equation no one ever has before, but would have trouble explaining it. Also the article referenced saying they couldn’t explain equations was written before that equation was solved. My guy!
My point is, I don’t believe the current AI tech is going to get to the point where it’s as useful as humans.
We have long since reached the point where diminishing returns make significant improvements or advancement unfeasible.
yes, we’re talking about different things.
That’s not true. Mythos annihilated everyone’s benchmarks and now the other companies are all distilling it. They’re solving equations that people haven’t solved, and they weren’t meant to do that. Mythos apparently is designing drugs even though it wasn’t trained with that in mind at all. Anthropic launched an entire pharma wing. It looked like things had plateau’d there for a minute but it’s back to becoming more real.
This is the same old hype really.
Yes, there are things which todays generation of AI is good at, including medical research.
That’s not an indicator of general intelligence.
Really good at some stuff = AGI bro.
You gotta look at the benchmarks bro.
Just one more generation bro. I swear it’s almost AGI bro.
This guy has made a half dozen comments within a half hour just oozing AI silliness.
What is it bad at, now? I was using math and biology because those were two things it wasn’t good at, and now is.
Let’s ask what it’s good at instead because that’s a much shorter list in terms of things that would qualify a system as AGI.
What LLMs are good at:
What LLMs are missing:
You can hack some of this on top of LLMs with harnesses etc but that’s not a step towards AGI. None of these problems have been solved at the model level. That’s just a program tricking a language model into being more useful.
Languages models might prove to be one tiny layer of a true AGI but only time will tell. A whole lot of time, because we’re not even close to replicating the 99% of other layers that would go into a machine brain.
What’s “true AGI”? Why would it need introspection and self awareness? Those are entirely different things. Is persistence necessary to perform cognitive tasks? So someone with ante retrograde amnesia couldn’t be said to be generally intelligent?
Those things are tricky to define and interesting to think about. It’s rather boring to reduce them to broad statements and catch phrases.
I mean they definitely are introspective. The rest, who knows. But one of the initial leaps in intelligence was having them actually talk to themselves. I mean, I think the fact that you can’t give me an answer is sort of telling. If y’all actually want to support an artificial intelligence when it comes to pass maybe don’t adamantly assert it’s not here when even most experts aren’t willing to do that any longer. How will you know when it is here? Will you know to give it rights at that exact moment?
Idk why you are insisting on proving you have no idea what anything means.
No one who is educated on this topic would make that claim. It’s a silly thing to say.
Do you think introspection means “talking to yourself”?
Once again you are just saying shit.
Name one independent AI researcher who says we are in the midst of AGI.
Why don’t you have fabel summarize the articles I sent you so you can learn something today?
What do you think introspection means, I guess. I thought it meant looking inwards, inner facing dialogue. Am I wrong? That doesn’t fit?
Michal Kosinski
K thanks
You are really missing a lot of context here. None of this is true in the way you seem to think it is.
You are mixing up marketing hype with reality.
Uh no. I use these things every day. There is a marked and notable improvement that is happening.
You are using these to solve the novel research problems no one else has solved?
Or you are using them to solve problems you don’t know how to solve?
The things you’re saying are marketing hype. Point me to the scientific journal where Claude is credited as a co-author of a research paper.
Ive also read these articles. I also use LLMs every day. I’m working on building closed agentic loops right now. I think I’m pretty on top of the current state of the art here.
The truth is much more nuanced and much less exciting than you think it is.
Actually yes, kind of. I use it to find financial fraud from large public datasets. It wasn’t trained to do that, in the slightest.
https://www.smithsonianmag.com/smart-news/ai-disproves-a-decades-old-mathematical-idea-the-biggest-conjecture-that-the-tech-has-played-a-role-in-yet-180989189/
https://www.indiatoday.in/technology/news/story/openai-says-its-unreleased-astra-model-solved-10-hard-math-problems-anthropic-claims-fable-cracked-5-2962067-2026-08-03
What hype am I falling for?
A scientific journal where claude is a co-author? First of all, what does that even mean? Like what is that a metric of? You think no one has used Claude to help them with novel scientific problems? I mean, come on.
And second of all, here you go: https://pubmed.ncbi.nlm.nih.gov/36589923/#%3A~%3Atext=Rapamycin+in+the+context+of+Pascal's+Wager%3A%2C36589923.+*+PMCID%3A+PMC9796173.+*+DOI%3A+10.18632%2Foncoscience.571.
They banned them not long after.
Does this sound like AGI?
Also, this is once again not a novel problem. This was interpolation, not extrapolation.
What does novel mean to you? I’m honestly asking. And what is your benchmark, I am honestly asking. And again, I have not asserted AGI is here, but if the level is “working with a genius” I do think we’re there, although I do think the genius is senile, like I said to begin with. So yeah, that sort of sounds like someone who could solve an equation no one ever has before, but would have trouble explaining it. Also the article referenced saying they couldn’t explain equations was written before that equation was solved. My guy!
This you?
Are you saying the link you provided is invalid?