It’s absurd that this is even a question for some people. We may get AGI someday, but a vector database that generates streams of tokens based on tokens that it receives as input is never going to be it no matter how huge you make it. It lacks a world model, it doesn’t have the capacity to update itself with new information, and - most importantly - it doesn’t even do anything until it receives input.
And the assholes pushing this as potential AGI fucking know all that already. They know that without input, LLMs literally do nothing. They’re not sitting there going, “I think therefore I am” while waiting around to hallucinate dangerous medical advice and fake case law at people. They are static and unchanging.
If you’re referring to the Chinese Room, I don’t personally find it to be a compelling argument against the hypothetical possibility of AGI. The proposed room has no way to store information, no way to incorporate that information into its system, and no way to initiate activity without external input.
Heck, now that I think about it, it’s actually a pretty good description of the way LLMs actually work. But not the way an autonomous thinking agent works.
My mind is still blown that llm is by definition intelligent because it certainly does not meet my definition of intelligence, but rather seems intelligent. 🤷
And I agree. AGI will not evolve from this, but still actions are taken as if it were so, bringing us in a very bad position.
The ability to acquire, understand, and use knowledge.
"a person of extraordinary intelligence."
Information, especially secret information gathered about an actual or potential enemy or adversary. The gathering of such information.
They acquire and somewhat understand it but it doesn’t get saved. It only lives in the context window.
The things the big ones can do now are amazing when you have tasks that can be iterated upon until completion. And then there’s Deepseek V4 flash, a much smaller model with a still huge context window that costs an order of magnitude (nearly 2) less than the American frontier models and managed to do some spectacular things for me with essentially no input from me beyond the original prompt. It took hours, but for me to learn the tools and everything would’ve taken days or weeks.
Still not intelligent the way humans are. Next session starts, it’s all wiped clean until it reads the last session’s notes. But in terms of understanding information and reasoning about it, it’s going to be better than a human that isn’t a domain expert.
But going devils advocate you could say the training is doing both, in a sense. And if it’s allowed to use the web it can also acquire and utilize (hesitant to say understand because it’s not the same) new information.
I personally would also expect to save and utilize all information permanently and to have own motivations but that is more sentience and dynamic memory.
I like the Bobbyverse explanation of AI, these are AMIs, Artificial Machine Intelligences, capable enough within the confines of their task but absolutely not self-aware, and not suitable for any high risk situations. The tools with some limited thinking capabilities, that’s it.
When the characters want to build an actual AI they need outside help in order to create it. As humans just aren’t their technologically yet.
That’s where we are right now we have the AMIs or at least a starting to get them, but We’re nowhere close to a true AGI.
It’s absurd that this is even a question for some people. We may get AGI someday, but a vector database that generates streams of tokens based on tokens that it receives as input is never going to be it no matter how huge you make it. It lacks a world model, it doesn’t have the capacity to update itself with new information, and - most importantly - it doesn’t even do anything until it receives input.
And the assholes pushing this as potential AGI fucking know all that already. They know that without input, LLMs literally do nothing. They’re not sitting there going, “I think therefore I am” while waiting around to hallucinate dangerous medical advice and fake case law at people. They are static and unchanging.
It’s a shame nobody has thought about these sort of philosophical arguments before…
If you’re referring to the Chinese Room, I don’t personally find it to be a compelling argument against the hypothetical possibility of AGI. The proposed room has no way to store information, no way to incorporate that information into its system, and no way to initiate activity without external input.
Heck, now that I think about it, it’s actually a pretty good description of the way LLMs actually work. But not the way an autonomous thinking agent works.
My mind is still blown that llm is by definition intelligent because it certainly does not meet my definition of intelligence, but rather seems intelligent. 🤷
And I agree. AGI will not evolve from this, but still actions are taken as if it were so, bringing us in a very bad position.
If an LLM is intelligent, then an encyclopaedia is intelligent.
intelligence /ĭn-tĕl′ə-jəns/ noun
/s
My point being that an LLM will never be more “intelligent” than the lump of text that it is.
It’s not intelligent, like an encyclopaedia is not intelligent.
Oh I do agree with you, but it sadly is, by definition, artificial intelligence.
But LLMs are incapable of acquiring or understanding information.
They acquire and somewhat understand it but it doesn’t get saved. It only lives in the context window.
The things the big ones can do now are amazing when you have tasks that can be iterated upon until completion. And then there’s Deepseek V4 flash, a much smaller model with a still huge context window that costs an order of magnitude (nearly 2) less than the American frontier models and managed to do some spectacular things for me with essentially no input from me beyond the original prompt. It took hours, but for me to learn the tools and everything would’ve taken days or weeks.
Still not intelligent the way humans are. Next session starts, it’s all wiped clean until it reads the last session’s notes. But in terms of understanding information and reasoning about it, it’s going to be better than a human that isn’t a domain expert.
I agree, it also does not meet my expectations.
But going devils advocate you could say the training is doing both, in a sense. And if it’s allowed to use the web it can also acquire and utilize (hesitant to say understand because it’s not the same) new information.
I personally would also expect to save and utilize all information permanently and to have own motivations but that is more sentience and dynamic memory.
I trained my book by putting words in it!
I like the Bobbyverse explanation of AI, these are AMIs, Artificial Machine Intelligences, capable enough within the confines of their task but absolutely not self-aware, and not suitable for any high risk situations. The tools with some limited thinking capabilities, that’s it.
When the characters want to build an actual AI they need outside help in order to create it. As humans just aren’t their technologically yet.
That’s where we are right now we have the AMIs or at least a starting to get them, but We’re nowhere close to a true AGI.