4 ms·
This doesn't seem like a major difference, since LLMs are also choosing from a probability distribution of tokens for the most likely one, which is why they res
by bytefactory 3y ago
This doesn't seem like a major difference, since LLMs are also choosing from a probability distribution of tokens for the most likely one, which is why they respond a token at a time. They can't "write out' the entire text at a time, which is why fascinating methods like "think step by step" work at all.
- Jensson 3y agoBut it can't improve its answer after it has written it, that is a major limitation. When a human writes an article or response or solution, that is likely not the first thing the human thought of, instead they write something down and works on it until it is tight and neat and communicates just what the human wants to communicate. Such answers will be very hard for an LLM to find, instead you mostly get very verbose messages since that is how our current LLM thinks.
- bytefactory 3y agoCompletely agree. The System 1/System 2 distinction seems relevant here. As powerful as transformers are with just next-token generation and context, which can be hacked to form a sort of short-term memory, some time of real-time learning + long-term memory storage seems like an important research direction.
- xcv123 3y ago> But it can't improve its answer after it has written it, that is a major limitation. It can be instructed to study its previous answer and find ways to improve it, or to make it more concise, etc, and that is working today. That can easily be automated by LLMs talking to each other.
- ewild 3y agothat is true and isnt. GPT4 has shown itself to halfway through a answer say "wait thats not correct im sorry let me fix that" and then correct itself. For example it stated a number was prime and why, and when showing the steps found it was divisible by 3 and said "oh i made a mistake it actually isnt prime"