That’s my point. OP doesn’t know the maths, has probably never implemented any sort of ML, and is smugly confident that people pointing out the flaws in a system generating one token at a time are just parroting some line.
These tools are excellent at manipulating text (factoring in the biases they have, I wouldn’t recommended trying to use one in a multinational corporation in internal communications for example, as they’ll clobber non euro derived culture) where the user controls both input and output.
Help me summarise my report, draft an abstract for my paper, remove jargon from my email, rewrite my email in the form of a numbered question list, analyse my tone here, write 5 similar versions of this action scene I drafted to help me refine it. All excellent.
Teach me something I don’t know (e.g. summarise article, answer question etc?) disaster!
No, they can summarise articles very convincingly! Big difference.
They have no model of what’s important, or truth. Most of the time they probably do ok but unless you go read the article you’ll never know if they left out something critical, hallucinated details, or inverted the truth or falsity of something.
That’s the problem, they’re not an intern they don’t have a human mind. They recognise patterns in articles and patterns in summaries, they non deterministically adjust the patterns in the article towards the patterns in summaries of articles. Do you see the problem? They produce stuff that looks very much like an article summary but do not summarise, there is no intent, no guarantee of truth, in fact no concern for truth at all except what incidentally falls out of the statistical probability wells.
I think it’s really important to keep in mind the separation between doing a task and producing something which looks like the output of a task when talking about these things. The reason being that their output is tremendously convincing regardless of its accuracy, and given that writing text is something we only see human minds do it’s so easy to ascribe intent behind the emission of the model that we have no reason to believe is there.
Amazingly it turns out that often merely producing something which looks like the output of a task apparently accidentally accomplishes the task on the way. I have no idea why merely predicting the next plausible word can mean that the model emits something similar to what I would write down if I tried to summarise an article! That’s fascinating! but because it isn’t actually setting out to do that there’s no guarantee it did that and if I don’t check the output will be indistinguishable to me because that’s what the models are built to do above all else.
So I think that’s why we to keep them in closed loops with person -> model -> person, and explaining why and intuiting if a particularly application is potentially dangerous or not is hard if we don’t maintain a clear separation between the different processes driving human vs llm text output.
You are so extremely outdated in your understanding,
For one that attacks others for not implementing their own llm
They are so far beyond the point you are discussing atm. Look at autogen and memgpt approaches, the way agent networks can solve and develop way beyond that point we were years ago.
It really does not matter if you implement your own llm
Then stay out of the loop for half a year
It turned out that it’s quite useless to debate the parrot catchphrase, because all intelligence is parroting
It’s just not useful to pretend they only “guess” what a summary of an article is
They don’t. It’s not how they work and you should know that if you made one
No they don’t, and your idiotic personal attacks won’t change how the tech works. You are just wrong. I don’t love them, I don’t care about Altman. I was just trying to tell you you are spreading misinformation. But nah defensive slurs it is
That’s my point. OP doesn’t know the maths, has probably never implemented any sort of ML, and is smugly confident that people pointing out the flaws in a system generating one token at a time are just parroting some line.
These tools are excellent at manipulating text (factoring in the biases they have, I wouldn’t recommended trying to use one in a multinational corporation in internal communications for example, as they’ll clobber non euro derived culture) where the user controls both input and output.
Help me summarise my report, draft an abstract for my paper, remove jargon from my email, rewrite my email in the form of a numbered question list, analyse my tone here, write 5 similar versions of this action scene I drafted to help me refine it. All excellent.
Teach me something I don’t know (e.g. summarise article, answer question etc?) disaster!
They can summarize articles fairly well
No, they can summarise articles very convincingly! Big difference.
They have no model of what’s important, or truth. Most of the time they probably do ok but unless you go read the article you’ll never know if they left out something critical, hallucinated details, or inverted the truth or falsity of something.
That’s the problem, they’re not an intern they don’t have a human mind. They recognise patterns in articles and patterns in summaries, they non deterministically adjust the patterns in the article towards the patterns in summaries of articles. Do you see the problem? They produce stuff that looks very much like an article summary but do not summarise, there is no intent, no guarantee of truth, in fact no concern for truth at all except what incidentally falls out of the statistical probability wells.
That’s a good way of explaining it. I suppose you’re using a stricter definition of summary than I was.
I think it’s really important to keep in mind the separation between doing a task and producing something which looks like the output of a task when talking about these things. The reason being that their output is tremendously convincing regardless of its accuracy, and given that writing text is something we only see human minds do it’s so easy to ascribe intent behind the emission of the model that we have no reason to believe is there.
Amazingly it turns out that often merely producing something which looks like the output of a task apparently accidentally accomplishes the task on the way. I have no idea why merely predicting the next plausible word can mean that the model emits something similar to what I would write down if I tried to summarise an article! That’s fascinating! but because it isn’t actually setting out to do that there’s no guarantee it did that and if I don’t check the output will be indistinguishable to me because that’s what the models are built to do above all else.
So I think that’s why we to keep them in closed loops with person -> model -> person, and explaining why and intuiting if a particularly application is potentially dangerous or not is hard if we don’t maintain a clear separation between the different processes driving human vs llm text output.
Yeah this is basically how I think of them too. It’s fascinating that it works.
You are so extremely outdated in your understanding, For one that attacks others for not implementing their own llm
They are so far beyond the point you are discussing atm. Look at autogen and memgpt approaches, the way agent networks can solve and develop way beyond that point we were years ago.
It really does not matter if you implement your own llm
Then stay out of the loop for half a year
It turned out that it’s quite useless to debate the parrot catchphrase, because all intelligence is parroting
It’s just not useful to pretend they only “guess” what a summary of an article is
They don’t. It’s not how they work and you should know that if you made one
They uh, still do the same thing fundamentally
Altman isn’t gonna let you blow him dude
No they don’t, and your idiotic personal attacks won’t change how the tech works. You are just wrong. I don’t love them, I don’t care about Altman. I was just trying to tell you you are spreading misinformation. But nah defensive slurs it is