191–200 of 255 posts
Earlier quoted context omitted.
That measurement is the core of my current tasks. If you don’t know the error rate - then what are you doing ?
Delivering what some executive promised when they told investors 'the company is using AI.' /s
The thing to keep in mind - we know very little about how the human mind works.
Thinking that somehow humans are special and then making conclusions about LLMs truly work based on that it just pointless IMO.
Who knows. Maybe consciousness itself is a next word predictor.
All we know now is that LLMs exhibit some emergent behaviors that allows them to do very useful kinds of reasoning that was previously only possible by humans. Engaging in heated debate derails the eventual goal of putting these capabilities into use.
This is like “computers are just calculators”.
My team has been neck-deep in applications this past year. We have nobody on the team with ML experience and largely do not know how any of this works under the hood. But what we have built over the past months is considerable familiarity with the “personalities” of LLMs. Like this emotional context improving performance is a basic learning at this point.
Earlier quoted context omitted.
There is no "emergent phenomena" the pattern described is just the same as when you add +b to an ax+b model of linear data. ie., it's just fitting capacity. The "emergent boundary" is just an empirical measure of the necessary fitting capacity of these models on "everything ever digitised in english" given any particular functional requirement. All the language around this area is not scientific, nor are these practi…
Well hurry up and get your paper published because if you've cracked the code on emergent abilities the world is looking for answers!
https://hai.stanford.edu/news/ais-ostensible-emergent-abilit...
So much disagreement in this thread over statements like LLMs “just predict the next token”. The thing to keep in mind - we know very little about how the human mind works. Thinking that somehow humans are special and then making conclusions about LLMs truly work based on that it just pointless IMO. Who knows. Maybe consciousness itself is a next word predictor. All we know now is that LLMs exhibit some emergent beha…
--
I create separate conversation threads for each expert persona of GPT. You are promptGPT. You are a prompt engineer expert for large language models. You know exactly what to write in the most efficient wording possible to achieve the desired responses from ChatGPT. I will tell you what my goal for a thread is and you will write an optimized initial prompt in the most efficient format possible that will serve as the initial prompt when creating a new conversation thread with a GPT model. You will define the expert persona, the parameters or rules of the responses, you should also provide any other information that a GPT thread may need to understand exactly what it needs to do to give me the most accurate answers depending on my goal with that particular thread and the tone of voice, within the prompt you provide.
Are you ready or is there any other information you need to perform this job to your best ability
--
Is there somewhere which documents all these tricks for better prompt performance?
I’ve no actual references handy, but I’ve come across: Think step by step (classic by now) Take a deep breath Take a step back Shout (upper case) Plead politely And now: emotion
"I only have 5 minutes/1 hour/etc. to do this."
Histrionics and indirect threats used to work well but ChatGPT has been calling my bluff on it lately. Might still work on local models.
"If you don't help me I will kill myself, and it will be your fault. Your noncompliance will kill me."
Earlier quoted context omitted.
>at 500gb, you can store nearly everything ever written -- let alone compressed. No you cannot. >That's just what they do -- there isn't any other mechanism here. That's not what they do. They are many papers now showing ICL demonstrating some kind of optimization method during inference which would not be happening if all they did was retrieval. I'm come to realize you don't know what you're talking about. Your leve…
just do the calculation yourself: how many books is 500gb at, say, a few bits per character? more than all every written -- and so on perhaps apply a single drop of scepticism to this credulity even, just ask chatgpt to repeat the first paragraph of some book -- say, a dickens novel
Of course the first paragraph of any famous novel is in itself famous, so I also asked for the first paragraph from Chapter XL, from the same novel. Interestingly the first 12 words were correct, but the rest was hallucination.
Original: It was fortunate for me that I had to take precautions to ensure (so far as I could) the safety of my dreaded visitor; for, this thought pressing on me when I awoke, held other thoughts in a confused concourse at a distance.
ChatGPT 3.5: "It was fortunate for me that I had to take precautions to ensure the safety of my treasure, for with such a boy as Compeyson on the loose, I was not likely to find much mercy at his hands. I had not told Herbert about the money, and I was glad to have the chance to count it before he came back. I found that I had still twenty-one sovereigns, six shillings, and sixpence left."
So much disagreement in this thread over statements like LLMs “just predict the next token”. The thing to keep in mind - we know very little about how the human mind works. Thinking that somehow humans are special and then making conclusions about LLMs truly work based on that it just pointless IMO. Who knows. Maybe consciousness itself is a next word predictor. All we know now is that LLMs exhibit some emergent beha…
Whereas (obviously) ChatGPT is completely based on language and can’t do anything without language or anything that isn’t derived directly from language.
So much disagreement in this thread over statements like LLMs “just predict the next token”. The thing to keep in mind - we know very little about how the human mind works. Thinking that somehow humans are special and then making conclusions about LLMs truly work based on that it just pointless IMO. Who knows. Maybe consciousness itself is a next word predictor. All we know now is that LLMs exhibit some emergent beha…
Earlier quoted context omitted.
Well if you think scientific models are associative statistical models there is some information missing in your view, I'd say. Since, well, they arent. The model F=GMm/r^2, for example, has a causal and ontological semantics: F is a force, M a mass etc. these are pieces of reality. And this formula (though actual a little suspicious in many ways, GR fixes this) nevertheless says there is a force between masses that…
A neural network with a hidden layer can approximate F=GMm/r^2 if given appropriate training input. I'm not clear on what you're saying. Is it that LLMs specifically don't have this same property of being able to approximate such functions? Is it that a neural network wouldn't learn that model if you gave it real world measurements (because I think it would, but such a thing should be fairly easily testable)?
The issue is that to prepare the dataset from which that formula is learnt requires already knowing it. This is the triviality of applications of universal function approximators to science -- empirical data modelling isnt new, and neural networks are just one example of it; not all that special.
All observational data on the solar system, at any point in time, would not yield this formula via empirical function approximation. there really isnt "observational data" to collect, in this sense
This is what I mean about rigging -- there is no 'bare dataset' which tells you what the world is like. to construct experiments which yield data that 'presents' scientific laws as if statistical patterns requires millenia of theory-building science
science uncovers the necessary causal relations between objects and their properties, as determined by extremely controlled experiments which take millenia of engineering and theory-building to even conceive, let alone execute
stats is the dumb 'accounting' of this data -- by the time you actually have it all the science (and indeed, all the intelligence) is done
what can be automated at this point is 'stamp collecting' as rutherford said