61–70 of 368 posts
Earlier quoted context omitted.
It's interesting for sure. There's also the fact that he didn't use the sota (GPT-4). I'm not saying that would be perfect. But i would be surprised if it wasn't big leap over 3.
it wasn't mentioned anywhere in there which version of the model he was using... how are you so sure?
I know everyone scoffs at the concept of 'prompt engineer', but it really is an essential craft that we're going to have to come to terms with when interacting with large language models. Seeking suggestions on a more comprehensive prompt: https://sharegpt.com/c/sckAPvV Trying it out: https://sharegpt.com/c/LbpEIxi
children: the OG AGI
Back in 2004, took a seminar in college regarding Decision Support Systems and how they a manager or doctor could ask it a question and get a response to help them make a decision. Went to the doctor couple a years ago and he charged $300 to google search the symptoms. No thanks.
You're expecting a doctor to have all relevant medical knowledge permanently memorized? That's the equivalent of coding interviews on random obscure topics where you can't look anything up. Like a SWE their value is not perfect recall of every area of CS/medicine but ability to decipher arcane documentation into actionable outcomes.
Would be great if OPENAI can get hospitals around the country could anonymize some trainable format data
Sometimes a patient will reply with something like “I can’t be.”
“But how do you know?”
If the response to that follow-up does not refer to an IUD or a specific medical condition, it’s more likely the patient is actually saying they don’t want to be pregnant for any number of reasons.
Funny how languages are ambiguous around "can't" and "don't want".
Being a scifi geek and AI geek and neuroscience geek for pretty much the past 40 years, I've read countless predictions and scenarios and stories about society's response as "true" AI begins to emerge. So watching it play out for real is creating this bizarre sense of deja vu combined with fascination and frustration and also some anxiety. This article and the comments in this thread are right up that alley. I mean,…
I'm also a Civil Engineering grad and it really cocked up when I asked a question about normal beam strength and I suspect it will get the pressure at depth thing wrong too. Both of these are related to depth only - one dimensional. That's why we have I beams - the flanges are to resist twisting and buckling but the strength is in the depth.
ChatGPT is a tool. It is basically a really good calculator. I do find myself saying please in my requests.
Being a scifi geek and AI geek and neuroscience geek for pretty much the past 40 years, I've read countless predictions and scenarios and stories about society's response as "true" AI begins to emerge. So watching it play out for real is creating this bizarre sense of deja vu combined with fascination and frustration and also some anxiety. This article and the comments in this thread are right up that alley. I mean,…
Half the time right is worse than nothing at all.
I'm not really sure what he expected here, ChatGPT was not trained to be a doctor, it is far more general than that. Asking ChatGPT for medical advice is like asking someone who is very well read but has no experience as a doctor, and in that context it's doing very well. He also brings up one of the most salient points without really visiting it enough: ChatGPT does not ask for clarification, because it is not a kno…
Earlier quoted context omitted.
But at the same time consider the problem, and if AI is even the tool for this job or if its total overkill. We are just matching symptoms to a diagnosis, you could get this utility from programming a simple dictionary with a fraction of the compute. Maybe even with a confidence estimate of the diagnosis this time. There's an old adage that rings true here I think: "keep it simple, stupid."
But I think that's missing the "magic" of it—for lack of a better word. In the situation you describe, you have to build a specific tool for the task. With LLMs, we appear to have stumbled on a general purpose tool for a myriad of situations. Yes, they're not perfect; but the fact that they work remotely as well as they do even at these early stages is astonishing.
On the other hand with chatgpt, you have to build a specific prompt, and you don't know if that will lead to a consistent result or what its even pulling from. It introduces uncertainty that you cannot reliably estimate, whereas with the above toolmaking example, you can estimate your uncertainty, you can calculate how much power you have and how much you should have to resolve what you expect, and you can calculate a false positive rate.
When decisions mean money or life, I wouldn't rely on the black box chatbot. I would take a little bit more time and make sure all i's are dotted and all t's crossed.
Earlier quoted context omitted.
> as "true" AI begins to emerge What, exactly, is that?
The definition has always been the Turing test. Nobody has been able to provide a more precise answer.