Hacker News Reader

Comment by Tostino | Hacker News Reader

Personally, the difference between GPT4 and 3.5 is, pretty immense for what I am using it for. I can use GPT 3.5 for things like summarization tasks (as long as the text isn't too complex), reformatting, and other transformation type tasks alright. I don't even bother with using it for logical or programming tasks though.

View on HN

One way that I've been framing this in my head (and in an application I'm building) is that gpt-3 will be useful for analytic tasks but gpt-4 will be required for synthetic tasks. I'm using "analytic" and "synthetic" in the same way as in this writeup https://github.com/williamcotton/empirical-philosophy/blob/m...

This is my experience too. While I'd really love the Open Source models to catch up, currently they struggle even with dead-simple summarization tasks: they hallucinate too much, or omit essential points. ChatGPT don't often hallucinate when summarizing, only when answering questions.

Would you please be more explicit? I'm curious about the relative strength's and weaknesses other's see

I can use GPT 4 for to work through problems that I have not actually figured out previously by talking to co-workers who work in my industry. I need to feed it contacts for my industry explicitly within the prompt and ensure that it understands and doesn't hallucinate its answers. However, that doesn't mean it's not useful, it just means you need to you understand the limitations.

I don't think I understand the process you are describing because it lowkey sounds like you are giving it information about your peers and using that as a basis to ask questions of chatGPT but get the benefit of the real people's perspectives?

Also I agree it can be super useful, its just that my own use of it is very limited (basically as the research AI I always wanted), so I am trying to broaden my perspective on what is possible

Heh, autocorrect messed up context for contacts. Sorry for that. I was talking about context for the problem I am working through. So it's not zero work to just "ask the AI" ang get an answer...you need to know essentially what you are missing first, and ask relevant questions.

Can you provide an example?

Reply on news.ycombinator.com