5 ms·
I think that insight is an important feature that GPT doesn't seem to have, at least not yet. For instance, I've seen people saying they can ask it for help wi
by Falkon1313 3y ago
I think that insight is an important feature that GPT doesn't seem to have, at least not yet.
For instance, I've seen people saying they can ask it for help with how to code something and it will help them. Although the results aren't perfect, they can be helpful.
However, I recall years ago asking a more senior developer how to do something. They could've just told me how, but instead they asked why I was doing that. Then they told me not to bother with writing code at all, and to instead do this other, much simpler thing which would have the same effect at solving the client's request. ChatGPT wouldn't have had that insight.
Years later, a junior dev came to me asking for help with his code because it wasn't working. We went over the code line by line and he explained what it was trying to do and it all looked good. But when he ran it, the web server crashed. I told him to try adding a comment:
// Do not remove this comment.
Sure enough, it worked perfectly and the server no longer crashed. Why? Because I realized that if his code was wrong he should get an error message, not crash the server. But sometime back I had read about a bug in a version of the interpreter that would crash when parsing a file that was an exact multiple of 4096 bytes. Would chatGPT have thought of that? Probably not. It would've just talked about the code.
Which is not to say that it's useless. But it lacks greater context and the insight to go beyond the parameters of the question on its own initiative. Then again, so do humans most of the time.
- bytefactory 3y ago> I think that insight is an important feature that GPT doesn't seem to have, at least not yet. I actually think this is a limitation of the RLHF that GPT has been put through. With open-ended questions, I've seen GPT4 come up with reasonable alternatives instead of just answering the question I've asked. This is often seen as the infamous, "however, please consider..." bits that it tacks on, which occasionally do consider actual insights into the problem I'm trying to solve. In most cases it seems to try very hard to mold the answer into what I want to hear, which in many cases isn't necessarily the best answer. A more powerful version of GPT with a less-restrictive RLHF seems like it would be more open to suggesting novel solutions, although this is just my speculation.