I teach my writing class today. I'm supposed to talk about ChatGPT. Here's what I will say.
NEVER USE CHATGPT. YOU ARE HERE TO LEARN HOW TO WRITE ABOUT SCIENCE, YOU WILL NOT ACCOMPLISH THAT BY USING A GODDAMNED CRUTCH THAT WILL JUST MAKE SHIT UP TO FILL THE SPACE. WRITE. WRITE WITH YOUR BRAIN AND YOUR HANDS. DON'T ASK A DUMB CYBERMONKEY TO DO IT FOR YOU.
I have strong opinions on this matter.
@pzmyers I'm right there with ya!
My school asked us to do an "AI" day and plan activities to explore LLM's and such.
I sat each of my classes down and we had a serious discussion about the ethics and shortcomings of LLM's. Almost none of them were aware of how the program gathers (copyrighted) data or the immense environmental costs.
Personally, I struggle to understand why anyone would want an LLM to write for them. The point of writing is self expression!
@AlexCorby @pzmyers
> The point of writing is self expression!
That may be true for some writing, but a *vast* amount of writing doesn't contain any personal expression at all; it just needs to be accurate and descriptive (think car manuals).
@AlexCorby @pzmyers Oh, you definitely wouldn't, no. But an LLM model trained on technical manuals, with a low temperature setting, and fed the facts it needs to impart, could do a pretty good job of it.
The key is that LLMs are tools. Like most tools, they can be used properly to enhance a work, or be misused as a crutch.
@AlexCorby @LouisIngenthron @pzmyers > I'm not sure I want my car manual written by ChatGPT either.
After finding out just how badly ChatGPT confabulates junk, I asked it to give me the oil change procedure for a 2017 Chevrolet Bolt.
It dutifully gave me a very generic oil change procedure summarized from who knows where, which was very amusing, since the 2017 Bolt is an electric car with no engine oil to change.
It would be less amusing if this were an urgent or emergency situation and I definitely needed the right answer for an actual car problem right away.
@dpnash @AlexCorby @pzmyers Lol, yeah, that's a great example of using a tool the wrong way. Trying to get specific facts out of a generalized language model is like trying to hammer a screw: Sometimes it works, but a lot of the time, it does more harm than good.
@LouisIngenthron @pzmyers
That's a fair point. However, with some of the nonsense that I'm catching coming out of these programs, I'm not sure I want my car manual written by ChatGPT either.