I don’t want to sound elitist, but I do wish that someone’s first use of ChatGPT would be locked behind a basic crash course in Natural Language Processing. Like, just 30 minutes worth of material would make such a difference in the user feedback. I made the mistake of subscribing to the subreddit, and 90% of it is the same complaints that would never come up if the users had a better understanding of:
What Large Language Models (LLMs) are
The limitations of LLMs and of ChatGPT as a service
The inherent biases resulting from the training data
The consequent need for active realignment and moderation
Don’t get me wrong, I desperately want this technology to be as accessible as possible, but if I see one more post to the tune of “y won’t the AI make a racist joke for me?”, I’m gonna lose it.
That’s a good idea.
You could probably have ChatGPT write the script. (Only half-joking!)
Maybe a few minutes? Very quick overview. I’m thinking a video tutorial that can’t be skipped? With a four or five question quiz at the end?
I do have to wonder if some of the “Hey, why didn’t the AI email me my dinner recipe? I’ve been waiting for seventeen hours now!” are just trolls. Most communities have them lurking about, and with AI writing, it’s not hard to sound this way or that way if you wanted to troll…
Anyway… I agree.
People aren’t ready for what’s about to unleashed…
Honestly, even if they aren’t, I can say with some certainty from my time in the trenches during the Great Troll War of ‘09: If there’s one thing that will dissuade a troll from engaging with a community, it’s education.
Lmk if this would be inappropiate for this forum, but I’d like to use this thread to compile examples of people completely missng the mark of what LLMs are capable of, as well as trolls sincerely complaining about the fact that alignment is even a thing.
P.S. Mind if I DM you about your projects? I foresee a rising interest in your work in the very near future.
They could have a crash course in ML as requirement to post on the forums.