I hired an editor once who added more mistakes to my writing he removed.
ChatGPT has a hypnotic ability that comes from its being trained to pick the least surprising next word. It is very fatiguing to catch its mistakes and if you falter and let one through your only consolation is that it will also hypnotize most of your readers.
And, very quickly its style will be recognizable and uninteresting.
Folks might want to keep in mind that the application they have in mind for this customer service and tech support automated assistants. Think, "Calling the cable company to ask about your bill." For this kind of stuff, it will be fantastic, especially if it is trained on real documents from the company who plans to offer it as a chat. There, too, accuracy is not a problem if the answers come from official sources.
It should not, however, be used as a substitute for your own human work. Anyone doing that is going to quickly look dumb, get people hurt, or get fired.
I don't get why people keep saying its style will start to become recognizable. You can have it write things in any number of styles and levels of verbosity.
Yes, and none of the styles are original or yours. This will become transparent and undesirable. There's no threat to humans who can do their own research and come up their own novel analyses and conclusions. Folks who can't do that? Well, it might help them churn out crap.
Like I've said previously but seemingly unheard, it's a tool for chat support agents. It's not a research or learning tool. It's not a writing or composition tool. People who use it that way will be disappointed.
Opposite. ChatGPT has lowered the bar to entry of creating drivel to probably 2c/750 words. When I saw it can produce text as good as some stuff I have written, it made me realize more imagination is needed to make my writing more interesting or go deeper into a subject than a bot can possibly do. i’m
14 comments
[ 1.7 ms ] story [ 45.8 ms ] threadChatGPT has a hypnotic ability that comes from its being trained to pick the least surprising next word. It is very fatiguing to catch its mistakes and if you falter and let one through your only consolation is that it will also hypnotize most of your readers.
Folks might want to keep in mind that the application they have in mind for this customer service and tech support automated assistants. Think, "Calling the cable company to ask about your bill." For this kind of stuff, it will be fantastic, especially if it is trained on real documents from the company who plans to offer it as a chat. There, too, accuracy is not a problem if the answers come from official sources.
It should not, however, be used as a substitute for your own human work. Anyone doing that is going to quickly look dumb, get people hurt, or get fired.
Like I've said previously but seemingly unheard, it's a tool for chat support agents. It's not a research or learning tool. It's not a writing or composition tool. People who use it that way will be disappointed.
No it won't.
It can't tell you what you need if you don't know to ask for it.
The how that ChatGPT can address is really the smallest part of my struggles, most of the time.