It would be helpful if GPT could identify when a user accidentally types just the ‘v’ key and sends it instead of executing a ‘ctrl + v’ to paste text, recognizing it as a likely error, and answering that GPT assumes this error happened, would be great.
You gave an interesting example. Nevertheless, don’t you think GPT can predict the general meaning of the topic as it did in your example, but when there is a topic such as providing part of code, it also predicts the error, etc?
By the way, this is not a problem, but it would be interesting if it could predict this error as well. In the end, this can add to “users typing features” where speed of typing, mistakes in the words, time it takes between sentences, etc.
It does not automatically, at least not as a single shot prompt.
My custom instructions put me in the realms of AI dev, scripting, and linguistics myself, and when I tried a one-shot prompt to get it to talk about Poe it gave me answers about Edgar Allen Poe, not Poe bots.
LLMs like GPT bases its predictions on the most probable usage based on its entire training data, filled with vast swaths of different use cases.
Something like what you’re asking would come with personalized models, but we’re not there yet (unless you’d like to fine-tune models yourself).
Models can’t predict errors. It can identify and analyze for certain errors based on provided code given context, and it can run code to identify errors and readjust its approach based off such errors (using the data analysis plugin), but in terms of predicting the most likely resulting error, I would not expect that any time soon.
1 Like