Many of you have been generous in your help for my free apps that help persons with multiple disabilities communicate, please see Redirecting...
A well known physics professor in Delhi has ALS and only has a little head movement left, which costs significant effort. If I am to use that movement as a signal to control a typing app, it needs to be good at predicting what he may wish to type next, so as to minimize that effort.
I'm thinking of using the experimental chatbot component that lets your app send queries to ChatGPT and receive the response. My app would send queries such as "give me a comma separated list of 100 words that are most likely to start a sentence" and let the user select one. Effectively, this would let him type word by word. While I do plan to locally and on the cloud back up the responses, so that the same request isn't sent again to ChatGPT, there will still be plenty of traffic on the chatbot.
- How can I minimize the cost?
- I do not wish to pay for the typing my users do. Will each of them need an individual API key, in that case?
- Is there a better way to solve this problem that you could suggest?
warmly.
Arun Mehta