It would be easy enough for OpenAI to block this kind of use, both in their TOS and technically. It seems like there would be plenty of other reasons not to support unlimited chats long term.
It should not be necessary to set up technical or written guards against every possible act of bad behaviour.
Instead be nice. Don't be that **ole who steals the whole candy bucket on Halloween just because you can.
> Do not attribute to malice that which can be attributed to really weird usecases.
see also, xkcd.com/1172 (spacebar heating)
I think the core issue isn't the abuse, but the very existence of a service that claims to provide infinite of something. Infinite is always BS because resources aren't infinite.
It's just like "all you can eat" places kicking out people after they've been in the restaurant for a while. We should discourage these marketing tricks because they are always a lie.
Regarding Halloween, we stopped leaving unattended buckets out because of bad actors taking everything. Now we hand out the candy ourselves. Our home no longer provides "all you can grab" treats...
We left a bucket out while we were out with the kids. No issue. Had too much at the end of the night, and had to tell the last stragglers to just take handfuls when we got back. Maybe that's a good, quick temperature check on whether your area is high-trust/low-trust.
* Lets people use whatever approach they prefer (- and if someone wants to keep their existing setup they can - with reasonable limits)
* Minimizes risk of users getting banned, including due to A) activity they didn’t know was disallowed, or B) false positives from OAI
* Heads off different future loopholes
* Avoids adding more TOS complexity
* Avoids users being uncertain if borderline activities will get them banned