Background:
In the year 2023, machine learning based chatbots finally became superior to the average human*. One interesting feature of these chatbots is that they provide the ability for the user to regenerate an answer.
Here’s how the ChatGPT system handles answer regeneration:
- USER: Please inform me of one vegetable that tastes good. Only provide one vegetable as a response.
- ChatGPT:
Asparagus is a vegetable that many people find delicious […]
- USER clicks “🔄 Regenerate Response”
- ChatGPT (Regenerated):
Eggplant is a vegetable that is known for its unique taste and texture […]
Not bad at all!
[*] If you don’t believe this, please spend 10 minutes (or however long it takes to change your mind) reading the comments on your local newspaper’s web site.
Proposal:
What would be even better, though, is the ability to use “regenerate response” in real-world conversations, too.
Imagine if any time someone said something dumb (and/or regretted saying something), anyone could utter the phrase “REGENERATE OUTPUT” to give the speaker a socially acceptable “second chance” to answer the question (Figure 1).

The ability to “rewind” and fix conversations would be hugely beneficial in nearly every social interaction.
Conclusion:
This could be expanded to include text messaging and other forms of textual chat as well. Unless you’re a huge moron, you should be able to see how this would be useful. REGENERATE OUTPUT. It’s likely that everyone can see how this would be valuable.
PROS: Would provide a straightforward way for people to improve their social interactions.
CONS: People might become dependent on the ability to revise their answers. This could be awkward in trials, e.g. Prosecutor: “Where were you on the night of the crime?” Defendant: “Uhh, committing the crime? Wait. Regenerate output. I was asleep at home.”
You must be logged in to post a comment.