This is peripheral to the question, but pertinent:
I'm perplexed as to why the company is seeking to destroy itself in this manner. While a useful amount of AI output is reasonably accurate and useful, a significant portion is either very misleading or entirely fabricated.
ChatGPT will:
- Produce totally fake references,
- Produce wholly non-existent web links.
- Make comments which, when challenged, it will disclaim and say something entirely different.
- In some cases contradict itself in the same answer.
The only way to deal with these issues is to either have an expert knowledge greater than the level in the answer, or to use it only as a source for cross referencing.
- By allowing essentially uncontrolled use the company will destroy the integrity of its core asset.
A relevant example: I am not a Quora member, but I read a significant number of their answers in my (relatively wide) areas of interest. I have been surprised to see some Quora members, obviously using ChatGPT for essentially their whole answers, producing many factually incorrect and hardly relevant answers, but receiving numerous upvotes from other members.
While Quora still contains a very large percentage of genuine and highly useful human-generated answers, it is obvious that they have not chosen to actively oppose the posting of complete rubbish to their site. The likely outcome seems clear. Why Stack Exchange inc would wish to follow the same path is puzzling.
Added:
The company may be inadvertently "throwing the baby out with the bathwater". The proposed use of AI to improve asked questions has someSOME prospect of being useful.
Experienced users can ignore the advice (as long as it is only advice) and many beginners would benefit from good advice.
Worst case, it may make questions worse in some cases (but possibly not too many) and bad questions can be and presently are treated badly. This would be a negative for newcomers, but it could be ameliorated by thoughtful community action if members are prepared to actively help newcomers. Based on present practices, this seems unlikely; I perceive the site and the majority of member actions creating a substantial barrier for new members. Improving this is hard.
Overall, AI may quite possibly improve questions on average. This impression may affect the company's perception of AI answer improvement, which is far less likely to be net-positive overall.
One area where I have reservations is the suggested use of AI generated code to improve questions. This is very much a "double edged sword" as AI code is often good, but may contain subtle errors leading to 'rubbish out'. If new user code related questions suddenly sprout AI-generated code examples, the quality would need to be extremely high on average to not produce negative reactions overall from the community.
