Will OpenAI remove, for >7d, any feature that was released to everyone in a tier, citing AI safety, before 2024-05-18?
May 19

"In a tier" means that all customers within a specific group of customers have that feature. For example, all ChatGPT Plus users have access to ChatGPT-4 -- if OpenAI removes ChatGPT-4, this market will resolve to YES, even though free-tier users never had access to ChatGPT-4.

"For 7 days" means that the feature has to be unavailable to users in that tier for at least 7 (continuous) days. An example of something that would have counted was the removal of "Browse with Bing" on 2023-07-03, followed by the reinstatement of that feature on 2023-09-27. An example of something that would not count is a series of intermittent hour-long outages of a particular feature leading to 60% aggregate uptime over a 1 month period.

If the feature was partially but not fully rolled out within its tier, it does not count. For example, as of now (2023-11-18), OpenAI states that GPT-4 fine tuning will be released incrementally to some, but not all, users. If OpenAI halts the rollout of GPT-4 fine-tuning without making it available to everyone, that would not count as "removing a product that was released to all paying customers".

If OpenAI removes a feature for reasons that do not touch on AI safety (e.g. cost, legal concerns), this market will not resolve to YES with respect to that feature. If they cite a reason due to the inability to guarantee safe results from their models, this market will resolve to YES, even if they do not use the exact words "AI safety" (e.g. "preventing prompt injection through uploaded files turned out to be intractable and so users can no longer process uploaded files").

This market specifically addresses OpenAI doing this to any feature between now (2023-11-18) and the resolution date (2024-05-18).

Get Ṁ200 play money

More related questions