Stackexchange (SE) is currently subject to 'Mod action' over its policy towards ChatGPT content.

SE's initial policy is effectively to ignore ChatGPT (advanced AI) generated content if it was posted as an answer on SE. The rationale was because ChatGPT (advanced AI) detectors produced anomalous rates of false positivity with non-native English speakers from SE's internal testing. In a subsequent, more lengthy explanation the ban on ChatGPT detectors seemed more nuanced and focused around the possibility or else actual occurrence of erroneous SE account suspensions/deletions.

The underlying concern from the Mod-perspective is AI content will overrun SE content. There are other factors however.

The Mod action affects the site as follows:

  • Mod @terdon is on all out strike;
  • I perform essential Mod duties, but act in solidarity to the strike and have requested a meeting with SE HQ;
  • Mod @gringer is fully active.

In my case I follow due process for Mod action.

Does ChatGPT affect Bioinformatics SE?

No. I do ML development, I know it's what it's about. It would be really hard for a supervised learning algorithm (the workhorse of ChatGPT) to train against the biological context of any or all bioinformatics problems. In bioinformatics the biological context is pivotal to the solution. That layer of abstraction makes training hard, because the data scientist constructing the AI needs to understand the abstraction. At a minimum any question is spilt across 5 kingdoms. It would need a biological data scientist to help build a successful AI.

In the worst case scenario it would mean loads of Biostars content turn up here and vice-versa. That happens anyway.

ChatGPT detectors and Bioinformatics SE?

We'll never use AI detectors because it's not relevant to bioinformatics biological context (above). Every question and answer its checked. We'd spot and comment if there's a gap between the biological context and the response, when not striking of course ...

If there are any urgent issues flagging them by posting on meta is a good approach to getting a fast response.



You must log in to answer this question.

Browse other questions tagged .