curtismchale
commented
6 months ago @lukefretwell @kevindherman
I think we should use AI like we should have used crypto...not at all. From the disclaimer on the tool you linked to.
You are solely responsible for monitoring, reviewing, and managing the content generated by the AI Engine plugin, as well as for dealing with any abuse or issues that may arise from its use. - AI Engine Disclaimer
According to a report today, all the major image generator AI tools have CSAM in them, and there is no way to "clean" them now that they've been poisoned (read Not with a Bug for more on that). Outside of giving our users bad information which they'll use in their content no matter how much we tell them not to, I'm very concerned that AI will leak sensitive data back into the model, because it happens all the time and there is no way to fully protect AI against adversarial attacks.
From the OpenAI Privacy policy on what they collect.
Log Data: Information that your browser or device automatically sends when you use our Services. Log data includes your Internet Protocol address, browser type and settings, the date and time of your request, and how you interact with our Services.
I highly doubt that many of our city users understand that they can leak private data from citizens to OpenAI by their usage of the service. We should be getting a legal opinion on any usage of AI with our sites, and our customers should have a legal opinion on this as well. We certainly should not be turning on anything AI without a deep understanding of exactly what we expose ourselves to in the way of legal risk.
Overall, I think that AI is a bubble which will leave nothing behind but a bunch of power used and cheap chairs from services trying to liquidate their office furniture because they can't afford it anymore.
Source
The source of the issue (ex: Customer: #CustomerName#).
LF
Feedback
Opening this thread up for discussion on how we might incorporate GPT/AI tools into ProudCity.
QA
Notes:
Links: