Support #19605
openAI question from Ned Benton
0%
Description
Ned Benton asks:
nbenton@jjay.cuny.edu | January 26, 2024 9:20 am
Will CUNY Commons install any plug-ins that connect to AI services. I am thinking of this one but there may be others.
https://wordpress.com/plugins/ai-engine
I have posted in NESRI an article about AI and slavery records. Here it is.
https://nesri.commons.gc.cuny.edu/artificial-intelligence-and-slavery-research/
While this article reveals risks and deficiencies, I continue to think that AI could become a useful way to explain records of slavery at the individual level. We are working on a project we call NESRI 2.0 which involves adding several more fields to will enrich the interconnections between the 90,000 records that we have.
If we eventually do this system-wide, we have the challenge of presenting this to users. I first approach is to add a table to our locality reports that lists the families, in the locality specified for the report, with a link to click into the records of any family, so people in Boston can learn of families of enslaved people in the 1700s. But I also think that I can code the database to construct a test string that can be submitted as a prompt to an AI service to get back an essay about the family - the names and relationships and events in their lives and in the community when they lived. The service would have to have guardrails to not make up information.
Ned Benton
Updated by Matt Gold 3 months ago
Hi Marilyn -- thanks for this. Can you please tell Ned that we appreciate his inquiry and that we'll be discussing this at our next team meeting in early February?
First, here's the correct plugin site (on the wp.org repo rather than wp.com) to look at -- https://wordpress.org/plugins/ai-engine/
I think we need to give some thought to how AI-generated content could affect our ToS and privacy policy. For instance, if AI generates posts, chatbot content, or writing suggestions that violate CUNY's Acceptable Use of Computers policy (which forbids, for instance, the use of CUNY computers/platforms for commercial gain), who would be responsible and how would we handle this? It's one thing if we can hold a member responsible for the content they post; it's another if a member is using a plugin that we have enabled on the site.
Of course, there are also the related issues around whether these plugins work well/pose issues for a WP multisite install like ours.
At any rate, I think this is worthy of discussion when we meet.
Updated by Marilyn Weber 3 months ago
Reminder to put this on the agenda tomorrow.
Updated by Colin McDonald 3 months ago
Thanks Marilyn, this is on the agenda for Friday's meeting.
https://docs.google.com/document/d/1A5jdUmLnwQbHC1IuzgsEeHuyWPnKLNppeYoaa9HAQfk/edit?usp=sharing
Updated by Colin McDonald 3 months ago
Recapping our conversation in the monthly meeting today, Matt is going to reach out to Ned and we'll see how he wants to proceed from here. Our main concern is what might happen if the AI produces inaccurate or even offensive information, given the subject material of NESRI, and how we protect against that.
Ned mentions "guardrails" being put in place, but who will do that, and what will they be? How will they align with our ToS and privacy policies. Perhaps we could come up with a disclaimer here or a universal one for other future AI projects, but that will take some work to craft and also set up proper display to users.
We recalled that in #19192 we had disabled the Jetpack AI assistant that had suddenly appeared due to some similar concerns.
Updated by Colin McDonald 2 months ago
Adding Matt as a watcher here. Matt, just nudging this if you haven't reached out to Ned yet, or if you want to update us on his response.
Updated by Matt Gold 25 days ago
Hi Boone -- let's discuss this during today's dev call. Here are the two plugins Ned is interested in:
AI Power - https://wordpress.org/plugins/gpt3-ai-content-generator/
AI Bud - https://wordpress.org/plugins/aibuddy-openai-chatgpt/