late 20s, AI-queasy, historian in training, frequent flier at museums.
Intellectually, I know there are theoretically ethical, reasonable, and appropriate things we can do under the vague "AI" banner at a museum/archive. In practice, I do not think most (if any) generative AI is a good idea at museums and archives.
Most of what is produced, particularly when we're talking about generative AI through LLMs, is just garbage though, so fairly or unfairly, my association with "oh we used AI" is suspicion at best and garbage at worst. It doesn't read as "we used this fancy tool, we're with the times" it makes me question if there's an AI-dependency issue there.
I have colleagues and mentors who use AI in limited ways that I think are Fine, I Guess, even if I wouldn't personally. There are case uses, particularly in processing massive amounts of data that hey, I get it.
But by volume, most of what I see both inside and outside the academy from LLMs is garbage.
Saying "we used AI" in a general vague sense is not a selling point because unless I have reason to believe otherwise, it's probably crap generated by a hype machine. It makes me take you less seriously.
I'm less curmudgeonly if you can specifically say what you use "AI" for and why. Like if you have 50K records and you're figuring out how many borough salesmen named Jack are listed, I'm not going to be a jackass and insist someone count by hand.
There are case uses, and I'm going to react differently to "we did x using y" than a wholesale buy-in to the hype machine.
I'm not fond of how GenAI is impacting communication (written and oral). I don't think any individual institution has the responsibility to fight this alone, but I'm going to take you less seriously if you're contributing to this. The same goes for the normalization of GenAI as a thing everyone "needs." It's not a need for the museum or me.
Proud, widespread use of GenAI at an institution makes me wonder how engaged your staff is with the material. Maybe that's not fair, but if I see a lot of generated text, generated images, or generated audio it would make me wonder. I also don't want to go to an archive, ask for human expertise, and get directed to the bot. If that happens, I'm going to think there isn't any human expertise to be found.
Maybe your staff is all great about checking behind AI's work, but most human's default setting seems to be not doing that. How many lawyers have been caught out on non-existent citations, doctors on impossible diagnoses, reporters with false claims, students leaving prompts in? Do you really trust all your interns and coworkers and admin that much? Volunteers, if they get access to it?
This is more of an institutional health thing than a personal preference thing, but what if the AI goes broke? None of the big names are profitable right now because it's extremely expensive to run LLMs and very few people pay. Do I trust an institution that is (or appears to be) dependent on a tech that could burn out financially in the next few years?