They say… one of the keys to successful blogging is regular and systematic publishing of posts. Well dear reader, it has been six months since my last entry and alas I must admit I’m just not that kind of blogger… but here we are.
The last few posts here really took off and far surpassed anything else I have ever published in this space, also garnering several new subscribers – hello new folks. This is likely because those posts were about generative AI and specifically ChatGPT. I’m of the firm belief that anything published on this topic in the last six months was going to skyrocket but I’m flattered that some found those posts useful.
Over the last six months I’ve found myself somewhere between fascination and boredom around the bots.
…They also say, the more something changes the more it stays the same. The headlines say this is big change in the fabric of the tech landscape but there is a part of me that can’t help but feel a little… meh. I’ll admit something seems big and earth shaking but something also seems blasé – like we have been here before. Everyone seems to be talking, all at once, and there is a lot of overlap in what is being said. It seems everyone is evangelizing about how this tech will change the world for good or bad, but the thing about the world is that it is in her nature to change no matter what.
I have just not been that interested in adding my voice to the chorus and saying even more of the same. I’ve been quiet on purpose. I’m not in a rush to push out my next post/article/hottake around generative ai. But I am still reading, I’m still listening, and I’m still thinking. And a six month update on where my head is at seems… reasonable?
So this is just an update with some of the things on my mind right now regarding generative AI in higher education:
Enterprise Access and Other Integrations
A big part of my past concerns with ChatGPT have been around privacy. Even with numerous examples of what say… social media companies, have done with our data, people still don’t seem to have a good sense of platform literacy. Many still sign up for accounts and apps with no regard. It is just an email address, a phone number, oh look I can log in with my google account – how convenient! [broken heart emoji] This led for me to call for better digital literacy/citizenship but I have been doing that for a long time now and it only goes so far. (No shade on others who do that work – I just want more of it).
But word on the street is that enterprise access might be on the way meaning you don’t sign up with a personal account but with an institutionally recognized account. OpenAI mentioned this when they were forced to expand privacy functions of the bot because of the Italy ban stating: “We are also working on a new ChatGPT Business subscription for professionals who need more control over their data as well as enterprises seeking to manage their end users.”
Though there was no mention of “educational” enterprise access this is an interesting prospect and I’d have to think if it is available for businesses that education could sign on if they so desired/can get the lawyers on board. It is a prospect that does give me some hope of relief. I would hope that this would mean there would be some education-specific data safeguards negotiated by university/college higher ups. That there would be some expanded restrictions on the sale of personal data to third parties for instance and oh I don’t know, maybe that some of that stuff in FERPA would be considered as another level of protections above commercial offerings. But these are just my hopes.
I also admit that I might just being naive. It is important to call out that this kind of access, if it were to come to pass, would simply be a power shift. Remember “end users will be managed” by someone (your boss or your boss’s boss rather than OpenAI) and the specifics of all of that are likely to be buried in technical deployment details and more contract legalize. And who even knows if that access or language will be accessible/understandable. I’ve seen folks asking data questions of their institutions go down that rabbit hole to be met by those in authority who tell them that language is buried in closed contracts and access to those interfaces are only available to certain administrators. They are then faced with filing a FOIA request to try to get answers. And that’s a just a great look when you decide to ask for that raise.
So, educational enterprise access is on my mind this summer. I’m wondering what it looks like in terms of licensing and pricing. I’m especially wondering what it will mean for educational data privacy. But I’m also wondering if institutions will be able to train these models with enterprise access to behave in specific ways the way that some other educational integrations (who have been granted special GPT-4 API access) have done. I imagine recent announcements about lower prices and “function calling” with APIs will further enable things? But I’m wondering what it means when some schools can afford such access and others can’t. Some may just prefer to use the free web access but, despite all the inevitability rhetoric, I question how sustainable a free web interface of this tech is given the cost of keeping it running and the regrets of its creators. But I imagine that free training data is worth a ton to them.
Okay, what else…
Cheating and Detectors
Detectors continue to be front of mind for me especially with the Turnitin mess.
I continue to be flabbergasted by this one. I mean why aren’t more people talking about this? I know there has been a good amount of press but I really don’t know why it is not the top story in every higher ed and edtech outlet.
In the this round of AI arms race madness in April Turnitin decided to turn on an AI detection feature that that they had never tested in a real world environment and to not allow any of their customers to opt-out. Rather than offering this functionality to a small number of test schools and piloting it to get some feedback, they just forced it on everyone.
They claimed it had a 1% false positive rate but, surprise, that proved to be low after they actually started using it in real schools. They have never provided access to the specifics of their internal research or testing data. They have just expected everyone to take their word for it that they can detect synthetic text (nevermind that is still very much a complex subject area). Oh oh and they are not just adding functionality to their product out of the goodness of their hearts – no this is a paid feature that is “free for now”. This is the drug dealer, first hit is free, sales tactic if I’ve ever seen it but it is worse because these are existing customers who can’t opt-out.
And the infuriating part is that, even though some pushed back and were able to get them to allow an opt out, most schools (in the US) just let them do it.
I’m still not sure where cheating begins and resourcefulness or collaboration ends but I’m skeptical of black boxes that give us nice percentages make it all look super clear and easy.
Final Thoughts – for now
There are a ton of other things running through my head too. Economic and labor impacts are a big one, as is climate. I especially enjoyed this Marketplace report featuring Hugging Face’s climate lead Sasha Luccioni because of the “is it worth it” positioning. She makes the case, for instance, that search is already very good, that it already runs with AI but that the current models are way more efficient than LLMs. From the article:
“I’m not against innovation. I don’t think we should all just stop doing AI research. But for me, it’s kind of the basis of that research to say “this thing costs this much” not just in terms of money, but also in terms of planetary and human costs. Then, if that calculation makes sense, then yes, we’re going to use the AI. But currently, we’re not making that logical kind of decision. Right now, it’s more like “why shouldn’t we use the ChatGPT to do web search?” But I think environmental factors should be considered, because the new technology is lot less efficient than the current model.”
I wonder what the “is it worth it” position is for education?
Of course I’m continuing to pay attention to all the different ways that people are proposing that we use ChatGPT in the classroom. I like some of the ideas. They seem neat. And I’ve seen plenty of people warning students about bias and how the bots can straight up get it wrong sometimes. And that is a good thing.
But I still continue to see many people skip the privacy issues when talking about these things. Even as companies continue to ban this tech and create all manner of policy around how it can be used by their employees for fear of leaking company information to the bots. But it is just students’ personal data – what could happen? With all this talk about teaching students prompting skills I’d think that there would be a place for this – it seems like it would transfer nicely to the workplace.
When is it worth it to use a chatbot in the classroom and when is it just window dressing/look at me being an innovative professor? I’m not exactly sure but it is the question I’m most interested in right now.
So, that is my summer check-in. Perhaps I’ll do another in six months – maybe I’ll even post on something other than generative ai.