can we please stop pretending generative AI has a future

The school district in which I work released its official policy on the use of generative AI today.

I haven't read the policy yet - I was busy picking up a book order - but based on the email that presented it to the staff, I expect to have Strong Opinions about it.

Said email, from the assistant superintendent, is unabashedly pro-gen AI. It breathlessly cites statistics about "gen AI startups" (approximately 5,000 in the US, per said email) and "investments" ($94 billion-ish). It announces excitedly that gen AI is most definitely the wave of the future and that, as educators, it would be irresponsible for us not to incorporate this technology into our work!

...Click one tab over in my beleaguered work browser, and there's a long email thread from the AASL listserv on whether we should even allow the use of generative AI in our classrooms, let alone promote it.

Among the biggest listed concerns is the fact that generative AI is still a plagiarism machine. Your prompt may not be plagiarism, and citing the output of that prompt in your work may not be plagiarism, but the output itself most definitely is.

It does us no good to show students how to "cite their work" when they use generative AI. Use of generative AI is still plagiarism. The machine does the plagiarizing.

Also, it's not "just" plagiarizing the data generative AI has already hoovered up, although that would be quite bad enough. Generative AI models continue to steal work from every single person who posts things online. Generative AI will probably steal this post. I do not give permissions, but I currently have no way to stop it and no way to hold its creators accountable.

Given the active plagiarism these models commit, using them remains an ongoing case of serious academic and intellectual dishonesty. As one AASL contributor put it, using generative AI to create classroom materials is like stealing a workbook, photocopying it, and handing out the photocopies to your students. Your students' use of the workbook might fall under "fair use," but the way you acquired and distributed that workbook definitely does not.

None of this appears in generative AI stuff aimed at teachers, however. The closest thing I've gotten to actual concern from any teaching listserv I read - including the district's own "ALLSTAFF" mailing list - is a link to a spreadsheet that attempts to rank the "privacy" of prompts entered into various gen-AI-using edtech products.

I can't explain the details, because I didn't read them, because I rolled my eyes and closed the tab about thirty seconds after opening it. Privacy and gen AI is a fractally bad subject. It is bad at every conceivable level of magnification. I don't even know where to begin explaining the problems with it.

...Which, honestly, is where I am with this district policy, too. I don't even know where to begin explaining to people who are so excited about AI that their glee is at best unfounded and at worst actively dangerous.

I don't care how many venture capitalists are pouring how much into how many "startups." There remains no meaningful use case for generative AI that will allow it to make the money it requires to run. Not even close. OpenAI's ChatGPT BURNED NINE BILLION last year. That's $5 billion after revenue. Revenue includes its paid subscribers. Even paid subscribers cost the company more than it makes in making them pay. And OpenAI has been renting GPU time from Microsoft at something like a 75% discount (though that's likely to stop soon, as Microsoft recently rolled back plans for expansion amounting to something like 14% of its total GPU offerings).

What happens when companies and individuals are forced to pay what it actually costs to generate each prompt and response? Can't nobody afford that. Certainly no one in their right mind would pay for it. Human labor becomes cheaper again at that point.

I don't mind that the district has a generative AI policy. I mind very much that we are all pretending generative AI has a future.

https://www.wheresyoured.at/power-cut/

https://www.sciencedirect.com/science/article/abs/pii/S0301051123001382