The New Grown-Ups: “Taxman Salamander”
With AI-generated lyrics.
Generative AI, trust, and distrust
At Inside Higher Ed, Jacob Riyeff writes about generative AI and its effect on teacher-student relationships. What breaks his heart, he says, are the ways in which AI makes it difficult for him to trust his students.
Off the bot
In response to a comment from Matthew Schmeer that describes inventive assignments to keep students from turning in AI-generated writing, I came up with a phrase that I’d like to share: “off the bot,” after “off the grid.” I am thinking and writing off the bot.
“Human interaction might be preferred”
I asked ChatGPT: “What would be some good reasons not to use ChatGPT?”
Another try at a Pinboard bookmarklet
I asked ChatGPT to write a bookmarklet that would add a URL to Pinboard and paste in text copied from the clipboard as a description. And it works.
In search of lost passage
I may have found the passage Elif Batuman asked ChatGPT to find.
Revising again
One more addition to How to e-mail a professor, now that I can get through to ChatGPT: Don’t ask AI to write an e-mail for you. At least not if you want your e-mail to sound like the work of a human being.
AI hallucinations
These are our lives on artificial intelligence.
Making HTML links with Alfred workflows
I finally found a good use for ChatGPT: I had it write two workflows for the Mac app Alfred. It took the bot about a half dozen tries over several days to produce workflows that work. Thanks, AI.
Chomsky et al. on ChatGPT
In The New York Times, Noam Chomsky, Ian Roberts, and Jeffrey Watumull consider (with examples) “the false promise of ChatGPT.”
Edwin who?
What can ChatGPT tell us about Edwin Mullhouse, of Steven Millhauser’s novel Edwin Mullhouse: The Life and Death of an American Writer 1943–1954, by Jeffrey Cartwright ?
ChatGPT’s twenty-line poems
ChatGPT has difficulty distinguishing “in the style of” from a hazy idea of subject matter. And ChatGPT assumes that almost all poets write in quatrains of iambic tetrameter, with clumsy rhymes and inspirational messages.
Another ChatGPT fail
What ChatGPT doesn’t understand is that Lillian Mountweazel was not a real person.
More ChatGPT fails
More 500-word analyses of well-known poems, from John Ashbery, Elizabeth Bishop, Emily Dickinson, Philip Larkin, and William Butler Yeats. In each case, ChatGPT produces phrases and lines nowhere to be found in the poem.
ChatGPT fail
I asked ChatGPT to write a 500-word analysis of Ted Berrigan’s poem “Red Shift.” It produced text with fake quotations that had nothing to do with the poem.
Spot the bot
The New York Times has a challenge: read ten short pieces of writing and figure out which ones were generated by a chatbot and which ones were written by children.
Teachers and chatbots
If I were still teaching, I’d adopt five strategies to counter the chatbots.
A 100-word blog post
Generated by ChatGPT in response to the prompt “write a 100-word blog post.” It’s fourteen words short.
I’m sorry too, ChatGPT
A couple of days ago Elaine was a celebrated pianist who had performed with orchestras around the world. Hot damn! And I was a writer who had won the Ruth Lilly Poetry Prize. $100,000! But now it has no info about us.