The 'Megalopolis' trailer attributed fake quotes to real critics, possibly because they asked AI and it lied.
Do not let ChatGPT do your homework, kids.
Welcome to The #Content Report, a newsletter by Vince Mancini. I’ve been writing about movies, culture, and food since I started FilmDrunk in 2007. Now I’m delivering it straight to you, with none of the autoplay videos, takeover ads, or chumboxes of the ad-ruined internet. Support my work and help me bring back the cool internet by subscribing, sharing, commenting, and keeping it real.
—
You may remember Megalopolis as the long-awaited, self-financed, $120 million epic from Godfather director Francis Ford Coppola. Or, you may remember it as the film whose on-set behavior by the director sparked a feud between two trade publications, both with the same parent company, based on which members of the production each trade writer owed their access to. There are many reasons you could be aware of it, but the point is that now it has caused a whole new kerfuffle, thanks to a latest trailer that misquoted a bunch of famous critics, some of them dead.
Often the way that this sort of thing happens is that the marketing people in charge of the trailer just invent some glowing quotes and assign them to some prominent critics. That does happen from time to time, though it’s much more common to just give those quotes to the kinds of flunkie critics whose names are always showing up in trailers and posters for bad movies, as everyone sort of assumes happens. Another option is to just take negative reviews out of context (as has happened to me), or use random Tweets or user reviews and treat them like pull-quotes. Point is, there are many established ways to cheat, which most of us now recognize, and are generally part of the “the numbers are all fake” phenomenon.
The interesting wrinkle this time is that Megalopolis’s marketing team was trying to lean into the idea of Megalopolis being divisive and controversial, by using pullquotes from famous critics bashing Coppola’s past masterpieces. The message there being something like “history proved him right before and will again.”
The ad apparently quoted Pauline Kael describing The Godfather as “diminished by its artsiness,” and Andrew Sarris calling it a “sloppy, self-indulgent” movie. This alongside Vincent Canby calling Apocalypse Now “hollow to the core,” Rex Reed bashing Apocalypse Now, and Roger Ebert calling Bram Stoker’s Dracula “a triumph of style over substance.”
If Ebert had called Dracula that he would be entirely correct (in reality he said it about Batman). Trouble is, he didn’t say it, and neither did any of the other critics about the movies they supposedly bashed (Rex Reed actually did bash Apocalypse Now, just not with the quote the ad used). Vulture’s Bilge Ebiri seems to be the first to have noticed this, and I’m linking him because Lionsgate (Megalopolis’s distributor) pulled the ad before I could embed it here.
“Lionsgate is immediately recalling our trailer for ‘Megalopolis,'” a Lionsgate spokesperson said in a statement provided to Variety. “We offer our sincere apologies to the critics involved and to Francis Ford Coppola and American Zoetrope for this inexcusable error in our vetting process. We screwed up. We are sorry.”
I mean sure, but saying “sorry” and pulling the trailer doesn’t address the important questions this incident raises. Such as “how did this happen,” and “did you really think people just wouldn’t notice?”
It’s not hard to find real critics who hated on great masterpieces, which is something I’ve done myself for virtually every movie retrospective I’ve ever written. And if you’re already putting Rex Reed up there in the pantheon of great critics alongside Pauline Kael, what’s to stop you from digging a little further to find some crank from the Kansas City Star or whatever? In the days before the mediapocalypse, dissenting voices were easy to find.
As for the question of how this actually happened, one of the first tweets I saw on the subject was the NY Times’ Kyle Buchanan saying “All the ‘mean’ critics’ quotes in the MEGALOPOLIS trailer were made up and likely came from an erroneous ChatGPT answer. Incredible.”
I thought he was making a joke at first, but Corey Atad over on Blusky screencapped a prompt in which he told Chat GPT "I want three negative quotes about the Godfather from famous '70s film critics."
ChatGPT in turn appears to have produced quotes from Pauline Kael, Roger Ebert, and Andrew Sarris, which, while they don’t match the quotes from the ad exactly, are all similar sounding quotes that notably don’t appear in any of the original reviews either. When I prompt ChatGPT myself (in slightly different words), I too get answers that usually include people like Ebert and Pauline Kael, quoted as writing things that, as far as I can tell, they never actually wrote.
And so, in the absence of a genuine explanation from Lionsgate, we’re left with the probable conclusion that someone in the marketing department really did try to get ChatGPT to do their homework for them and nearly got away with it.
On the one hand, my general sense has been that no one really ever uses generative AI for anything important, and that tech companies have been essentially throwing billions of dollars at a thing that no one besides its promoters has ever been much interested in. Here, we actually managed to find a use-case in the wild!
Of course, generative AI failed miserably. Asked to complete the seemingly simple task of finding negative quotes from famous critics about classic movies it confidently just made some shit up.
My main conclusion here is that generative AI is terrible at producing factual, actionable information, but pretty good at coming up with believable-sounding lies. Maybe it’s best we stop imagining it as some kind of Albert Einstein standing in a magical library, and start imagining it as Sam Rockwell in a wrinkly suit with a three-day old beard sitting behind a desk in an upstairs office in a strip mall somewhere. “You need some movie quotes? Heh, I gotcha covered, pal.”
Chat GPT is a Ben Mendolsohn character.
The best explanation as to how/why ChatGPT does things like this that I’ve found is in a paper called “ChatGPT is Bullshit;” it helps that it is written by philosophers and not tech autists:
https://link.springer.com/article/10.1007/s10676-024-09775-5
The basic gist is that the entire point of ChatGPT isn’t to do tasks or provide information, but rather to fake being human.
AI in the entertainment industry is so noxious that we let a whole piece about finding contrarian critics go and not one mention of Armond White? The future is dumb