McNeal and Robert Downey Jr. Dance With ChatGPT

Robert Downey Jr. and Brittany Bellizeare in McNeal, at the Vivian Beaumont. Photo: Matthew Murphy and Evan Zimmerman

Here’s an AI prompt: Write me a vehicle for a movie star intent on making a “serious” debut on Broadway. Let’s say he’s a veteran of a decade-plus of superhero movies, so we want a character akin to his persona there and a subject that comes with some contemporary relevance. Maybe because he played a tech genius onscreen, we have him wrestle with the cutting edge of technology onstage. He’s also acclaimed as a dramatic actor, meaning we need to give weight to this, so let’s throw in a few hefty themes: addiction, suicide, adultery, family trauma, and, for that genuine “great-man problems” zing, a pinch of misogyny.

If you plugged that prompt into a large language model like ChatGPT, it might spit out something a little like McNeal, though you could spare the climate a few glugs of processing power and simply imagine the thing yourself. Either way, the end product, written by Ayad Akhtar and directed by Lincoln Center Theater’s go-to, Bartlett Sher, doesn’t stray far from what is algorithmically plausible based on past inputs. Robert Downey Jr., on shore leave between being Iron Man, winning an Oscar for Oppenheimer, and then returning to the CGI fold to play Doctor Doom, is in New York to star as an acclaimed writer named Jacob McNeal or, rather, to stand tweedily in the the center of the vast Vivian Beaumont stage, just below a projection of a giant iPhone screen (the sci-fi scenic design is by Michael Yeargan and Jake Barton, and Barton did the projections). As we meet McNeal, he is in the midst of a liver check-up where his doctor (Ruthie Ann Miles, underutilized here but always nice to see) is telling him to stop drinking; McNeal pays her little attention and frets about getting a call. The iPhone indicates that the date is early October, a time of year he dreads because he keeps not winning a Nobel Prize.

The immense phone screen is an early tell of the other angle McNeal is working: If the events of the play are plausibly algorithmically generated, perhaps they are meant to seem so. Before we meet McNeal, we hear him typing on that phone, asking a chatbot who will win the Nobel, the conversation projected onstage. When McNeal does win — this is no spoiler; the announcement happens within the first 15 minutes — he admits he had a bot draft his acceptance speech but says he didn’t like it. When his agent (Andrea Martin, bouncing around the stage trying to generate enough comedic energy to escape this play’s container) suggests he sign a rider on his contract for his next novel to affirm that he used no LLM, he balks. Has McNeal been using these programs to write his books? Akhtar wants us to ask. And should we care? Since they do, Sher and Akhtar bring to mind narratively and aesthetically the unreal tang of AI-generated content. (Though I think they should have gone the extra mile and outfitted an actor with a spare sixth finger.)

Before we get to the big questions of morality and contract law, there’s that central performance you’re here to see: McNeal turns out, unsurprisingly, to be an analogue to Tony Stark, still cracking wise, now transposed from arms trader to great American author (with a Jewish mother and a Catholic father and raised in Texas, so basically Roth/DeLillo/McCarthy in a blender). He has won acclaim for fiction about Reagan and Goldwater. He remains prickly about his legacy, his reputation for not understanding women, and the damage he has done to his relationships in the name of his art. Downey carries off McNeal’s charm well — courting applause when McNeal delivers showboating, professorial tirades about Shakespeare and Saul Bellow — but struggles to downshift into subtlety. An extended scene between McNeal and his son (Rafi Gavron, lost at sea) that pivots around their diverging memories of McNeal’s dead wife takes place in a eerily empty re-creation of a country lodge, with limp blocking and half-hearted references to Chekhov. You better believe someone’s polishing a gun. Sher has Downey and Gavron wander in odd circuits around the living area as they talk, while the ghostly figure of McNeal’s wife walks backstage past their windows. It’s all in the shape of a serious drama, yet it has no grip, no immediacy.

Akhtar is a synthesizing playwright interested in picking apart vast systems, whether America’s relationship to Islam or the financial markets, but his writing about AI disappointingly doesn’t untangle much, instead falling into familiar tropes and arguments. On the advent of the technology, Akhtar took the stance in an interview with The Atlantic, “I’m also certainly not making a claim about whether it’s good or bad. I just want to understand it, because it’s coming.” So as much as McNeal grouses about how AI has produced tons of word slop, the play itself tilts more in the direction of what Akhtar imagines are the technology’s possibilities: He compares the work of a language-learning model to Shakespeare, who himself downloaded the input of banal Elizabethan dramas he would have known, such as the tale of “King Leir,” and output one of the greatest tragedies ever written. The idea may be, to garble Anna Karenina in a botlike way, that all plagiarisms are alike — but that obfuscates the idiosyncrasies of art and, indeed, of technology. Art is stealing, sure, but some of it is transformative, intentional, and personal, while other thefts are procedural, for the purpose of idle amusement and shareholder value. I don’t think an LLM can do the former, and it’s silly to take the exaggerated, apocalyptic marketing language of Sam Altman types at face value in the way McNeal tends to. An LLM’s cognition is simply different from a human’s. We don’t have to pretend those two processes approach the same ends — and there’s plenty of material in looking at the ways a brain and a language model run differently. For her play Prometheus Firebringer, Annie Dorsen had a computer program complete a lost Greek tragedy while she painstakingly assembled a speech about her work made up entirely of quotation from other works. On film, The Beast, a very French genre-hopping hybrid of David Lynch and Henry James, traps characters in the not-quite-human scripts generated for them and runs, thrillingly, on the friction between what a machine may think a person wants and the sometimes slightly different tangle that constitutes their actual emotions and impulses. And buried in a corner of streaming television, there is Mrs. Davis, which imagines a world run by a godlike AI program that thought it was doing good but operated with unhuman shortcuts (the joke of that show was also that the program was originally intended to sell Buffalo Wild Wings). There’s also much to analyze about the socioeconomic conditions that might encourage the proliferation of AI-generated word slop, and it was disappointing that Akhtar, who elsewhere has looked more closely at the undersides of systems, avoids in McNeal the thornier downsides of LLMs: rampant energy use, violation of copyright, a corporation’s desire to cut labor costs. I don’t think it’s likely that LLMs will start creating art as good as great human opuses; it does seem more plausible that the forces of capital will pressure us not to care about the difference. Slap a recognizable name in the center of something ill formed and people will still pay top dollar for it, speaking from hypothetical experience.

McNeal’s woozy ruminations about art and technology might strike with more force if the actual drama around them had more tensile strength. The human dynamics Akhtar and Sher hang all of this on never get past cliché: McNeal confronts a cadre of women, including Martin’s assistant (Saisha Talwar), a horrifyingly underprepared magazine journalist (Brittany Bellizeare), and a former New York Times books editor who is pretty obviously based on Pamela Paul (Melora Hardin). In his exchanges with them, they get to do little except absorb his rants about everything from the work of Annie Ernaux to Harvey Weinstein. If McNeal wants to tell us that a great artist — specifically a great man, in this case — is some unique force in the universe, it requires a more finely crafted rendering of that man. But on the opposite end, if the play wants to suggest AI could offer an alternative to that figure that’s equal in stature? Well, there’s an issue of rendering there, too. Later on in the production, we see a giant Wizard of Oz–style projection of a “Jacob McNeal Digital Composite” that mimics Downey’s face, leaving it gooey and claylike — contra the script, which suggests the deepfake should appear “as if the computer image was delivering a level of emotion we would not imagine it capable of doing.” The deepfake is credited to AGBO, a studio founded by Avengers directors the Russo brothers, and it seems a lot like a product demonstration stuck in the middle of a play.

Then, right at the end, Akhtar has AI step into the role of Shakespeare himself. Downey ends the play with a farewell speech in imitation of Prospero’s from The Tempest, which Akhtar says he constructed after spending a lot of time probing a chatbot with just the right inputs. That in itself sounds like creative work, but to Akhtar, there’s something going on: “I wanted the final speech to have a quality of magic to it that resembles the kind of amazement … I have sometimes felt when I see the language being generated. I want the audience to have that experience.” When that speech was projected onscreen at the performance I attended, there were indeed oohs and aahs at the bot’s sort-of eloquence. It had figured out some nifty rhymes and approximated a Shakespearean poetic register but, in my very human critical opinion, without any surprising imagery or wit. There’s no there there: This is just gunk.

McNeal is at Lincoln Center Theater.