Ayad Akhtar and Robert Downey Jr. Confront AI

Ayad Akhtar’s good new play, McNeal, presently on the Lincoln Middle Theater, is transfixing partially as a result of it tracks with out flinching the disintegration of a celebrated author, and partially as a result of Akhtar goes to a spot that few writers have visited so successfully—the very close to future, by which massive language fashions threaten to undo our self-satisfied understanding of creativity, plagiarism, and originality. And in addition as a result of Robert Downey Jr., performing onstage for the primary time in additional than 40 years, completely embodies the genius and brokenness of the title character.

Discover the November 2024 Concern

Take a look at extra from this concern and discover your subsequent story to learn.

View Extra

I’ve been in dialog for fairly a while with Akhtar, whose play Disgraced received the Pulitzer Prize in 2013, about synthetic generative intelligence and its influence on cognition and creation. He’s one of many few writers I do know whose place on AI can’t be decreased to the (comprehensible) plea For God’s sake, cease threatening my existence! In McNeal, he not solely means that LLMs is perhaps nondestructive utilities for human writers, but in addition deployed LLMs as he wrote (he’s used lots of them, ChatGPT, Claude, and Gemini included). To my chagrin and astonishment, they appear to have helped him make a fair higher play. As you will notice in our dialog, he doesn’t consider that this needs to be controversial.

In early September, Akhtar, Downey, Bartlett Sher—the Tony Award winner who directed McNeal—and I met at Downey’s house in New York for what turned out to be an amusing, sometimes frenetic, and typically even borderline profound dialogue of the play, its origins, the flummoxing points it raises, and, sure, Avengers: Age of Ultron. (Oppenheimer, for which Downey received an Academy Award, additionally got here up.) We have been joined intermittently by Susan Downey, Robert’s spouse (and producing companion), and the one who believed that Akhtar’s play would tempt her husband to return to the stage. The dialog that follows is a condensed and edited model of our sprawling dialogue, however I feel it captures one thing about artwork and AI, and it actually captures the distinctive qualities of three folks, author, director, and actor, who’re working on the pinnacle of their commerce, with out worry—maybe with out sufficient worry—of what’s inescapably coming.


Jeffrey Goldberg: Did you write a play a couple of author within the age of AI since you’re attempting to determine what your future is perhaps?

Ayad Akhtar: We’ve been residing in a regime of automated cognition, digital cognition, for a decade and a half. With AI, we’re now seeing a late downstream impact of that, and we expect it’s one thing new, however it’s not. Know-how has been remodeling us now for fairly a while. It’s remodeling our neurochemistry. It’s remodeling our societies, you recognize, and it’s making our emotionality throughout the social house completely different as properly. It’s making us much less able to being bored, much less keen to be bored, extra keen to be distracted, much less desirous about studying.

Within the midst of all this, what does it imply to be a author attempting to put in writing in the best way that I wish to write? What would the brand new applied sciences imply for writers like Saul Bellow or Philip Roth, who I am keen on, and for the richness of their language?

Goldberg: Each of them inform the character of McNeal.

Akhtar: There are a lot of writers inside McNeal—older writers of a sure technology whose work speaks to what’s everlasting in us as people, however who possibly don’t communicate as a lot to what’s altering round us. I used to be really considering of Wallace Stevens within the age of AI sooner or later—“The Auroras of Autumn.” That poem is about Stevens eyeing the top of his life by the dazzling, otherworldly gentle of the northern lights. It’s a poem of extraordinary magnificence. On this play, that dazzling show of pure marvel is definitely AI. It’s now not the chic of nature.

Goldberg: Had been you picturing Robert as you wrote this character?

Akhtar: I write to a really perfect; it’s not essentially an individual.

Robert Downey Jr.: I really feel that me and splendid are synonymous.

Akhtar: Robert’s embodiment of McNeal is in some methods a lot richer than what I wrote.

Downey: I’ve a extremely heavy, heavy aller­gy to paper. I’m allergic to issues written on paper.

Akhtar: As I’ve found!

Downey: However the writing was transcendent. The final time that occurred, I used to be studying Oppenheimer.

Goldberg: There’s Oppenheimer on this, however there’s additionally Age of Ultron, proper?

Downey: Truly, I used to be fascinated by that whereas I used to be studying this. And I’ll catch you guys up within the combination. I’m solely ever doing two issues: Both I’m attempting to keep away from threats or I’m looking for alternatives. This one is the latter. And I used to be considering, Why would I be studying this? As a result of, I imply, I’ve been a little bit of an oddball, and I used to be considering, Why is that this occurring to me; why is that this play with me? And I’m having this response, and it took me proper again to Paul Bettany.

So that you just guys perceive what’s occurring, that is the second Avengers movie, Age of Ultron, and Bettany was taking part in this AI, my private butler. The butler had gone by these iterations, and [the writer and director] Joss Whedon determined, “Let’s have you ever turn out to be a sentient being, a sentient being that’s created from AI.” So first Bettany is the voice, after which he turned this purple creature. After which there was at the present time when Bettany needed to do a form of soliloquy that Joss had written for him, as we’re all launched to him, questioning, Is he a risk? Can we belief him? Is he going to destroy us? And there comes this second once we notice that he’s simply looking for to know, and be understood. And this was the second in the course of this style movie once we all stopped and thought, Wait, I feel we would really be speaking about one thing vital.

Goldberg: Bart, what are you exploring right here?

Bartlett Sher: I’m principally exploring the deep tragedy of the lifetime of Jacob McNeal. That’s the central concern. AI and all the pieces round it, these are supply techniques to that exploration.

Akhtar: Robert has this excellent second within the play, the best way he does it, by which he’s arguing for artwork on this very sophisticated dialog with a former lover. And it will get to one of many essences of the play, which is that that is an try to defend artwork even when it’s made by an indefensible particular person. As a result of in the long run, human creation continues to be superior, and none of us is ideal. So the bigger dialog round who will get to put in writing, the morality of writing, all of that? In a approach, it’s form of rising from that.

Goldberg: I can’t say for certain, however I feel that is the primary play that’s concurrently about AI and #MeToo.

Downey: And id and intergenerational battle and cancel tradition and misunderstanding and sub­intentional contempt and unconscious bias.

Goldberg: Are there any third rails you don’t contact?

Akhtar: McNeal is the third rail. He’s a imaginative and prescient of the artist in oppo­sition to society. Not a flatterer of the present values, however somebody who questions them: “That’s a lie. That’s not true.”

Goldberg: The timing is superb.

Downey: In films, you at all times miss the second, or you’re preempted by one thing. With Oppenheimer, we occurred to be popping out proper across the time of sure different world occasions, however we couldn’t have identified. With this, we are actually first to market. Theater is the shortest distance between two factors. You could have one thing pressing to say, and also you don’t dawdle, and you’ve got an area like Lincoln Middle that isn’t within the backside line, however within the kind. And you’ve got Ayad inspiring Bart, and then you definately get me, the bronze medalist. However I’m tremendous fucking motivated, as a result of I by no means get this sense of immediacy and emergence occurring in actual time.

Goldberg: Let’s discuss for a minute in regards to the AI artistic apocalypse, or if it’s a artistic apocalypse in any respect. I prompted Claude to put in writing a play identical to McNeal, with the identical plot turns and characters as your play, and I requested it to put in writing it in your fashion. What emerged was a play known as The Plagiarist’s Lament. I went backwards and forwards with Claude for some time, primarily to attempt to get one thing much less hackish. However in the long run, I failed. What got here out was one thing like an Ayad play, besides it was dangerous, not good.

Akhtar: However right here’s the factor. You’re simply utilizing an off-the-shelf product, not leading-­edge story know-how that’s now changing into more and more frequent in sure circles.

Goldberg: So don’t fear about in the present day, however tomorrow?

Akhtar: The know-how’s shifting shortly, so it’s a actuality. And worrying? I’m not attempting to foretell the longer term. And I’m additionally actually not making a declare about whether or not it’s good or dangerous. I simply wish to perceive it, as a result of it’s coming.

Downey: To borrow from current expertise, I feel we could also be at a post-Trinity, pre-Hiroshima, pre-Nagasaki second, although some folks would say that we’re simply at Hiroshima.

Goldberg: Hiroshima being the primary real-world use of ChatGPT?

Downey: Trinity confirmed us that the bomb was purpose-built, and Hiroshima was displaying us that the aim was, probably, not completely obligatory, however that it additionally didn’t matter, as a result of, traditionally, it had already occurred.

Goldberg: Proper now, I’m assuming that a part of the issue I had with the LLM was that I used to be giving it dangerous prompts.

Downey: One concern is that LLMs don’t get bored. We’ll be working one thing and Bart will go, “I’ve seen this earlier than. I’ve performed this earlier than.” After which he says, “How can I make this new?”

The individuals who transfer tradition ahead are often the high-ADD people that we’ve tended to assume both must be medicated or all go into one line of labor. They’ve a low threshold for boredom. And since they’ve this low threshold, they are saying, “I don’t wish to do that. Do one thing completely different.” And it’s virtually simply to maintain themselves awake. However what an incredible reward for creativity.

Goldberg: The three of you symbolize the performing facet, and directing, and writing. Who’s in probably the most existential hazard right here from AI?

Downey: Anybody however me.

Akhtar: The Display screen Actors Guild has handled the image-likeness concern in a that means­ful approach.

Downey: We’ve made probably the most noise—­we, SAG—­and we’re probably the most dramatic about all the pieces. I keep in mind after I was doing Chaplin, the discuss was about how important the top of the silent period was.

Goldberg: Is that this the identical stage of disruption?

Downey: I doubt it, however not as a result of Claude can’t presently pin his ass with each fingers. There are variations which can be going to be considerably extra superior. However applied sciences that folks have argued would impede artwork and tradition have typically assisted and enhanced. So is that this time completely different? That’s what we’re at all times worrying about. I reside in California, at all times questioning, Is that little rumble within the kitchen, is that this the massive one?

Sher: For me, I feel directing could be very plastic. It requires integrating quite a lot of completely different ranges of exercise. So really discovering a technique to course of that into a pc’s considering, and really having it work in three dimensions when it comes to organizing and creating, appears very tough to me. And I basically do the work of the interpreter and synthesizer.

A machine can let you know what to do, however it might probably’t work together and join and pull collectively the completely different strands.

Akhtar: There’s a management dimension to what Bart does. I imply, you wouldn’t need a pc doing that.

Sher: This might sound geeky, however what’s the distinguishing high quality of creating artwork? It’s to take part in one thing uniquely human, one thing that may’t be performed another approach.

So if the Greeks are gathering on the hillside as a result of they’re constructing an area the place they will hear their tales and take part in them, that’s a uniquely human expertise.

Akhtar: I do assume that there’s something irreducibly human in regards to the theater, and that in all probability over time, it’ll proceed to show its worth in a world the place virtuality is more and more the norm. The financial drawback for the theater has been that it occurs solely right here and solely now. So it’s at all times been arduous to monetize.

Goldberg: However I’ve two phrases for you: ABBA Voyage. I imply, it’s an extraor­dinarily in style present that makes use of CGI and movement seize to give the expertise of liveness with out ABBA really being there. Not exactly theater, however it’s scalable, seemingly reside know-how.

Downey: Surprisingly, that is the true trifecta: IP, know-how, and style. I consider this model of music—which, you recognize, it’s not my bag, however I nonetheless actually admired that someone was captivated with that after which purpose-built the venue. After which they mentioned, “We’re not going to go for ‘Oh my God, that appears so actual.’ We’re really going to go for a extra two-dimensional impact that’s rendered in a approach by which the viewers can full it themselves.”

Akhtar: ABBA Voyage is an exception. Nevertheless it’s nonetheless not reside theater.

Sher: It’s additionally not doable with out the ABBA expertise that preceded it. It’s an augmentation; it’s not unique.

Goldberg: When it comes to writing, Ayad, I did what you steered I do and requested Claude to critique its personal writing, and it was really fairly good at that. I felt like I used to be really speaking with somebody. We have been in a dialogue about pacing, readability, phrase selection.

Sher: Nevertheless it has no instinct in any respect, no instinct for Ayad’s mindset in the course of this exercise, and no understanding of how he’s seeing it.

Downey: It does have context, and context is crucial. I feel it’s going to begin shortly modeling all of these issues that we maintain expensive as ­subtleties which can be un­assailable. It’s going to see what’s lacking in its sequence, and it’s going to focus all of its cloud-bursting vitality on that.

Goldberg: It is perhaps the producers or the studios who’re in bother, as a result of the notes are delivered sequentially, logically, and with out defensiveness. Do you assume that these applied sciences may give higher notes than the common government?

Akhtar: I do know producers in Hollywood who’re already utilizing these instruments for his or her writers. They usually’re utilizing them empirically, saying, “That is what I feel. Let’s see what the AI thinks.” And it seems that the AI is definitely fairly good at understanding sure varieties. For those who’ve received a corpus of texts—like, say, Regulation & Order ; you’ve received many, many seasons of that, otherwise you’ve received many seasons of a youngsters’s present—these are codified varieties. And the AI, if it has all these texts, can perceive how phrases are formed in that kind.

Goldberg: So you can add a thousand Regulation & Order scripts and Claude might give you the thousandth and first.

Akhtar: A couple of 12 months and a half in the past, after I began taking part in with ChatGPT, the very first thing that I began to see have been processes of language that jogged my memory of studying Shakespeare. No author is best at presenting context than Shakespeare. What I imply by that’s Shakespeare units all the pieces shortly in movement. It’s virtually like a chess sport—you’ve received items, and also you wish to get them out as shortly as doable so you have got choices. Shakespeare units the choices out shortly and begins creating variations. So there’s a sequence of phrases or linguistic tropes for each single play, each poem cycle, each sonnet. All of them have their universe of linguistic context that’s being deployed and redeployed and redeployed. And it’s in that play of language that you just discover an accretion of that means. It was not fairly as thrilling to see the chatbot do it, however it was really very attention-grabbing to acknowledge the identical course of.

photo of bald man in glasses wearing black t-shirt, tan blazer, and jeans standing with one hand in pocket against brown background
OK McCausland for The Atlantic

Goldberg: Shakespeare was his personal AI.

Downey: As a result of he carried out as a youthful man, it was all uploaded into Shakespeare’s system. So he was so conversant in the template, and he had all this expertise. And equally, all of those LLMs are on this stage the place they’re simply starting to be taken critically. It’s like we’re pre–bar mitzvah, however these are sharp youngsters.

Goldberg: Would you employ ChatGPT to put in writing a complete piece?

Sher: Quickly we’ll be having conversations about whether or not Claude is a greater artist than ChatGPT. Might you think about folks saying, “Effectively, I’m not going to see that play, as a result of it was written by this machine; I wish to see this one, as a result of it’s written by Gemini as an alternative.”

Goldberg: Sadly, I can simply think about it.

Akhtar: I’m unsure that I’d use an LLM to put in writing a play, as a result of they’re simply not superb at doing that but, as you found in your personal play by Claude. I don’t assume they’re adequate to be making the varieties of choices that go into making a murals.

Goldberg: However you’re instructing the software the best way to get higher.

Akhtar: So what? They’ve already gone to highschool on my physique of labor.

Goldberg: So what? So what? 600 years of Gutenberg, and the printing press by no means made choices by itself.

Akhtar: However we’re already inside this regime the place energy and monetized scale exist throughout the fingers of only a few. We’re doing it daily with our telephones; you’re instructing the machine all the pieces about you and your loved ones and your needs. That is the paradigm for the twenty first century. All human exercise is passing by the fingers of only a few folks and quite a lot of machines.

Goldberg: McNeal is about lack of management.

Akhtar: It’s. I’m simply making the purpose that we’re not likely in a distinct regime of energy with AI. It could be much more concentrated and much more consequential, however on the finish of the day, to take part within the public house within the twenty first century is to take part on this construction. That’s simply what it’s. We don’t have another, as a result of our authorities has not regulated this.

Goldberg: You see the LLM as a collaborator in some methods. The place will the purple line be for writers, between collaboration and plagiarism?

Akhtar: From my perspective, there are any variety of artists we might take a look at, however the one which I’d in all probability at all times spend probably the most time is Shakespeare, and it’s powerful to say that he wasn’t copying. As McNeal explains at one level within the play, King Lear shares 70 p.c of its phrases with a earlier play known as King Leir, which Shakespeare knew properly and used to put in writing Lear. And it’s not simply Leir. There’s that nice scene in Lear the place Gloucester is led to this plain and advised it’s a cliff over which he’s going to leap, and that subplot is taken proper out of Sir Philip Sidney. It could replicate deeper processes of cognition. It could replicate, as Bart has mentioned, how we imitate in an effort to study. All of that’s simply a part of what we do. When that will get married to a corporate-ownership mannequin, that may be a separate concern, one thing that should get labored out over time, social­ly and legally. Or not, if our legislators don’t have the need to take action.

Goldberg: The ultimate soliloquy of the play—no spoilers right here—is augmented by AI.

Akhtar: This has actually been an interesting collaboration. As a result of I needed some a part of the play to really be meaningfully generated by ChatGPT or some massive language mannequin—Gemini, Claude. I attempted all of them. And I needed to do it as a result of it was a part of what the play was about. However the LLMs had a troublesome time really delivering the products till this week. I’ve lastly had some experiences now, after many months of working with them, which can be bearing fruit.

I needed the ultimate speech to have a top quality of magic to it that resembles the form of amazement that I knew you had felt working with the mannequin, and that I’ve typically felt after I see the language being generated. I need the viewers to have that have.

Sher: , I feel the issue you have been going through might have been with any of your collaborators. We simply had this new collaborator to assist with that second.

Goldberg: You’re blowing my thoughts.

Akhtar: It’s not likely that controversial.

Goldberg: Sure it’s. It’s completely controversial.

Downey: Effectively, let’s discover out!

Goldberg: It’s extra of a leap than you guys assume.

Akhtar: It’s a play about AI. It stands to motive that I used to be ready, over the course of many months, to lastly get the AI to offer me one thing that I might use within the play.

Downey: what the leap was like? A colicky little child lastly gave us a giant ol’ burp.

Akhtar: That’s precisely proper. That’s what occurred. Quite a lot of unsatisfying work, after which, unprompted, it lastly got here up with an excellent ultimate couplet! And that’s what I’m utilizing for the top of the play’s ultimate speech.

Goldberg: Wonderful, and threatening.

Sher: I simply can’t think about a world by which ChatGPT might take all expertise and unify it with Ayad’s curiosity in magnificence and that means and his obsession with classical tragedy and pull all these forces along with emotion and feeling. As a result of irrespective of what number of instances you prompted it, you’re nonetheless going to get The Pestilential Plagiarist, or no matter it’s known as.

Downey: The rationale that we’re all sitting right here proper now could be as a result of this motherfucker, Ayad, is so searingly refined, but in addition occasionally—greater than sometimes—scorching beneath the collar. My new favourite cable channel known as Ayad Has Fucking Had It. He’s like probably the most collaborative superintelligence you’ll ever come throughout, and subsequently he’s letting all this slack out to everybody round him, however every now and then, if this intelligence is completely unappreciated for hours or days at a time, he’ll flare. He’ll simply remind us that he can break the sound barrier if he desires to. And I get chills from that. And that’s why we’re right here. It’s the human factor.

Akhtar: It’s not new for people to make use of instruments.

Sher: Are we going to be required to add a system of ethics into the machines as they get an increasing number of highly effective?

Downey: Too late.

Goldberg: That’s what they promise in Silicon Valley, alignment with human values.

Downey: Two years in the past was the time to do one thing.

Akhtar: You guys are considering massive. However I simply don’t know the way that is going to play out. I don’t know what it’s. I’m simply desirous about what I’m experiencing now and in working with the know-how. What’s the expertise I’m having now?

Goldberg: There’s a distinction between a human hack and a very good human author. The human hack doesn’t know that they’re dangerous.

Downey: This can be a harebrained rabbit gap the place we might consistently maintain considering of an increasing number of ramifications. One other concern right here is that sure nice artists do one thing that most individuals would labor a life-time or profession to come back near, and the second they’re performed with it, they’ve contempt for it, as a result of they go, “Eh, that’s not my finest.”

Akhtar: I acknowledge somebody in that.

Downey: All I’m saying is that I simply need the sensation of these sparks flying, that new neural pathway being pressured. I wish to push the bounds. It’s that entire factor of pushing limits. Once I really feel good, after I can inform Bart is kicking me, when Ayad is simply lighting up, and after I’m realizing that I simply received a notice that revolutionized the best way I’m going to attempt to painting one thing, you go, “Ooh!” And even when it’s previous information to another person, for me, it’s revolutionary.

Akhtar: One other approach of placing this, what Robert is saying, is that what he’s engaged in will not be problem-solving, per se. It’s not that there’s an recognized drawback that he’s attempting to unravel. That is how a pc is commonly considering, with a gamification type of mindset. For Robert, there’s a richness of the current for him as he’s working that’s figuring out prospects, not issues.

Sher: I’ve thought lots about this, attempting to know the problem of GPT and creativity, and I’m lots much less anxious now, as a result of I really feel that the depth of the inventive course of within the theater isn’t replicable.

The amalgam of human expertise and emotion and feeling that passes by artists is uniquely human and never capturable. Phrase orders might be taken from all types of sources. They are often imitated; they are often replicated; they are often reproduced in numerous methods. However the important exercise of what we do right here on this approach, and what we construct, has by no means been safer.

Downey: And if our job is to carry the mirror as much as nature, that is now a part of nature. It’s now a part of the firmament. Nature is now inclusive of this. We’re onstage and we’re reflecting this again to you. What do you see? Do you see your self inside this image?


This text seems within the November 2024 print version with the headline “The Playwright within the Age of AI.”