Let's get the jokes out of the way first: "I, for one, welcome our new AI overlords and look forward to fifty-seven Matrix sequels where [PLOT TWIST] the robots are now the good guys!"
But, seriously, the news that Warner Brothers is adopting a new AI system to help it greenlight new films is as scary as it is dumb as it is banal. Amazingly, it is both the biggest, most frightening change to the assembly line nature of filmmaking many studios engage in and also the most inevitable — and really unlikely to change all that much. Which is exactly the problem.
Sadly, this isn't a plot from Futurama, which predicted this specifically in the classic "Bender Should Not Be Allowed on Television," where we meet the Network Execubots who run television. They are programmed to like what they've seen before, roll dice to determine the fall schedule, and underestimate Middle America. ("It's funny, but will it get them off their tractors?") Sadly, this satire is becoming reality, but mostly in that this how executive already behave when approaching creative decisions, so literally outsourcing it to an AI seems like an inevitable next step. But what does that really mean? And how soon will we have computers greenlighting pitches?
I: What is AI?
Part of the confusion here is a misunderstanding of exactly what AI is in this situation. It is, essentially, a smart spreadsheet. Warner Bros joins STX and other production companies using a program called Cinelytic, which, essentially, crunches numbers on how much they think a certain star's value will add in ticket sales in markets across the globe. According to Cinelytic founder Tobias Queisser, "it will reduce the amount of time executives spend on low-value, repetitive tasks and instead give them better dollar-figure parameters for packaging, marketing and distribution decisions, including release dates."
"The system can calculate in seconds what used to take days to assess by a human when it comes to general film package evaluation or a star's worth," says Queisser. . . "Artificial intelligence sounds scary. But right now, an AI cannot make any creative decisions. What it is good at is crunching numbers and breaking down huge data sets and showing patterns that would not be visible to humans. But for creative decision-making, you still need experience and gut instinct."
No doubt, it will save time from number crunching. But what it promises to do is take all worldwide box office data, look at how much specific films starring specific actors made in specific markets, and then predict how some other film, all other things being equal, will fare at the box office. If $X to hire said actor is less than $Y of box office receipts, then "We got ourselves a picture!" If not, then back to the drawing board.
Yikes. But here's the dirty secret:
II: Hollywood is Already Doing This
This is the sad truth we have to face. Too many decisions are made already by guys with spreadsheets who are calculating whether Cynthia Erivo is a box office draw or not. This is how Emma Stone gets cast as a Chinese-Hawaiian woman in Aloha. It's most certainly how we ended up with Johnny Depp in the Fantastic Beasts films.
AI is just going to make it faster. And that's part of the problem.
Because the mind-numbing, soul-sucking act of running a bunch of calculations in a spreadsheet to determine who the best person is to lead a film is in itself dehumanizing and awful. The fact that it really, really sucks to do that work should wake people up to the idea that maybe, just maybe, that isn't how we should do things.
I'm not so naive as to believe that somehow Hollywood will, anytime soon, put art over commerce. (Yay, late stage capitalism and consolidation of corporate media!) But it would be nice if it weren't so blatant that now we're just letting computers make the decisions for us and that the value of any human being can be summed up in how much money they're likely to make for them next quarter.
But outsourcing that menial data task to computers, who never met a person they couldn't dehumanize and instead dispassionately look at as only a conglomeration of aggregated data, we double down on the idea that, yes, actually, this is the best way to make creative decisions.
It will also, no doubt, be nice when it comes time for accountability, when the franchise film falters despite a mix of "bankable" stars, executives can merely blame the dataset. You fire one big data vendor and hire another.
III: The Fallacy of Predictive Analytics in Hollywood Hitmaking
The problem with all computer systems and data sets is the adage of "garbage in, garbage out." And it's time that we faced facts: all this data is bullshit.
The first fallacy is the same warning you hear about literally any financial investment: "past performance is no guarantee of future results." This is literally Business School 101 stuff, but the folks making some of the most consequential decisions about our movies seem to think that because Brad Pitt's last two films made X in Malaysia then that tells us that he'll likely make Y. That's just silly.
Second, it's time to reconcile the idea that past box office performance data is skewed heavily towards white men and perpetuates racism and sexism in casting and what types of films get made. Back to Garbage In, Garbage Out, one of the main problems with AI and predictive analytics are the inherent biases that get put into any data set. "But it's just data," you say. Yes, but which data points get collected and why and how the data is supposed to be interpreted ultimately get reflected in the data output.
When we assume that data, AI and predictive analytics will be dispassionate, we actually see the exact opposite in the real world. From information bubbles perpetuated by social media AI algorithms to how those bubbles are exploited by savvy practitioners and pols, we see the effects in everything from Brexit to Trump to Snyder Cut Bronies. Because the old Hollywood system valued leading (white) men, paid men more than their female costars, refused to put women and non-whites in lead roles, we have decades of data that reinforce those very specific biases. Therefore, any AI drawing inference from that data will perpetuate those biases ad nauseam. Congratulations: you taught computers to be racist and sexist.
And third, our most recent box office trends show anything but. The landscape is changing quickly. Our top films star Chadwick Boseman, Brie Larson, John Boyega, Daisy Ridley, and Oscar Isaac. And yes, they also star Robert Downey Jr, who, a decade and a half ago was uninsurable as an actor, and Chris Evans, who (same timeframe) had recently been in several major studio flops, and Chris Pratt, who was the third or fourth male lead on an NBC single camera sitcom. NONE of these people would've been spreadsheet-bankable. No AI would ever greenlight a movie with these people in them. And yet. . .
I challenge anyone to name a critically or commercially successful film from the last decade (sans a franchise sequel with already known, bankable stars) that an AI would have greenlit. I think Knives Out could pass that test, or Once Upon a Time… In Hollywood, making them, arguably, the exceptions that prove the rule. But Tarantino and Rian Johnson don't need AI to help them greenlight their next films: Greta Gerwig and Lulu Wang do. And it's in those marginal cases, especially with up and coming filmmakers, where it's most important.
I fear for the casting fights Ava DuVernay will have to go through to get who she wants to deliver the New Gods film we all deserve. And now she'll have to fight a damn robot powered by decades of racism and sexism to do it?
Since you asked so nicely. Working on the script with @TomKingTK and dreaming of actors. Seeing most of the incredible fan casting that's going on. May in fact be taking some notes. xo https://t.co/4aIPXrmjHs pic.twitter.com/NtybLesLd6
— Ava DuVernay (@ava) January 8, 2020
In some ways, that couldn't be any more 2020.
Let's hope Birds of Prey, Wonder Woman '84 and The Suicide Squad are all good — and crazy successful enough to break Cinelytic's data sets, perhaps teaching it to say "Diversity and fresh faces will help sell your movies." And may Warner Bros decide to look up from the datasets and spreadsheets and say "Let's be better than this."
But if not, bring on The Matrix sequels, I guess.