Lexus Used A.I. to Make the 'Perfect' Car Ad. Here's How It Turned Out
The bots continue to struggle with intuition
There exist all kinds of experiments in trying to get artificial intelligence to do our creative heavy lifting: A.I. paint names. A.I. journalism. A.I. literature. A Beatles-style A.I. song (and also a Christmas carol!). There’s even an A.I. short film featuring Thomas Middleditch, and metrics for gauging whether a script was written by an artificial mind.
Whether these are proof of A.I.’s impending colonization over creative giggery is debatable; many of these instances tell us artificial productions are hilariously experimental at best (something Burger King poked fun at in recent ads), and creepy at worst. All get some kind of human assistance.
Even so, Lexus is doubling down on the idea that A.I. may well be the future of cutting out bloated agency costs (tomorrow, if not today). To promote its new ES Sedan, it gives us “Driven by Intuition,” scripted entirely by IBM’s Watson with help from The&Partnership London and Visual Voice.
The eerie little work, directed by “award-winning human” Kevin Macdonald (The Last King of Scotland; Whitney) opens with a Takumi master craftsman examining his last great oeuvre—a Lexus ES. The car is taken away and threatened with destruction for reasons unclear (was it just too beautiful?), but automatic emergency braking kicks in just in time to save itself, and its maker’s fragile heart.
The subtext is that the car is so intuitive that you don’t have to worry when you drive it; it wants to survive just as much as you do. It’s got super-responsive kit, reinforced by the new Global Architecture-K platform and Lexus Safety System+ functions, built to help navigate volatile road and traffic conditions. Automatic braking, steering and driver alerts can prevent an accident or reduce one’s consequences.
If you’ve driven a car made in the last five years, you’re probably used to similar functionalities. The difference lies in how they’re marketed, and there’s a lot to be said about the unimaginative, slow-moving character of “car porn.”
To break that mold, Watson was outfitted with all kinds of data from past ads and what people liked about them, the fruit of a study commissioned with MindX, the applied science division of the University of South Wales. Here’s what the team learned about what makes “the perfect car ad”:
• The car doesn’t need to drive at all, unless that’s part of the story
• Driving should be peripheral to the story
• Viewers care more about performance than they do luxury or features
• Characters should be emotionally relatable—think a dad, not a driver or an engineer
• Favor strong facial expressions over strong language
• Twists maintain interest
• Midway through, incorporate an unexpected event—like a crash! (Or a near miss)
Taken separately, this is good advice that can probably be summed up in the following insight: “Car ads suck and are too predictable or engineery. Be a better storyteller.”
But there’s no insight to sum up here. One thing about going all-in on A.I. is that you want to see how the machine decides to value this information. So all these points were tossed into a big data stew, spiced thus: Fearful that Watson would produce something derivative, Lexus added its own brand and project guidelines “to keep the script original and on-brand.”
We can feel all the ad people in the room cringing. This is the machine-learning equivalent of “Make the logo bigger,” except there’s no creative director standing in front of the boiler and shouting, “NO!”
What we receive are vestiges of a story, nipping but never quite cohering. The fatherly quality of the Takumi craftsman comes off disconcertingly strong; he actually cries when his creation lights up on its own and self-pilots out into the brave new world, an eager little Sparky ready to embrace the day.
There’s just enough driving to illustrate some semblance of exploration, but nothing, really, to suggest it’s a threat people would want to lock up—just the standard winding-road shot, offset by an inexplicably sinister stormy sky.
We then see disembodied hands chain it up in a warehouse that contains crumpled, dead vehicles—the machine version of a hostage situation. But the paparazzi are there, too, silently snapping photos as the Takumi master watches it all on television with his actual child.
Tension is mounting! A crash test is imminent!
Yet it never quite arrives. The car is dragged at speed toward the back of a truck (why a truck?). We see it brake, stressing the cables, which aren’t strong enough to resist the sedan’s sense of self-preservation.
And our Takumi dad is so happy! He kisses his daughter, and we cut to the product shot.
It’s quiet, chilly work, futuristic in a past-life, Flight of the Navigator way—being, as it is, informed by old car ads. You can also feel the unrestrained impact of brand guidelines; there are no emotional stakes apart from a murky desire to not see the craftsman cry again. The car will remain untouched, like a hero in an old Tarantino film, and even the atmosphere feels distinctly Lexus—muted and opaque in color, a metallic egg.
In short, it feels like the ultimate triumph of work by committee, nearly a contrast piece to Apple’s “1984.” The ghost of Steve Jobs can be heard in the wind, shouting, “This is exactly why we hurled the hammer!”
We aren’t totally sure what to extract from this. Maybe it’s best to let the experts talk.
“The truth is that the machines are coming. I think that means that we have more magic tricks to play with,” says IPA president Sarah Golding at the outset of the making-of video below.
“The magic of storytelling will always come to life in the human creative process, and using Watson to identify the common attributes for truly award-winning creative work is an example of how man and machine will collaborate in the A.I. era,” says Reece Medway, media and entertainment specialist for IBM Watson U.K. and Ireland.
“Collaborate” is an interesting word. There are a ton of ways in which data can facilitate creativity, bringing to light critical points we may well have missed in human bias. Netflix’s desire to reach men of a certain age group, examine the data and combine their results to create “House of Cards” is a good example of this.
But there’s also something to be said about trusting only data at the expense of intuition: Netflix’s “Marco Polo” was equally data-driven, this time to appeal to the Asian market, and it failed to stick the landing. Data doesn’t really give us an excuse not to get to know one another better, and understand what stakes steer our emotional responses. It doesn’t really replace intuition.
There’s plenty of precedent for how intuition-driven work both succeeds and fails in advertising, so this isn’t about valuing one over the other. This is probably more a question of raising false gods. Technology is a trick, one more tool in the arsenal. But the magic comes from connectivity, which is an almost organic thing: Its variables are constantly shifting, alive and moving with us.
Despite its title, “Driven by Intuition,” ironically, lacks human intuition. It’s cool and soulless, a ghost in the shell—even with a strong director’s credit. We’ve established that data left on its own isn’t the future of TV. But if people still wonder whether it’s the future of advertising, then somebody needs to hand us a hammer.