COULD A COM­PUTER TAKE YOUR JOB?

AI is a pow­er­ful cre­ative tool; can it re­place the hu­man touch?

Campaign Middle East - - FRONT PAGE -

Ar­ti­fi­cial in­tel­li­gence al­go­rithms can now re­place so­cial me­dia mar­ket­ing agen­cies, paint a “new Rem­brandt”, project the im­age of a per­fect mother, com­pose com­mer­cially vi­able mu­sic and even di­rect ads, trail­ers and short films. Just how wor­ried should cre­atives be?

Ac­cord­ing to Matt Webb, global chief tech­nol­ogy of­fi­cer at Mirum, an AI al­go­rithm is the­o­ret­i­cally ca­pa­ble of min­ing data from a brand and a brief and com­bin­ing it with in­for­ma­tion gath­ered to solve prob­lems. In that sense, AI could con­ceiv­ably an­swer a cre­ative brief, he con­cludes.

The rea­son­ing and pro­cesses used to teach AI to com­pose mu­sic turn out to be re­mark­ably sim­i­lar.

Pa­trick Sto­bbs, co­founder and chief op­er­at­ing of­fi­cer of AI mu­sic com­poser Jukedeck, doesn’t seem to believe in the di­vin­ity of cre­ativ­ity. Sto­bbs, a clas­si­cally trained mu­si­cian with a mu­sic de­gree from Cam­bridge, says: “Cre­ativ­ity comes down to im­mers­ing your­self in a field, as­sim­i­lat­ing the in­for­ma­tion of that world and re­com­bin­ing it in dif­fer­ent ways.”

Mu­si­cians ev­ery­where, he adds, learn by gain­ing a deep un­der­stand­ing of their field, break­ing down pat­terns, analysing the ef­fects of what they do, and per­fect­ing and re­com­bin­ing to achieve an in­tended ef­fect.

“That’s how we taught the AI mu­sic,” Sto­bbs ex­plains. “It analy­ses large data sets of mu­sic, looks for pat­terns, in­fers the rules of com­po­si­tion and cre­ates a set of prob­a­bil­i­ties and rules – such as, 70 per cent of the time, a piece of folk mu­sic starts on a tonic cord and, 80 per cent of the time, it ends on a per­fect cadence. It then com­poses, but does not copy, based on these rules.”

And it’s im­prov­ing and learn­ing all the time. Jukedeck’s com­po­si­tions from 2014 were re­ceived with de­ri­sion but, by 2015, they were re­sponded to favourably and, by 2016, were in­dis­tin­guish­able from that of a hu­man be­ing.

Jukedeck is ca­pa­ble of pro­duc­ing unique com­po­si­tions in sec­onds, tai­lored to the mood, taste or genre se­lected by the user. They largely grace YouTube videos but have started be­ing used by brands such as Sotheby’s and Coca- Cola ( in Mex­ico). The start-up has re­cently signed a deal with a ma­jor ad­ver­tis­ing net­work.

But there’s no need for cre­atives to panic. Not for the next 30 or 40 years, any­way, ac­cord­ing to Stu­art Wapling­ton. He is chief ex­ec­u­tive of Happy Fin­ish, a cre­ative tech and con­tent stu­dio with its own “cre­ative AI” di­vi­sion.

“At this point, AI al­go­rithms are still not ca­pa­ble of solv­ing un­bounded prob­lems. They can­not, for ex­am­ple, sit down with blank pa­per and pro­duce copy for you. How­ever, within cer­tain pa­ram­e­ters, an AI can solve prob­lems bet­ter than a hu­man be­ing,” he says.

And it would be a mis­take to think that the ideas and work com­ing from AI are purely ster­ile and de­riv­a­tive. “It’s not cre­ative in a hu­man sense, but it has its own oth­er­worldly style,” Wapling­ton says.

Still, cre­at­ing “The Next Rem­brandt” with the aid of AI for ING proved to Bas Korsten, cre­ative part­ner at J Wal­ter Thomp­son Am­s­ter­dam, that the hu­man abil­ity to think crit­i­cally is still very much needed. “I’m not wor­ried about los­ing my job any time soon,” he says. Cre­ative col­lab­o­ra­tors The team at Happy Fin­ish view their AI cre­ations as col­lab­o­ra­tors. As­sis­tants that make the im­pos­si­ble pos­si­ble and im­prove the cre­ative process.

“We use AI as a ba­sis for per­son­al­is­ing a cam­paign, for ex­am­ple, or we use it as a pro­duc­tion tool or cre­ative as­sis­tant,” Wapling­ton says. “The best work from AI comes from when it is used col­lab­o­ra­tively.”

With­out AI, cam­paigns such as Dove’s “Per­fect Mom” (see box) or “The Next Rem­brandt” wouldn’t have been pos­si­ble.

Per­haps a skilled im­per­son­ator of the great Dutch masters could have cre­ated “the next Rem­brandt” based on ex­pe­ri­ence, but the paint­ing wouldn’t have held the same al­lure be­cause it would have been cre­ated by a per­son who is not Rem­brandt. AI is a tool and, by study­ing and as­sim­i­lat­ing Rem­brandt’s work, it has, in a sense, taken in and re­com­bined the soul of the artist.

Korsten says: “All 80 bil­lion pix­els in the paint­ing were Rem­brandt pix­els. All, at one time or an­other, painted by Rem­brandt. We didn’t truly create some­thing new – we re­com­bined. We tried to get into his skin and his hands and create some­thing he could have cre­ated.” Even McCann World­group Ja­pan’s “AI cre­ative di­rec­tor” could not func­tion with­out hu­man be­ings. “It can­not di­rect every­thing from idea gen­er­a­tion to edit­ing,” Shun Mat­suzaka, the cre­ative plan­ner be­hind AI-CD, ad­mits. “It can only give the big cre­ative di­rec­tion at the very begin­ning, based on the client’s cre­ative brief.”

To trans­late the cre­ative brief, clients were asked to fill in a spe­cific form de­signed for the al­go­rithm. A hu­man cre­ative team would then man­age the pro­duc­tion process based on the big idea pro­vided by AI-CD.

While Korsten can’t en­vis­age an AI cre­ative, he can po­ten­tially see one as a cre­ative di­rec­tor. “You could feed the ma­chine every­thing and it could judge on ef­fec­tive­ness,” he says. “As a judg­ing me­chanic, I would be in­ter­ested to see it in ac­tion.” Cre­ativ­ity adapting to AI Al­ready, when it comes to on­line mar­ket­ing, AI al­go­rithms such as Al­bert by Ad­gorithms are ca­pa­ble of rapidly test­ing dif­fer­ent cre­ative com­bi­na­tions and op­ti­mis­ing for ef­fec­tive­ness.

Court­ney Con­nell, mar­ket­ing di­rec­tor at Cos­abella, an on­line lin­gerie com­pany that has re­placed its so­cial mar­ket­ing agency with Al­bert, says that the tech­nol­ogy has com­pletely changed her team’s cre­ative process.

She ex­plains: “Instead of giving the AI a set cre­ative fo­cus, we pro­vide dif­fer­ent el­e­ments. Al­bert then tests dif­fer­ent copy with dif­fer­ent photos and spends the first cou­ple of weeks op­ti­mis­ing. Once he’s op­ti­mised your cam­paigns, he’ll start to make his own.”

The hard­est thing to grasp, Con­nell says, is that Al­bert doesn’t view peo­ple the way peo­ple do: “Hu­mans are far more un­sub­tle.”

Take, for ex­am­ple, the back­lash in China over tar­geted ads by BMW on WeChat two years ago. Some con­sumers were out­raged at be­ing judged “too poor” to view ads by BMW when the so­cial plat­form rolled out the ads.

“I would say that hu­mans con­tam­i­nated the al­go­rithm,” Con­nell com­ments. “Hu­mans cre­ated those pa­ram­e­ters and think they [made] the right de­ci­sion.”

Al­bert, she con­tin­ues, doesn’t base his tar­get­ing on any­thing so crude as de­mo­graph­ics or in­come but “mi­crore­ac­tions he no­tices about a per­son”, Con­nell says. “How they in­ter­act, what they are click­ing on, what they en­gage with.”

“It’s pos­si­ble that ma­chines can read hu­man emo­tions bet­ter than hu­mans,” Alas­tair Green, ex­ec­u­tive cre­ative di­rec­tor at Publi­cis agency Team One, ob­serves. His team set out to create a film for the Saatchi & Saatchi New Di­rec­tors’ Show­case that was shot en­tirely by AI. The film, “Eclipse”, was screened at Cannes last year along­side 19 oth­ers di­rected by hu­mans.

Green says: “We hu­mans sat back and let the ma­chines do the cast­ing. We had all in­ter­nalised our favourite

AI- CD: McCann’s ‘AI cre­ative di­rec­tor’ comes up with the big idea

‘Eclipse’: the film was di­rected by AI tech­nol­ogy, which used ‘emo­tional data and mo­tion’ to make cast­ing de­ci­sions

‘The Next Rem­brandt’: ‘re­com­bined’ the artist’s work

Newspapers in English

Newspapers from UAE

© PressReader. All rights reserved.