I mean, if the big win is that marketing teams get to make ever more samey and irritating pre-roll trailers for YT, then I guess that's a sort of win, at least in the short term. If you're not employing teams of creatives in the creative side of marketing, you'll end up with a lot of bland ad work, spun up from prompts by some product manager who hasn't a creative bone in their body. The returns will be rapidly diminishing because the cost might go down, but so will the overall efficacy of these campaigns.
For meeting short-term profitability targets, I can definitely see the utility here, but in the long term, you're replacing innovation with iteration and speeding up the brain drain. When you've finally managed to replace all your creative teams with a giant, cheap AI that turns stuff around in minutes rather than weeks, you'll have nothing left but a bunch of middle-aged, middle-class, middle-managers feeding it dull prompts and getting the same iterative dross out the other end. And because it's cheap and easily, they'll make dozens of cheap crap ones, rather spend three or four months putting together something great. The internet, meanwhile, is flooded with all this bland AI content created from the same training data sources, which in turn becomes the training data itself - a sort of information inbreeding that makes everything gradually more dull and unremarkable.
Diminishing returns force product managers to panic and produce even more variants and the situation worsens...
The thing is that we've already seen what happens when algorithms make content, even with human steer: it's bland. Big publishers have been running business on the basis of big data for the last fifteen years and you can already see what that's done: more and more iterative dross informed by business intelligence and data analytics that no matter how pretty or impressive, is repetitive, iterative and fails to draw in audiences. The MCU and DCEU both fell prey to this kind of cultural inbreeding, where brands are creating content that's only ever in conversation with itself based on 'proven' templates, and output is becoming noticeably more malformed and wall-eyed with every new iteration.
Games, in broad brush strokes, are doing the same, relying on an iterative process to force a trend line from 2018 to keep going up and up into infinity, Disco Stu style. It's not working and the games that are blowing up now are the ones that ignore the algorithms and do something that seems right to them: Larian bring back the turn-based CRPG junked by most publishers 20 years ago and create one of the biggest-selling games of the generation. Balatro sells a bajillion copies and scoops up awards for a pixel-art poker roguelike. Clair Obscur mixes old-school FF with Persona and Sekiro and sets it in a fucked up Paris and create the year's most talked-about title. Disco Elysium revives the old-skool point-and-click adventure to create an off-the-wall political drama about an alcoholic cop, sells gangbusters. Helldivers 2 is a tongue-in-cheek send-up of corporate democracy, riffing hard on the work of Paul Verhoven that belongs in some genre all of its own and is Sony's fastest-selling title to date.
Generative AI just doubles down and turbo charges that already failed iterative model and speeds up the much need collapse, turning out twice much of the same shit we're already ignoring at a fraction of the time and they'll sell even less. The real winners in the long-term will be the creatives and the people who enjoy great ideas. The big losers will be shareholders and those bizarre fan boys who flag-wave for mega corps.