You are reading an exclusive WrapPRO article for free. Want to level up your entertainment career? Click here for more information.
“Last year, generative AI was on no one’s radar screen,” Peter Csathy, chairman of Creative Media said Wednesday during TheWrap’s annual industry conference TheGrill.
Csathy — alongside Erick Opeka, president and chief strategy officer for Cineverse; Mark Mothersbaugh of DEVO; explAIner-in-chief at Human Artistry Campaign Dr. Moiya McTier; head of media and entertainment partnerships for Intel, Rick Hack; Lore Machine founder Thobey Campion; and What’s Trending founder Shira Lazar — discussed the future of AI, which, according to Opeka, isn’t as expansive as it’s been feared to be.
“The bigger picture is we’re using a lot of the technology we have to think about where is this going and how do we build systems to build the next generation of entertainment?” Opeka said.
“There’s very limited data today,” he continued. “Most of the LLMs today are very focused on textual-based data or individual images.” This is because the datasets created for AI were done over a decade ago and are still limited in scope. “The datasets that exist today are very, very far away from saying, ‘Hey, create ‘Breakfast at Tiffany’s’ but put this actor in it,” he said.
What would it take for that to happen, then? As Opeka laid out, there are several things that needed to be negotiated starting with regulation.
“Are we, as a society, going to allow that?” he said. There’s also a logistical component. The sheer amount of processing power and data that comes with building a system to allow that are, as Opeka explained, the “exclusive playground” of the largest companies in the world. It wouldn’t be something available to the masses.
And because the current data is all textual, it forces AI generators to craft things using descriptive tags. “How do you have descriptive tags at a level of depth to do full-length movies? It takes an incredible amount of processing power [and] you need experts that are watching,” Opeka said.
Opeksa explained that despite this being AI, there is still a lot of manpower that goes into processing. Most language models today, he said, have 40-50 people working on them: “Literally armies of people behind the curtain [are] training these models.”
That also begs the question of how people would be compensated, whether working on the models or being the basis for them. One would also have to answer the age-old question of whether there’s a business model for that.
As Opeka further explained, it’s one thing to craft something that could let a film viewer watch “Robocop” with a different actor in it, or that’s a comedy, but would a mass audience be willing to pay for that? He compared the tech to the early 1900s, when people were watching dime-store nickelodeons (short videos).
“It’s a far cry from ‘Avatar,’ but it’s the genesis that led to ‘Avatar.’”
Watch the full panel in the video below.
About TheGrill: For more than a decade, TheGrill event series has led the conversation on the convergence of entertainment, media and technology, bringing together newsmakers to debate the challenges of and opportunities for making content in the digital age. TheGrill delivers a unique series of curated discussions, industry panels and networking activations that explore the ever-changing media landscape.
TheGrill is powered by the essential source for entertainment insiders, WrapPRO, TheWrap’s premium content subscription platform. This members-only service and community provides deep analysis and access — that can’t be found anywhere else — on the business of entertainment, streaming and media. Click here for more information on WrapPRO.
Leave a Reply