Impact and opportunity surfaced as the dominant theme of a full day of Digital Hollywood sessions devoted to artificial intelligence at CES 2024. We are in a period of disruption similar to the early 90s when the Internet went mainstream, said Forbes columnist Charlie Fink, moderating a panel of industry leaders from CAA, Paramount, HTC, Nvidia and Google. Yet despite the transformation already underway, panelists agreed that this is neither the first nor last technology to shift the status quo, more the latest example of inevitable change and adjustment. The current conversations around AI at CES are a refreshing departure after a few years of evolutionary, not revolutionary tech confabs.

During an opening session, Steve Canepa of IBM and Richard Kerris from Nvidia, along with Nokia’s Leslie Shannon who moderated, key points were made about the quality and sourcing of the data coming into large learning models and guardrails for when it comes out. “Trust in, trust out,” said Kerris.

From left to right: Leslie Shannon, Steve Canepa, Richard Kerris

Canepa noted that the notion of bigger models being better is not the case. “A lot of models have vacuumed up so much data,” he said. “Now we have to clean up the data and piles of rights.”

For generative AI, rights issues are a growing concern, which companies like Adobe, Getty and Shutterstock are addressing by managing rights within their generative models.

Kerris suggested that “we will start to see watermarks and [ways to] indemnify use cases. These things we worry about are an opportunity for companies to build security, governance, rights management,” and more, similar to the industry that has developed around cybersecurity.

Shannon cited a Stanford study that looked at big image video models that revealed “thousands of inappropriate images.”

Despite the security, ethical and intellectual property concerns, AI is already proving to be extremely valuable in creating new experiences. Canepa gave an example of the brand conscious Masters golf tournament and how “AI analyzes and classifies video from the tournament. Then our models can generate commentary, trained on their data to create branded content.”

Nvidia plans to make several AI announcements in March at its GTC conference, including more on the company’s Omniverse platform and its relation to digital twins.

The ability to create physically accurate, large-scale simulations — digital twins — is a powerful tool at enterprise-scale in animation and manufacturing. Yet, said Kerris, it also applies to the home. Instead of robot vacuum cleaners spending their first weeks of service bumping into things to learn the layout, with a digital twin of a home the robot will know from day one.