- We need to take a longer-term view of AI as a tool to speed production but not one that replaces human talent, a panel of experts say.
- The industry is hesitant about using AI because of ethical and copyrights concerns, but perhaps blockchain technology could help track and verify content.
- In just a decade we could have AI-driven real-time personalized storytelling.
Pragmatism has replaced fear as the sentiment most likely to be directed at generative AI by creative media companies and tech developers.
“If you’re not using AI it’s sort of like saying you’re not going to have a mobile strategy,” says Jeremy Toeman, Founder, AugX Labs. “AI was commoditized in less and under a year and now it’s just as much a building block for doing things than anything else.”
He was speaking on the video panel “Beyond ChatGPT: How AI Is Transforming Streaming Workflows and Businesses,” moderated by Ben Ratner, director of news technology at Boston 25 News.
Like Toeman, Steve Vonder Haar, senior analyst for Intelligent Video & Enterprise at IntelliVid Research, is taking a long-term view:
“In a decade we will not be using the term artificial intelligence at all because it’s not really descriptive of the value that these capabilities are delivering to the marketplace, just as was the case in the early 1990s when ‘information superhighway’ was in fashion. As we move forward, the term AI is going to fade away.”
All these experts agree that AI helps by speeding up processes, but that if users really want to make creative content with value then AI only gets you part of the way. It may automate 80% of the previously manual process, but human skill, knowledge and taste making is crucial for finesse and polish.
As Mobeon CEO Mark Alamares puts it, AI enables teams and individuals to amplify their capabilities to make a much more efficient production process overall. “AI will enhance what we’re doing both on a creative and technical level, in the broad sense,” he says.
Vonder Haar likens generative AI to “getting people past that blank sheet of paper.”
Toeman says his colleagues use ChatGPT to evolve their own strategies and question their own thinking, resulting in higher-caliber of outcomes than they would otherwise achieve.
The downsides of reliance on generative AI include stifling independent creative thinking and learning.
“I still remember when I used to know phone numbers. And then, thanks to Google, we don’t know facts. And now, thanks to GPT, we might not know anything,” says Toeman. “That’s obviously hyper cynical, but there’s somewhat of a concern.”
Alamares reports a lot of hesitancy in the creative industries, because of ethical concerns, but most of the panelists are pessimistic that anything can be done to concretely vet every copyright infringement or deepfake.
“For me, it’s a garbage in-garbage out,” says Vonder Haar. “If you’re going to take information from the web, then God bless you, because you’re not going to have a trusted source of information from which to draw.
“The real future for AI in the business sense is going to be in the development of limited datasets that are used to inform decision making within a specific corporate network or a specific realm of individuals.”
There is some discussion about whether regulatory bodies like MPEG could devise a scheme to watermark video directly into video codecs as a way of tracking and verifying content. Vonder Haar suggests that Vizrt’s NDI video-over-IP protocol could be used to the same end.
“NDI is relatively widely distributed among devices [and they] would also have the opportunity to create some sort of standard that would help in this type of watermarking of real content.”
Perhaps blockchain technology holds the most promise for differentiating real from fake video.
“If you’ve got blockchain-certified video that was recorded from a blockchain-certified camera you can know that that’s the original,” says Toeman. “I think that’s going to be hopefully our savior through this all, I don’t know.”
On the storytelling side, AI will be able to jumpstart creative thinking and fast forward the scripting process but there’s a belief that traditional storytelling talent will rise to the top.
“The point is ChatGPT drives toward the norm, toward the middle, toward the average,” says Toeman. “So you use it to do average things. If you want to be a great storyteller, you will write a better prompt to get more out of it using it, but all the tools are kind of the same from this perspective.”
The consequence, he says, is that reliance on generative AI tools alone will only churn out very average mediocre content. The flip side is this trend will enable master storytellers to use the tool to shine.
“A great example might be Jon Favreau, who made the first two seasons of The Mandalorian using a lot of new tech to make that storytelling cheaper, faster, easier. But it’s not the tech that made the story.
“So, if you can write a very clever prompt for your unique story angle, and then add your own special sauce on top of that, that’s where [AI is going].”
Corey Behnke, co-founder and producer at LiveX, says he believes there will be more demand for producer oversight and moderation of AI than ever before when it comes to live streaming.
AI could help with 80% of the mundane tasks in video production but that still leaves 20% for actual humans to get the product right.
“The place where you are going to deliver your value is that last 20% of broadcast quality video; that’s [what] separates a basic piece of video from a high polished piece of content,” agrees Vonder Haar.
“Because so much more content will be developed there will be more opportunities for high-end producers. The folk who were only going to be in trouble will be those who [operate] at the very basic levels of video production rather than the high end of the market space.”
Asked where he thinks AI will be a year from now, Toeman anticipates we will start seeing the first AI-generated content but no one will like it: “That’s my hunch. I think the content industry stays away from AI for [maybe] five years and by only then we’ll start seeing it used sort of the way CGI showed up in movies.”
In a decade we could have AI-driven real-time bespoke storytelling: “Make me a video in the style of Harry Potter about a sci-fi wizard on an asteroid, and I want it to be 90 minutes long, and make it seem like it’s written by Quentin Tarantino. I think that’s 10 years from now, tops.”