StreamGPT? AMG’s Senzon dreams up streaming's AI-powered future

When it comes to streaming, engagement is often the name of the game, with services and platform players all working on ways to not only have users tune in or sign up but stay longer and keep coming back often.

During a keynote session at the recent StreamTV Show in Colorado, Michael Senzon, president of digital at Allen Media Group, discussed engagement and the importance of creating a daily viewing habit –  one where the executive shared big dreams for how AI could eventually change the game, not just for FAST but streaming at large.

Before getting into his vision of an AI-powered future that entices and caters to viewers so they engage daily, Senzon disclosed some interesting stats and advice based on current efforts for AMG’s Local Now, where the free ad-supported streaming TV (FAST) service has chosen to lead with a niche focus of local programming – leaning on proprietary tech, human capital and several partners - to stand out in a crowded space.  When it comes to creating a daily habit, he said the most valuable ARPU users usually come into services and platforms at least a couple of times a week. For services, engaged users mean they’re being monetized and don’t require as much marketing spend. At Local Now specifically, time spent per unique has “grown exponentially,” according to Senzon.

The free ad-supported streamer counts distribution across major platforms and has worked to build a user journey that includes multiple viewing sessions throughout the day. For its most-loyal users, Senzon described how viewers first come in for their exclusive local news station (of which it has 223 geotargeted Local Now channels in the U.S.) for about an hour, move on to a favorite affiliate like CBS, tend to follow it up with national news like Bloomberg or Dateline NBC, later lean in to entertainment with single-series IP channels from content partners, and - on a very successful day - round it out with a movie at night.

And it’s working. Through focus on content and curation and by leading with something unique followed by a range of broader news or entertainment options, Senzon said 70% of the Local Now audience now count as loyal users who engage with the service every day, marking a jump in the last year from just 15% previously.

StreamGPT?

To be sure, Senzon wasn’t the only one talking about AI at the StreamTV Show. As he noted, there were some commonly referred to AI streaming use cases that are likely to happen soon, such as AI determining what FAST channels to promote, or helping predict what movies should be in channels or that viewers should watch, and assisting with FAST channel creation. However, the executive took time to dream up and lay out several could-be scenarios of how he envisions AI changing FAST and streaming over the next five years and beyond.

To set the scene, Senzon pointed back to the early 2000s when going to several webpages a day was a common habit before the advent of social platforms like Facebook and others that became dominate and were keen to keep users within their platform confines. He feels the industry is “on the precipice of that with streaming,” pointing to smart TV makers like Vizio, Roku, Amazon and others that are investing in TV OS with an aim of being the only go-to destination for viewers.

“I think AI is going to take it a step further, and it’s going to be to streaming what social was to web pages,” Senzon said on stage during a keynote presentation.

But he also sees AI potentially transforming the TV OS and OEM’s altogether.

Senzon cited the current ecosystem of several and fragmented gateways for services and content partners with various ways to integrate. For example, services like Local Now can do direct deals to be included as a standalone app embedded on a smart TV interface or streaming sticks (of which there are many), act as a content partner to platforms, and/or have deals so channels are included in FAST service lineups, like on The Roku Channel or several others.

Fast forward five years and he envisions a broader picture where GenAI predictive capabilities could even replace OEM’s own operating software.

“Imagine if the next version of an OEM…the OS of an OEM is actually a ‘streamGPT’,” Senzon said. 

He described this as a model that understands the user, their patterns and gets to the heart of what they actually want to watch or feel. One where a person powers on the TV and it says, “’Hello Michael, by the way I’ve also created a TV show you might like’.”

Outside of the streaming realm he mentioned tech efforts like Apple Watches and Elon Musk’s Neuralink brain-chip company as examples of how AI might be able to connect emotion and figure out what humans want.

Senzon also told the audience to imagine an AI-curated FAST channel designed just for the viewer alongside a feed of their favorite TikTok videos, all delivered seamlessly through an AI-powered model that knows what content the person wants to watch.

 Of course, many questions around such capabilities or features could easily arise, with Senzon bringing up the point that, tech aside, rights and privacy are both aspects that would need to be figured out.  

GenAI TV – an end to content arms dealers?

Continuing further down the AI-powered path, Senzon talked about GenAI’s potential to disrupt content creation.

Some companies are already working on AI-generated content and TV series. Senzon used the examples of OpenAI’s Sora text-to-video generation tool that’s to be made available for public use. He also pointed to Showrunner, a text-to-episode platform from The Simulation, a company which considers itself “the Netflix of AI”. It already debuted a South Park AI episode generator last year and in May launched a creation platform that lets users make their own TV shows with AI. The company announced 10 web series TV shows made with the Showrunner platform – which, according to Forbes, can generate scenes and episodes between two and 16 minutes, including AI dialoge, voice, editing, consistent characters and story development based on 10-15-word prompts. It currently can generate episodes in styles such as cutout, anime, and 3D animation (not live action).

In an interview with Forbes, The Simulation (formerly Fable Studio) CEO Edward Saatchi noted that AI in general and its generation platform specifically isn’t really suited for entertainment series that have a full series arc like Breaking Bad.

“Today AI can’t sustain a story beyond one episode,” Saatchi told Forbes. “What AI is strongest at is deeply episodic shows with characters largely resetting every episode – sitcoms, police procedurals, space exploration.”

At the StreamTV Show, Senzon hypothesized efforts like Showrunner could advance and posed an interesting question.

“What if the [content] arms dealers don’t matter in the future? What if the next TV show I watch was created by AI and is better than Breaking Bad?” he posited.

That said, his dreams for AI in streaming may be quite a bit into the future – and the exec acknowledged they would require other moving pieces and players to align. For one, getting Hollywood unions representing actors, writers and directors, such as the Directors Guild of America, Writers Guild of America and Screen Actors Guild (SAG-AFTRA), on board, as well as IP rights.

Here are a few other fun ideas Senzon floated for AI-powered streaming (again, caveat that these are ideas and a successful leap to reality would mean figuring out not only tech but appetite, rights, and not to mention potential costs associated).

  •  An AI-generated NFL game that looks like a real pro football event but features the greatest players of all time and airs in the middle of summer
  •  A sequel to your favorite movie (in Senzon’s case Back to the Future 4)
  •   Evening news produced specifically for the user
  • FAST channels curated at the user-level, including shows made just for them
  • An AI-generated presidential debate between long-past and recent presidents or candidates (ie Abraham Lincoln versus former President Trump)

 

“There’s things if you open your mind to it, it’s really remarkable,” Senzon said. “And I think that, ultimately, is why AI will create the ultimate streaming service and make it the ultimate daily habit.”

Work on personalized recommendation engines

And while admitting some “lofty ideas” that might not come cheap for the future of AI in streaming, Senzon also noted separate work already being done by AMG as a long-term partner with Google on personalized recommendation engines that are tailored to the user.

For example, a feature where one user’s FAST EPG could look different than another’s as one has a preference for romantic comedies while the other likes Architectural Digest. AMG and Senzon have been talking about personalized FAST channels for at least a couple of years, including back in September 2022 when it partnered with Quickplay for its OTT platform. Quickplay earlier this year announced teaming with Google to use GenAI for streaming video discovery, enabling content programmers to create dynamic storefronts with personalized recommendations.

Some iterations of similar features – like surfacing recommendations based on other titles viewed – are in use today, but the recommendation engine mentioned at the StreamTV Show being built on the back end for Local Now hasn’t been implemented on the front end yet, according to Senzon.