A senior executive at Scale AI has drawn a striking parallel between AI foundational labs and major Hollywood studios, suggesting that companies like OpenAI and Meta AI operate more like movie franchises than traditional software firms. In an episode of the a16z Podcast published on Tuesday, Ben Scharfstein, Scale AI’s head of product for enterprise applications, shared his most controversial take: that foundation model companies are best understood as content studios. Scharfstein argued that these AI labs make massive investments in high-profile models—akin to blockbuster films—expecting rapid returns, but with limited long-term relevance. “My hottest take is I think the right way to think about foundation labs is that they're like movie studios,” he said. “What I mean by that is they invest a ton of money in blockbusters that have a relatively short time span to pay them back.” He compared the development of GPT models to Marvel Studios’ strategy with franchises like The Avengers. “Marvel, you're investing in 'The Avengers' and you invest a billion dollars in it and then it maybe pays you back over the next 18 months. It's pretty irrelevant after that,” he explained. Similarly, he said, companies like OpenAI build successive versions of their models—GPT-3, GPT-4, and beyond—creating a franchise that gains momentum quickly but may lose relevance just as fast. Scharfstein noted that while these models might eventually evolve into more enduring software products—like turning a movie into a video game—the current state of AI model development resembles content creation more than software engineering. “It's just very similar. You're able to turn that franchise maybe into a video game and then maybe it is more like software. But actually right now the model companies look more like a content studio than they do like a software company.” Scale AI, which recently received a $14.3 billion investment from Meta, is known for its data annotation services that support major AI players like Google, Meta, and OpenAI. Its enterprise solutions, however, leverage models from OpenAI, Cohere, and Meta’s Llama 2 to build custom generative AI tools for clients. Despite its strong position, Scale AI recently cut 200 jobs—14% of its workforce—focusing on its generative AI division. In an internal email, interim CEO Jason Droege attributed the layoffs to overexpansion in the GenAI space, which left other divisions, including those serving the public sector, under-resourced. Scale AI has not yet responded to requests for comment on Scharfstein’s remarks.