Key Takeaways

  • AI is in its 'Atari command line stage': Bill Maris, founder of Section 32 and former CEO of Google Ventures, argues that current AI feels rudimentary, like early text-based video games. It's clunky, inconsistent, and often lacks memory.
  • Explosive 5-year leap: Maris predicts AI will evolve to a 'PlayStation 10 stage' within the next five years. This isn't incremental progress; he sees a fundamental leap that will solve present limitations like memory and consistency at an unprecedented pace.
  • Invest in the 'machinery,' not the models: Forget chasing the next large language model. Maris's investment strategy focuses on the underlying infrastructure that enables this rapid evolution—think platforms, physics engines, controllers, and GPUs.
  • Hardware and tools drive progress: Just as better hardware and input devices transformed gaming more than better stories, Maris believes advances in foundational AI components will accelerate ambient computing more than simply scaling current models.

The Atari Analogy and a Compressed Decade

Imagine the clunky, text-based games of the 1980s. That’s where Bill Maris says we are with AI today. “I think we're at the Atari command line stage of AI,” Maris observed, contrasting it with the sophisticated, immersive worlds of modern gaming. But here’s the kicker: the leap from Atari to PlayStation 10, a transition that took decades for gaming, will happen in AI in just five years. “What's happened to the gaming industry from the 80s to today is going to happen in AI but in the next like 5 years,” Maris stated on the All-In Podcast.

This isn't a casual prediction. Maris, known for his early bets at Google Ventures on companies like Nest and Uber, anchors this accelerated timeline in the current pace of technological advancement and capital deployment. He sees a rapid compounding of capabilities, pushing AI past its current limitations of memory and consistent output to a point where ambient computing becomes a practical reality.

Bet on the Picks and Shovels, Not the Gold Rush

Most founders are thinking about building the next great AI application or training a slightly bigger, better model. Maris urges a different perspective. His investment thesis isn't about the gold itself, but the picks and shovels the prospectors will need. He makes it clear: “I don't plan on investing in kind of larger models, right? Just like it wasn't better stories that would make better games. It was controllers and physics engines and GPUs.”

This is a critical distinction for anyone building in the space. The true leverage, Maris suggests, isn't in iterating on existing models or trying to outcompete giants like Google or OpenAI on model size. Instead, it lies in the enabling technologies – the computational platforms, the specialized hardware, the simulation environments, and the low-level tools that will allow these 'PlayStation 10' AI systems to function. These are the parts of the AI cycle that Maris finds interesting, believing they will make a fully ambient computing reality happen within five years. “And it's not just bigger models,” he emphasized.

What to Do With This

If you're building in AI, stop obsessing over marginally better model performance. Instead, evaluate where your core competence intersects with the foundational 'machinery' Maris describes: GPUs, physics engines, specialized data pipelines, or new interaction paradigms. Spend this week brainstorming three product ideas that directly enable the next generation of AI builders, focusing on tools or infrastructure that resolve current AI limitations like memory, consistency, or real-time interaction, rather than just building on top of today's 'Atari-stage' models.