Beyond the Hype: Separating AI Reality from Silicon Valley's Pipe Dreams

The AI Hype Machine: Where Enthusiasm Meets Reality Imagine browsing through film reviews on the Internet Movie Database. You'll find critics' reviews, but also amateur reviews from anyone who feels like sharing their two cents. It's amusing to see people gush over movies that were briefly popular, then vanished into obscurity. Take Darkness Falls, a forgettable horror flick about an evil tooth fairy. It has a Metascore of 23 and an IMDB rating of 5.0, yet someone gave it a glowing 10-star review.

This reminds me of an article by Kevin Roose in The New York Times, where he discusses how Silicon Valley is convinced we're on the cusp of achieving Artificial General Intelligence (AGI). AGI, roughly defined, is a general-purpose AI system that can perform almost all cognitive tasks a human can do. I'm not an AI skeptic; I think Large Language Models (LLMs) are powerful tools with real-world applications. However, we're nowhere near AGI, and the reason Silicon Valley thinks otherwise is that their understanding of human cognitive tasks is, well, a bit narrow.

Take Sam Altman's enthusiasm for OpenAI's creative writing model. He thinks it's got the vibe of metafiction just right. But, if we're being honest, the "vibes" might be convincing only if your literary sophistication is on par with a precocious high schooler. The writing samples Altman shared are, at best, amateurish attempts at metafiction. It's like saying a ChatGPT-generated essay on the trolley problem is profound simply because it sounds good to someone who's never delved deeper into philosophy.

This pattern repeats itself. AI can be helpful in discussing humanities concepts, but only up to a point. If you're looking for high school or undergraduate-level understanding, these tools might be sufficient. However, if you're aiming for graduate-level analysis, you'll quickly find yourself led astray. The same applies to art; AI can produce passable, mediocre art, but that's not exactly the pinnacle of human creativity.

So, what's driving this misplaced enthusiasm? It's a combination of two features of Silicon Valley tech bro culture. First, there's the conviction that they're geniuses who can excel in any field, not just computer programming. This is the same character trait that makes Elon Musk so... interesting. He's convinced of his own genius, even when he's talking about things he knows nothing about. Second, there's a basic lack of taste and a tendency to overestimate the significance of their own discoveries.

The trouble is, the Silicon Valley tech bro scene is incredibly insular. They don't often interact with people who can challenge their intro-level understanding of humanities and cognitive tasks. There's also an inherent incentive to spin exuberant narratives and exaggerate the potential of their products. In Silicon Valley, you don't get rich by selling products to consumers; you get rich by selling a story to investors who believe in it. It's not about creating something truly innovative; it's about crafting a compelling narrative that convinces people to put their money behind it.

This means that, even if tech bros recognized the limitations of their AI models, they couldn't admit it without deflating the narrative they need to sell. As Kevin Roose noted, AI executives have a vested interest in inflating AGI hype, but it's more than that. It's a culture where all ideas are big ideas, and all big ideas are unexamined. The result is a hype machine that's more focused on selling a story than on creating something truly groundbreaking.