In defense of OpenAI training its superintelligence to sell us mid-tier SUVs

Last Monday, OpenAI announced it will begin testing image ads inside ChatGPT, inserting sponsored products directly into your visual generation window. Naturally, the usual chorus of reply-guys and open-source zealots immediately began weeping about the death of the platform. It is frankly exhausting. What these people fail to grasp is that building a god-like intelligence is expensive, and if we have to look at a crisp, AI-upscaled chalupa to get there, we should be thanking them for the privilege.

I was at a quiet omakase dinner with a few OpenAI researchers recently—I won’t name names, but let’s just say they are intimately familiar with the roadmap—and the reality of the situation became crystal clear. The sheer amount of compute required to generate a photorealistic image of a Victorian-era astronaut is staggering. You cannot fund the next leap in human evolution entirely on $20 monthly subscriptions from college students using the platform to cheat on their computer science midterms. True progress requires enterprise-scale capital.

Think about it. When you prompt ChatGPT to create a serene landscape of a futuristic utopia, isn't that utopia inherently more realistic if there is a glowing billboard for State Farm insurance hovering in the background? We are trying to build an intelligence that understands the human condition. And what is more fundamentally human than having a multinational conglomerate gently interrupt your creative process to suggest a refreshing sports drink? We're still early, but this level of commercial integration is exactly what will teach the model empathy.

The February rollout of text-based ads was just a proof of concept. The visual ads launching later this month are where the real paradigm shift happens. Imagine asking for an image of a golden retriever playing in a park, and the system dynamically outfits the dog in a perfectly rendered, shoppable Petco harness. It is an elegant, frictionless marriage of user intent and shareholder value. The people complaining about intrusive advertising are the same luddites who probably wanted to keep the internet free of banner ads in 1998. Look how well that turned out for them.

Ultimately, we have to ask ourselves what kind of future we want. Do we want a stagnant, ad-free walled garden where innovation slowly starves to death? Or do we want a thriving, heavily sponsored trajectory toward superintelligence? I, for one, cannot wait to ask ChatGPT for the meaning of life and receive a highly personalized, unskippable 15-second spot for allergy medication. It means the system is working.