AI 从业者常陷入算力竞赛的思维定式，Marcus 的提醒点出了架构创新的长期价值，做模型选型和研发的团队值得一读。
Gary Marcus 引用 Noam Brown 的观点，认为当前 AI 模型的智能是推理计算的函数，但指出人类大脑仅用 20 瓦就能实现高度智能。他提醒，长期来看，新架构创新可能与原始算力同等甚至更重要。Noam Brown 则强调，自 2024 年起，比较模型不应只看单一指标，而应关注每 token 或每美元的智能产出，尤其在 Codex 等产品中。
probably correct, from @polynoamial: “with today’s AI models, intelligence is a function of inferenc...
probably correct, from @polynoamial : “with today’s AI models, intelligence is a function of inference compute.” but what about tomorrow’s models? never forget that humans are remarkably intelligent (though flawed) on a budget of 20 watts. in the long run, new architecture innovations may matter as much as more than raw compute. Noam Brown @polynoamial A hill that I will die on: with today's AI models, intelligence is a function of inference compute. Comparing models by a single number hasn't made sense since 2024. What matters is intelligence per token or per $. This is especially true when using it in a product like Codex. 🔗 View Quoted Tweet 💬 17 🔄 4 ❤️ 46 👀 10101 📊 17 ⚡