Fantastic piece Lynne. Markets are an information ecosystem. Going to read this a couple more times.
Some of these dynamics are illustrated by LLM degradation research. Rather than improving, devolve incoherently when fed LLM output. Superficially the problem is computational, but actually cognitive process, implicit behavior, and situationally contextual. The knowledge ecosystem is deeply embedded in culture, imprinted in neural networks, and genetics; perhaps even inaccessible today.
This has strong implications for tech driven market systems and the limits of AI feasibility today.
I love these posts, they make some very deep economic questions accessible to engineers like myself! I found this blog because of my interest in energy, and this knowledge problem is a fascinating one in that context too. The LaCroix analogy (which we don't get in Australia, but it featured in an episode of Succession recently) sound to me like asking consumers to put a value of using electricity. There's some much context and individual knowledge wrapped up in that question that we typically gloss over so we can estimate a $/kWh amount and just treat that as a fixed assumption going forward.
Agreed—pampelmousse is without a doubt the best LaCroix flavor.
Fantastic piece Lynne. Markets are an information ecosystem. Going to read this a couple more times.
Some of these dynamics are illustrated by LLM degradation research. Rather than improving, devolve incoherently when fed LLM output. Superficially the problem is computational, but actually cognitive process, implicit behavior, and situationally contextual. The knowledge ecosystem is deeply embedded in culture, imprinted in neural networks, and genetics; perhaps even inaccessible today.
This has strong implications for tech driven market systems and the limits of AI feasibility today.
I love these posts, they make some very deep economic questions accessible to engineers like myself! I found this blog because of my interest in energy, and this knowledge problem is a fascinating one in that context too. The LaCroix analogy (which we don't get in Australia, but it featured in an episode of Succession recently) sound to me like asking consumers to put a value of using electricity. There's some much context and individual knowledge wrapped up in that question that we typically gloss over so we can estimate a $/kWh amount and just treat that as a fixed assumption going forward.
Every day I find another great substack like this one. Which is awesome, but it's getting hard to keep up :)
This is fantastic stuff. Can you provide the full reference for your 2015 book/article?