12 Comments

Azeem and Paul, Regarding the move to substack, I'm concerned about its invisibility to AI as source material. Wikipedia does not recognize it as a validating source for facts, unlike the NY Times. How can substack, or parts of substack, become source materials for Aggregated Intelligence? If not, then the plutocratically controlled press will control the contents available to AI.

Expand full comment

it is a good question. i am sure Substack thinking will make its way into AI systems.

Expand full comment

How so, Azeem? I realize you know a lot more about what is going on than I, but I've written to the Wikimedia committee about this problem with no response. Wikipedia is a major source of info. Obviously, AI can't be training on every substack, just as it doesn't train on newspaper Op Eds. I don't see anyone anywhere discussing this problem.

Expand full comment

If they can crawl it, they will.

Expand full comment

afaik the big AI models train on pretty much any publicly-available data, i.e. the entire internet

Expand full comment

So, Kelin, when it answers a question, how does it rate the reliability of the material it crawls?

Expand full comment

Your mental model for this is wrong. There is no internal rating system. It uses different relationships.

Expand full comment

So the fact that you are a node in many key networks gives your posts higher standing... i.e., more likely to be cited in an answer?

Expand full comment

This is a great talk, thank you, Azeem. "Technology as a substitute for globalization" ... true, but maybe I would re-phrase it "technology as a way to globalise moving bits instead of atoms".

Expand full comment

If the US places tariffs on trading partners like Canada and Mexico, how soon before they retaliate? Did global tariffs help bring about the Great Depression?

Expand full comment

If the surge in inflation was mostly due to supply chain disruptions, was it necessary to raise interest rates?

Expand full comment