On Sat, Aug 1, 2026 at 12:06 PM Matt Mahoney <[email protected]> wrote:
> So to answer my earlier questions, the data you would compress to settle > questions about macro social theories would be Hume's guillotine > No. Perhaps I should delete the dataset there since it was merely a "toy" example of macrosocial data to illustrate the general idea. This should be obvious from the README where I discuss data selection as involving conflicting interests. A realistic contest to bring conflicting interests to the table with an objective criterion would involve a lot of negotiation about what data to include -- just as I described in the README. The LLMs also get confused by this since they, too, can't seem to think at the level of game theoretic incentives to resolve conflicts that could end up killing like 20% of the population over who gets control of the central levers of an out of control central government. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5b58bcc51c493d41-M3f3dfba4c6f91671ffc03024 Delivery options: https://agi.topicbox.com/groups/agi/subscription
