Tweet by Balaji that has been on my mind : AI doesn’t do it end-to-end. It does it middle-to-middle. The new bottlenecks are prompting and verifying.

Intelligence has become cheap Prompting = agency Verification = regulated industries

To survive in an post AI world, you either need crazy agency to create something yourself, or be part of a community where your role will be in verification (I’m thinking as a radiologist who might become a medicolegal wrapper around a sufficiently capable AI system)

I went to an amazing lecture/talk yesterday hosted by the adventures in awareness society - with Shamil Chanderia.

I don’t think I’ve ever seen such an excellent speaker in a while.

I think the lecture has been recorded, but the topics he covered :

• Your brain constructs your model of reality (Bayesian brain / predictive processing) ala Anil Seth
• A model of meditation/psychedelics that integrates this scientific understanding
• The self model ala Thomas Metzinger
• A model for attention/awareness
• The role of meditation in ‘deconstructing’ the world model, and then reconstructing

I think I need to watch the recorded lecture again! One epistemological stance that I’d like to explore was that he said that it is ‘stories’ all the way down’, and that there are good and bad stories. How do we actively guide the reconstruction process? Which stories are good/bad…

Reading about Friction. Digital services give the illusion that they are reducing friction, but they are really outsourcing it to somewhere else.