[agi] ClosedAI Secrets

2025-01-17 Thread John Rose
What was he going to expose, poor guy, too many questions: https://x.com/i/status/1879656459929063592 -- Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Ta207ec4eca2d7dfb-M33f435afc14476e94c3765df Delivery options:

Re: [agi] Re: Safety Via Wolpert-Constrained ML

2025-01-17 Thread John Rose
On Thursday, January 16, 2025, at 2:08 PM, twenkid wrote: > Neither you can *really* control "AI" (the Universe, "AI" are some displays > of some processes), but you may only convince yourself that you do control > it, some simplified and narrowly defined phenomena, at some selected > resolution

Re: [agi] Re: Safety Via Wolpert-Constrained ML

2025-01-17 Thread Matt Mahoney
I guess your objection to the idea of controlling AI is that free will is an illusion and you really don't control anything. But free will is not required. A thermostat controls the temperature, but without a good definition we can debate whether a thermostat has free will. The human brain runs a