ChatGPT and I agree on this regarding agentic programming: "The goal shouldn't be to minimize human involvement. It should be to maximize useful implementation progress per unit of human attention, while preserving the engineer's ability to understand and redirect the system."
@Carter_Shade This is a New York thing, not any other major city thing. New York has a steam system bigger than all the other steam systems in the world combined.
@mathelirium Treating AI as inscrutable and giving up on understanding and controlling it is akin to surrendering yourself to a higher power, which is a stance as old as humanity. It will invite charlatans to act as interpreters of AI and manipulate people to serve their agenda.
@littmath the sad part is after that nobody wanted to look through them. why do that when instead you can just attach another telescope to the first telescope’s eyepiece to get an even better telescope, and so on ad infinitum
@poweruserlg@littmath Re 1. nature knows laws of nature better than humans but we still have Physicists and governments still allocate resources to all kinds of Physics research.
@mathewguy6@fchollet I misspoke, I meant that it was falsifiable, not falsified. In the sense that it was practical to design experiments that could rule it out shortly after it came out. Which is not the case for ST in general. The catastrophy stuff makes you sound 12, not sure if you are.
@poweruserlg@littmath It's the other way around. If nobody understands the math results AI produces, what is the role of AI in math? Math is marketing for frontier labs. They will move on once it becomes old news. Nobody will spend resources on something nobody understands.
@mathewguy6@fchollet QM was a response to observations of atomic spectra, atomic structure, stability of atoms, black body radiation etc. which were unexplainable by pre-QM physics. GR was more of a leap of faith, but got falsified pretty quickly. ST has no chance of being falsified.