That's funny coming from you, Elon. You're easily the second if not the first most hated man in America. You're proof that self and social awareness just aren't very acute for those on the spectrum.
Has anyone explored the question of whether recursive self-improvement will actually be financially viable?
It seems to me that if AI can actually deploy capital at this scale in a way that reliably gets a return, it has to be at least as aligned with human values as capital markets are.
I get this, but i have yet to see a serious explanation of how that would happen. At some level you need a proxy for energy and where investment should go. Every computation system has to decide what to keep and what to throw out, which process to run, which experiment to perform, etc.
RSI could potentially be focussed on smaller, faster, better models which cost less to train and serve. We are still some ways off human genius level intelligence running on circa 20w.
I’m curious about the floor plan example - on how much are we using this general intelligence model for very specific tasks? What I mean to say is that I could not convert a picture into a floor plan if you ask 100 average humans, they would probably not be able to do it except for people who are trained in architecture. And I would think you could train the new newest AI model in architecture for a week or a month and it would perform any human very quickly.
AI becomes investable when it changes distribution for capital, not just workflow for operators. These JVs read less like software partnerships and more like private-market channel design.
The agentic stack reshuffling and Singularity threshold talk is fascinating. You connect AI capex pressures with deeper questions of jobs and meaning in a way that feels both technically sharp and philosophically grounded.
That's funny coming from you, Elon. You're easily the second if not the first most hated man in America. You're proof that self and social awareness just aren't very acute for those on the spectrum.
Has anyone explored the question of whether recursive self-improvement will actually be financially viable?
It seems to me that if AI can actually deploy capital at this scale in a way that reliably gets a return, it has to be at least as aligned with human values as capital markets are.
There is a group of people who
actually believe money will go bye-bye … hard to even think about …
I get this, but i have yet to see a serious explanation of how that would happen. At some level you need a proxy for energy and where investment should go. Every computation system has to decide what to keep and what to throw out, which process to run, which experiment to perform, etc.
RSI could potentially be focussed on smaller, faster, better models which cost less to train and serve. We are still some ways off human genius level intelligence running on circa 20w.
I’m curious about the floor plan example - on how much are we using this general intelligence model for very specific tasks? What I mean to say is that I could not convert a picture into a floor plan if you ask 100 average humans, they would probably not be able to do it except for people who are trained in architecture. And I would think you could train the new newest AI model in architecture for a week or a month and it would perform any human very quickly.
AI becomes investable when it changes distribution for capital, not just workflow for operators. These JVs read less like software partnerships and more like private-market channel design.
The agentic stack reshuffling and Singularity threshold talk is fascinating. You connect AI capex pressures with deeper questions of jobs and meaning in a way that feels both technically sharp and philosophically grounded.