This seems to skip over or underplay the main reason why we will not get a “FOOM” intelligence explosion. It's that it's constrained by hardware and the chips and electricity supply are maxed out.
It's like for a real explosion you need a lot of energetic stuff, say methane and oxygen, mixed up and ready to go bang. Instead we are burning all the gas as it comes out of the stove.
I'm not sure I believe in FOOM, but I don't think the hardware argument is so cut and dry. LLMs have had vast improvements to capabilities attributable purely to software improvements.
Considering how calorically efficient the human brain is when it comes to the breadth and depth of general problems it can solve, I don't feel like LLM software improvements have hit a theoretical limit yet.
Minor correction: OpenAI’s solution does not address the “smoothness problem”; this article states they solved the existence and smoothness problem. They claim to have solved parts (C) and (D) of the Millenium prize statement, but the smoothness problem refers to part (A).
“The first statement (A), which is known as the "smoothness" problem, states that there should always exist smooth and globally defined solutions to the Navier–Stokes equations in three-dimensional space. The second statement (C), known as the "breakdown" problem, states that there should be at least one set of initial conditions and external forces for which there are no smooth solutions to the Navier–Stokes equations.“
Edit: Though I notice now the OpenAI blogpost does claim to solve the “smoothness and existence problem”, so maybe there’s simply terminology differences here. I’ll note many mathematicians such as Dennis Sullivan suggest part (A) is the more important part.
You’re shifting your point now, so I’m not sure what you’re trying to argue for. Regardless, it’s both clearly false that art never involved skill and that there aren’t highly skilled artists today. It’s also clear that many artists, certainly every artist I’ve ever met, see self-improvement as a core feature of their life work. So I’m not really sure where you’re getting these ideas from. I know a bit about your background. You attended school in a place where the primary ambition was to get a job at a FANG company and accrue wealth. Is that where your ideas about art were formed?
Sorry, your idea of my background is incorrect. I also studied art history in college along with CS so you don't really know anything about me (or rather, maybe the college I went had CS people want to go to FAANG but that wasn't everyone's or mine and my art history clasmates did not have such an ambition anyway). Regardless my point was that art may have had technical skill but technical skill alone is not sufficient to create good art. There have been many periods in history where art technically has so-called regressed at least by the standards of technical skill maximalists which you see online these days, but again art was never about technical skill primarily. So fret not, regardless of how well a bot paints a picture it doesn't mean your or my art is invalid or pointless although I can understand your existential crisis.
This seems to skip over or underplay the main reason why we will not get a “FOOM” intelligence explosion. It's that it's constrained by hardware and the chips and electricity supply are maxed out.
It's like for a real explosion you need a lot of energetic stuff, say methane and oxygen, mixed up and ready to go bang. Instead we are burning all the gas as it comes out of the stove.
I'm not sure I believe in FOOM, but I don't think the hardware argument is so cut and dry. LLMs have had vast improvements to capabilities attributable purely to software improvements.
Considering how calorically efficient the human brain is when it comes to the breadth and depth of general problems it can solve, I don't feel like LLM software improvements have hit a theoretical limit yet.
I'm not saying they won't improve or self improve, just that it'll be steady more than explosive.
Minor correction: OpenAI’s solution does not address the “smoothness problem”; this article states they solved the existence and smoothness problem. They claim to have solved parts (C) and (D) of the Millenium prize statement, but the smoothness problem refers to part (A).
See https://en.wikipedia.org/wiki/Navier–Stokes_existence_and_sm...
“The first statement (A), which is known as the "smoothness" problem, states that there should always exist smooth and globally defined solutions to the Navier–Stokes equations in three-dimensional space. The second statement (C), known as the "breakdown" problem, states that there should be at least one set of initial conditions and external forces for which there are no smooth solutions to the Navier–Stokes equations.“
Edit: Though I notice now the OpenAI blogpost does claim to solve the “smoothness and existence problem”, so maybe there’s simply terminology differences here. I’ll note many mathematicians such as Dennis Sullivan suggest part (A) is the more important part.
What is the meaning of life when GPT6 can make such a painting? It disturbs me to no end.
It's not a painting. No paint was involved. Just a bunch of pixels extruded from a dataset.
Art was never about technical skill but expression of one's feelings.
Why do you believe this?
Why wouldn't you? Lots of art especially modern art has little technical skill involved especially compared to the old masters.
You’re shifting your point now, so I’m not sure what you’re trying to argue for. Regardless, it’s both clearly false that art never involved skill and that there aren’t highly skilled artists today. It’s also clear that many artists, certainly every artist I’ve ever met, see self-improvement as a core feature of their life work. So I’m not really sure where you’re getting these ideas from. I know a bit about your background. You attended school in a place where the primary ambition was to get a job at a FANG company and accrue wealth. Is that where your ideas about art were formed?
Sorry, your idea of my background is incorrect. I also studied art history in college along with CS so you don't really know anything about me (or rather, maybe the college I went had CS people want to go to FAANG but that wasn't everyone's or mine and my art history clasmates did not have such an ambition anyway). Regardless my point was that art may have had technical skill but technical skill alone is not sufficient to create good art. There have been many periods in history where art technically has so-called regressed at least by the standards of technical skill maximalists which you see online these days, but again art was never about technical skill primarily. So fret not, regardless of how well a bot paints a picture it doesn't mean your or my art is invalid or pointless although I can understand your existential crisis.
Be patient. It's coming. So I heard.