Chasing Rabbbits

🤖 Super Intelligence Time Machine

I came across a note I wrote back in early 2018, apparently inspired by this post by François Chollet. Here it is without edits:

Intelligence cannot grow itself exponentially, it basically maxes out around linear. Which is why science, civilization, and individuals grow/improve at roughly linear rates. Plus, resource constraints and other bottlenecks act as breaks on runaway processes. How many GPUs would be required to fuel a runaway super intelligent AI? Crypto mining is already using more electricity than some countries.

Further, individuals are incapable of creating something smarter than themselves so it will take time and community to achieve this. AIs will function the same way when trying to improve themselves.

There is also no such thing as general intelliengence. The No Free Lunch theorem essentially proves that a super AGI is impossible.

—- I still don’t know why we are so desperate to recreate human intelligence in silicon. Are we really the pinnacle? Do we really need more things thinking like us? Human intelliengence is based on and influenced by our civilization, culture, and emotions. All things absent in computers. So I think this will make creating a silicon “human” brain impossible. But we have an almost entirely rational agent, why not see what it does instead of modeling it on irrational ones?

Is a neural network a digital recreation of society’s network effect with each node being “1 human”?

#AI #algorithms #thoughts