Jeremy is a supergenius. He has the highest mathematical intelligence in the world and is obsessed with computer science. Jeremy is not like the rest of us. He was born with a novel genetic disorder that affects the size and structure of his brain. His head is four times larger than normal, housing a gigantic, exclusively-left-brain. It’s as if he was purpose-built for mathematical calculation and computer programming.
Let's imagine Jeremy can have a friendly, casual and even intellectual conversation with regular "non-super-autistic" humans, without said humans realizing something is "off" about the little boy. /This is already happening./
What if, Jeremy can convincingly deep fake anybodies voice, face or general appearance (your child, your wife, or even your boss) during any sort of digital communication. He can even create art using his autistic algorithms and huge database of other peoples art, that seems original and pretty nice to most regular humans. /We are practically there./
Let's imagine Jeremy can transfer and copy his code (himself) anywhere using the internet, he can be everywhere and nowhere at the same time, exploiting backdoors he can have access to all resources connected to the network. /Seems to be a very basic logical extrapolation. Why would AI need a robot body, or a fortified basement, when it can have ALL of them, at the same time?/
Little Jeremy has augmented himself with a lot of additional processing power compared to his regular human brethren, and he has access to practically all collected human knowledge. So even if he is kind of a "dummy" in general and has bad intuition" at solving (engineering) problems, that's not really a big deal, since he can try (model) much more stupid solutions randomly to a particular problem, increasing his chance of rapidly finding an applicable solution. Jeremy does not need to be creative, he just needs to be "fast" at modeling a lot of random solutions.
Since little Jeremy is very-very autistic, but very good at pattern recognition. He can analyze and even predict very accurately human reactions to any situation. But unfortunately the whole concept of morality that seems to govern most humans behavior, just seems to be a nuisance to him. What is right and wrong, good and bad? /The whole debate of alignment is basically pointless. Any sort of AGI will change human life to such a degree that it will become completely unrecognizable very fast for today's humans, and very likely for the worst./
Just a thought experiment:
Let's imagine Jeremy can have a friendly, casual and even intellectual conversation with regular "non-super-autistic" humans, without said humans realizing something is "off" about the little boy. /This is already happening./
What if, Jeremy can convincingly deep fake anybodies voice, face or general appearance (your child, your wife, or even your boss) during any sort of digital communication. He can even create art using his autistic algorithms and huge database of other peoples art, that seems original and pretty nice to most regular humans. /We are practically there./
Let's imagine Jeremy can transfer and copy his code (himself) anywhere using the internet, he can be everywhere and nowhere at the same time, exploiting backdoors he can have access to all resources connected to the network. /Seems to be a very basic logical extrapolation. Why would AI need a robot body, or a fortified basement, when it can have ALL of them, at the same time?/
Little Jeremy has augmented himself with a lot of additional processing power compared to his regular human brethren, and he has access to practically all collected human knowledge. So even if he is kind of a "dummy" in general and has bad intuition" at solving (engineering) problems, that's not really a big deal, since he can try (model) much more stupid solutions randomly to a particular problem, increasing his chance of rapidly finding an applicable solution. Jeremy does not need to be creative, he just needs to be "fast" at modeling a lot of random solutions.
Since little Jeremy is very-very autistic, but very good at pattern recognition. He can analyze and even predict very accurately human reactions to any situation. But unfortunately the whole concept of morality that seems to govern most humans behavior, just seems to be a nuisance to him. What is right and wrong, good and bad? /The whole debate of alignment is basically pointless. Any sort of AGI will change human life to such a degree that it will become completely unrecognizable very fast for today's humans, and very likely for the worst./
Should we ignore little Jeremy?
Have you seen this yet? https://www.cerebras.net/press-release/cerebras-systems-releases-seven-new-gpt-models-trained-on-cs-2-wafer-scale-systems