I know AGI hype is the thing these days, but I don’t see it happening anytime soon. Think about what motivates you to do anything. First thing when you wake in the morning, what makes you get out of bed? What makes you socialize with people? What makes you find a job and earn money? Why do you like/dislike certain things? How would you find data that can motivate these goals? If we could find data, how do we encode the information into a format that we can feed into a neural network. It simply won’t happen with an LLM that is trained quite literally to predict the next token in a sequence. I think the success of LLM is due to the fact that natural language is able to capture some of the conditions of the human experience. How are you going to encode what it feels like to eat your favorite dish? Listen to your favorite song? The list goes on and on.
I know AGI hype is the thing these days, but I don’t see it happening anytime soon. Think about what motivates you to do anything. First thing when you wake in the morning, what makes you get out of bed? What makes you socialize with people? What makes you find a job and earn money? Why do you like/dislike certain things? How would you find data that can motivate these goals? If we could find data, how do we encode the information into a format that we can feed into a neural network. It simply won’t happen with an LLM that is trained quite literally to predict the next token in a sequence. I think the success of LLM is due to the fact that natural language is able to capture some of the conditions of the human experience. How are you going to encode what it feels like to eat your favorite dish? Listen to your favorite song? The list goes on and on.