4 ms·

> Will people stop working altogether? No. > Then what do we do with our time? Work. History is littered with "labor saving" inventions that were going to c

> Will people stop working altogether? No. > Then what do we do with our time? Work. History is littered with "labor saving" inventions that were going to change the way we work and lead to a leisure-focused future where machines did all the work for us. It never happened. Not once. The nature of work changed, and often, the "new" work takes up more of your time than the "old" work. I have no reason to believe that this also won't be the case, this time.

p-e-w 4y ago
AGI is not an "invention", it's a being. A being that is vastly superior to even the best humans at every imaginable task. The history of machinery is irrelevant when it comes to predicting how such a being will affect human labor. Nobody is going to continue paying people to do something that AIs can do better, 1000 times faster, and for 1/1000th the price. That just isn't going to happen. Humans working (other than perhaps as a pastime, i.e. pretending to work) won't be a thing in the future.
- LeoPanthera 4y ago
  I respect your opinion, but I disagree with it. I see GPT as a tool. An adaptive, flexible tool, but a tool.
  - sophiabits 4y ago
    As it exists today I’d agree with you that GPT is just a tool. I think without some architectural improvements it will likely always be a tool—predicting the next token probably accounts for some part of human intelligence, but I find it difficult to believe that’s /all/ our brain is doing. There’s presumably some ceiling to how good GPT can become without some new fundamental insight into what intelligence is. I don’t think AGI could be categorized as merely a tool. Even “just” the ability to run human-level intellect and reasoning at scale would be a game changer—humans are really expensive to produce and maintain compared to GPUs—and there’s no reason to believe real AGI would be limited to the capabilities of the human brain.
  - p-e-w 4y ago
    Yes, GPT (more accurately, GPT-4) is a tool. But future AIs will be more than that. Obviously, the current generation of AIs hasn't made humans obsolete yet. But consider that 5 years ago, AIs couldn't pass elementary school reading tests, and today they perform in the top 10% of law students in the bar exam. Unless something unforeseen happens, superhuman performance is just a matter of time.
- esalman 4y ago
  Will AI grow food at scale enough to support billions of people?

p-e-w 4y ago

AGI is not an "invention", it's a being. A being that is vastly superior to even the best humans at every imaginable task. The history of machinery is irrelevant when it comes to predicting how such a being will affect human labor. Nobody is going to continue paying people to do something that AIs can do better, 1000 times faster, and for 1/1000th the price. That just isn't going to happen. Humans working (other than perhaps as a pastime, i.e. pretending to work) won't be a thing in the future.

LeoPanthera 4y ago
I respect your opinion, but I disagree with it. I see GPT as a tool. An adaptive, flexible tool, but a tool.
- sophiabits 4y ago
  As it exists today I’d agree with you that GPT is just a tool. I think without some architectural improvements it will likely always be a tool—predicting the next token probably accounts for some part of human intelligence, but I find it difficult to believe that’s /all/ our brain is doing. There’s presumably some ceiling to how good GPT can become without some new fundamental insight into what intelligence is. I don’t think AGI could be categorized as merely a tool. Even “just” the ability to run human-level intellect and reasoning at scale would be a game changer—humans are really expensive to produce and maintain compared to GPUs—and there’s no reason to believe real AGI would be limited to the capabilities of the human brain.
- p-e-w 4y ago
  Yes, GPT (more accurately, GPT-4) is a tool. But future AIs will be more than that. Obviously, the current generation of AIs hasn't made humans obsolete yet. But consider that 5 years ago, AIs couldn't pass elementary school reading tests, and today they perform in the top 10% of law students in the bar exam. Unless something unforeseen happens, superhuman performance is just a matter of time.
esalman 4y ago
Will AI grow food at scale enough to support billions of people?