3 ms·

The problem is the jobs you imagine people moving to won't materialize because AI will do those too. In the past, efficiency increases lead to the realization o

The problem is the jobs you imagine people moving to won't materialize because AI will do those too. In the past, efficiency increases lead to the realization of new modes of transactions becoming economically viable. The people made obsolete moved to entirely new jobs because the tech was narrowly applicable. AGI is fully general, there is no space of jobs that humans could do but AGI cannot do. This is the end game for human labor. Society is not ready, and probably never will be.

tombert 2y ago
What makes you think that we're actually close to AGI? I'm actually asking, because it seems like a buzzword that the media throws around a lot of handwaves the details away and just acts like it's inevitable. The latest thing I saw about it seemed to indicate that OpenAI was looking to build a dedicated nuclear power plant because AGI on current hardware might be possible with that much energy [1]. Even if they do crack AGI with that, it's not like its energy consumption will be justifiable. [1] https://www.independent.co.uk/tech/openai-nuclear-fusion-energy-ai-b2557064.html https://www.independent.co.uk/tech/openai-nuclear-fusion-ene...
- hackinthebochs 2y ago
  This article[1] does a good job of laying out the argument from trends and arguing that taking the trends seriously suggests reaching AGI in a few years. Admittedly, my own argument in favor of near-term AGI is much less sophisticated but is also largely based on extrapolating trends and noting the leaps in capabilities that can come from singular architectural innovations. The GPT family has been amazingly productive, and they result from largely a single architectural innovation. They have glaring deficiencies, but their failures are fairly well understood and overcoming them are basically engineering problems at this point. Whereas 10 years ago no one had any idea how to get from the SOTA at the time to AGI, I see a clear path to it from here. [1] https://situational-awareness.ai/from-gpt-4-to-agi/ https://situational-awareness.ai/from-gpt-4-to-agi/

tombert 2y ago

What makes you think that we're actually close to AGI? I'm actually asking, because it seems like a buzzword that the media throws around a lot of handwaves the details away and just acts like it's inevitable. The latest thing I saw about it seemed to indicate that OpenAI was looking to build a dedicated nuclear power plant because AGI on current hardware might be possible with that much energy [1]. Even if they do crack AGI with that, it's not like its energy consumption will be justifiable. [1] https://www.independent.co.uk/tech/openai-nuclear-fusion-energy-ai-b2557064.html https://www.independent.co.uk/tech/openai-nuclear-fusion-ene...

hackinthebochs 2y ago
This article[1] does a good job of laying out the argument from trends and arguing that taking the trends seriously suggests reaching AGI in a few years. Admittedly, my own argument in favor of near-term AGI is much less sophisticated but is also largely based on extrapolating trends and noting the leaps in capabilities that can come from singular architectural innovations. The GPT family has been amazingly productive, and they result from largely a single architectural innovation. They have glaring deficiencies, but their failures are fairly well understood and overcoming them are basically engineering problems at this point. Whereas 10 years ago no one had any idea how to get from the SOTA at the time to AGI, I see a clear path to it from here. [1] https://situational-awareness.ai/from-gpt-4-to-agi/ https://situational-awareness.ai/from-gpt-4-to-agi/