3 ms·

There’s a danger of confusing the problems with actually existing AI ethics work for evidence that there’s no need to ethically evaluate AI. On any moral reali

There’s a danger of confusing the problems with actually existing AI ethics work for evidence that there’s no need to ethically evaluate AI. On any moral realist view, AI will likely be extremely important. And insofar as ethical standards are used to organise society, ethical views on the use of AI will likely be developed in the same way as they are at least very weakly developed and observed in most other fields, so an understanding of them will be important for understanding society simpliciter.

pixl97 4y ago
To say AI will be important is like saying 'guns' will be important or 'internal combustion engines' will be important around the time they are created. In the past these creations took some time to massively change the world, but they did so in unimaginable ways. With AI we are not going to get the benefit of time, especially if the growth rate of the intelligence of AI keeps at is recent, or even faster trajectories. If there is any chance that we could, even unintentionally, create super intelligence the only ethical answer is to not create it at all. It would be a species ending mistake at our current level of knowledge.
- lame-robot-hoax 4y ago
  > especially if the growth rate of the intelligence of AI keeps at is recent, or even faster trajectories. How are you measuring intelligence? How are you defining intelligence? Formal measures of intelligence like ARC still show even SOTA LLM’s doing incredibly poorly. As Chollet said, people are misinterpreting LLMs as looking or seeming intelligent, therefore they are, just in the same sense that they think there must be lemon in this lemon flavored candy since it tastes like lemon, when there’s no lemon at all. This just seems like more Yudkowsky BS about how AGI is right around the corner, despite industry leaders (Ng, Chollet, etc.) saying we aren’t much, if any closer to AGI despite recent developments. There’s no indications that LLM -> AGI.
  - pixl97 4y ago
    Chollet really reminds me of someone in 1901 saying men will never fly because their is no way we can flap our wings fast enough.
- otabdeveloper4 4y ago
  I think you should lay off the sci-fi.

pixl97 4y ago

To say AI will be important is like saying 'guns' will be important or 'internal combustion engines' will be important around the time they are created. In the past these creations took some time to massively change the world, but they did so in unimaginable ways. With AI we are not going to get the benefit of time, especially if the growth rate of the intelligence of AI keeps at is recent, or even faster trajectories. If there is any chance that we could, even unintentionally, create super intelligence the only ethical answer is to not create it at all. It would be a species ending mistake at our current level of knowledge.

lame-robot-hoax 4y ago
> especially if the growth rate of the intelligence of AI keeps at is recent, or even faster trajectories. How are you measuring intelligence? How are you defining intelligence? Formal measures of intelligence like ARC still show even SOTA LLM’s doing incredibly poorly. As Chollet said, people are misinterpreting LLMs as looking or seeming intelligent, therefore they are, just in the same sense that they think there must be lemon in this lemon flavored candy since it tastes like lemon, when there’s no lemon at all. This just seems like more Yudkowsky BS about how AGI is right around the corner, despite industry leaders (Ng, Chollet, etc.) saying we aren’t much, if any closer to AGI despite recent developments. There’s no indications that LLM -> AGI.
- pixl97 4y ago
  Chollet really reminds me of someone in 1901 saying men will never fly because their is no way we can flap our wings fast enough.
otabdeveloper4 4y ago
I think you should lay off the sci-fi.