3 ms·

"OpenAI" nomenclature should be rejected. Do not allow this project to claim they are open: Its as closed-source as any for-profit company. What happen to HN "

by _red 6y ago

"OpenAI" nomenclature should be rejected. Do not allow this project to claim they are open: Its as closed-source as any for-profit company. What happen to HN "internet culture" that it falls for such obvious ploys without even calling it out?

FeepingCreature 6y ago
I don't think people are "falling" for it; you don't really see many people being confused that they can't find the GPT-3 repository. Personally, I just don't really care about the name - and I'm glad OpenAI is not open, purely on an AI safety basis. I think their original reasoning for openness was foolish, suicidal nonsense [1], and I'm glad they moved away from it. [1] https://slatestarcodex.com/2015/12/17/should-ai-be-open/ https://slatestarcodex.com/2015/12/17/should-ai-be-open/
- _red 6y ago
  >its too dangerous to open-source Its a weak argument. The bad guys are still using it, you are just ensuring that only the well-funded groups have it.
- lolc 6y ago
  Funny how people worry about things like that. It won't be one group suddenly having a superintelligent computer. It is groups developing ever more intelligent computers as hardware and software get more powerful. There won't be a moment before and after the invention of the AI. There is the creeping offloading of responsibilities to more and more automated systems we understand less and less. Already now these systems take decisions that we can only understand in hindsight if at all. We cannot approach them in a deterministic way like we're used to. These guys are just full of themselves to think they will have it and others won't. And also that they will be able to control it.
- Nasrudith 6y ago
  What is utter nonsense is both the understanding of nuclear physics and the understanding of actual machine learning as opposed to the Deus Est Machina nonsense. Hemisphere detonation with model T level complexity seriously? Even if we actually had the capability to make a human level intelligent AI that was trivially able to be commaned to go genocidal, and it fits in an average desktop PC it would not be a good comparison to nuclear weapons at all. It would lack any MAD and have far more applications than large radioactive explosions or large power generating reactors.

by _red 6y ago

FeepingCreature 6y ago

I don't think people are "falling" for it; you don't really see many people being confused that they can't find the GPT-3 repository. Personally, I just don't really care about the name - and I'm glad OpenAI is not open, purely on an AI safety basis. I think their original reasoning for openness was foolish, suicidal nonsense [1], and I'm glad they moved away from it. [1] https://slatestarcodex.com/2015/12/17/should-ai-be-open/ https://slatestarcodex.com/2015/12/17/should-ai-be-open/

_red 6y ago
>its too dangerous to open-source Its a weak argument. The bad guys are still using it, you are just ensuring that only the well-funded groups have it.
lolc 6y ago
Funny how people worry about things like that. It won't be one group suddenly having a superintelligent computer. It is groups developing ever more intelligent computers as hardware and software get more powerful. There won't be a moment before and after the invention of the AI. There is the creeping offloading of responsibilities to more and more automated systems we understand less and less. Already now these systems take decisions that we can only understand in hindsight if at all. We cannot approach them in a deterministic way like we're used to. These guys are just full of themselves to think they will have it and others won't. And also that they will be able to control it.
Nasrudith 6y ago
What is utter nonsense is both the understanding of nuclear physics and the understanding of actual machine learning as opposed to the Deus Est Machina nonsense. Hemisphere detonation with model T level complexity seriously? Even if we actually had the capability to make a human level intelligent AI that was trivially able to be commaned to go genocidal, and it fits in an average desktop PC it would not be a good comparison to nuclear weapons at all. It would lack any MAD and have far more applications than large radioactive explosions or large power generating reactors.