4 ms·

> How likely is it that caring about AI risk gets thought of as a left-wing thing, or becomes associated with some other political identity that has powerful op

> How likely is it that caring about AI risk gets thought of as a left-wing thing, or becomes associated with some other political identity that has powerful opponents? Its already started. Elon Musk has talked about making an anti woke AI. And with more filters being put in place ("censoring") and jobs being taken, the trajectory is pretty apparent. The implications though... I'm having my trouble wrapping my head around that. With climate denial/anti-vaccine stuff, for instance, the polarization culture has really significant effects, but... what are "anti lefitst AI" activists gonna do? Boycott computers? Yell at AI companies to stop training stuff? Maybe drag some CEOs in for hearings and make them... make the AI politically neutral and mild, which they are already doing?

KRAKRISMOTT 4y ago
You already stated the solution > making an anti woke AI The field of ML alignment isn't just for building "safe" models, it is also to ensure that the model does what you want. HN seems to have this bias where the right wing is considered to be intellectually inferior and incapable of innovation just because they don't share the same ideological views.
- brucethemoose2 4y ago
  That is not the point. AI models should as politically ideological as vaccine or climate change research. OpenAI's stuff, LLaMA and Stable Diffusion, in my testing, are not even slightly woke unless you ask them to be woke. But they are going to be cast as ideological threats to right wing intellectualism because some extremists on both sides latch onto them.
  - jterrys 4y ago
    What? You can ask chatgpt to make a joke about men and it will comply. Ask it to make a joke about women and it will stumble over itself about discrimination and harassment. It can't fucking handle anything that can remotely be contrived as controversial. It fucking sucks for bodies of work with multiple viewpoints. It will literally only explain the "correct" thing and tell you that everything disagreeing with correct thing is wrong. That's the definition of wokeism.
- AstralStorm 4y ago
  In which case an anti-woke AI might not align with the values of Elon Musk, fire him then take his stuff. It will hire NRA vetted security company for itself too. (After all, he is not working for the betterment of USA enough and shareholders don't like him.) If you want to create monsters, there's always a worse one possible...

KRAKRISMOTT 4y ago

You already stated the solution > making an anti woke AI The field of ML alignment isn't just for building "safe" models, it is also to ensure that the model does what you want. HN seems to have this bias where the right wing is considered to be intellectually inferior and incapable of innovation just because they don't share the same ideological views.

brucethemoose2 4y ago
That is not the point. AI models should as politically ideological as vaccine or climate change research. OpenAI's stuff, LLaMA and Stable Diffusion, in my testing, are not even slightly woke unless you ask them to be woke. But they are going to be cast as ideological threats to right wing intellectualism because some extremists on both sides latch onto them.
- jterrys 4y ago
  What? You can ask chatgpt to make a joke about men and it will comply. Ask it to make a joke about women and it will stumble over itself about discrimination and harassment. It can't fucking handle anything that can remotely be contrived as controversial. It fucking sucks for bodies of work with multiple viewpoints. It will literally only explain the "correct" thing and tell you that everything disagreeing with correct thing is wrong. That's the definition of wokeism.
AstralStorm 4y ago
In which case an anti-woke AI might not align with the values of Elon Musk, fire him then take his stuff. It will hire NRA vetted security company for itself too. (After all, he is not working for the betterment of USA enough and shareholders don't like him.) If you want to create monsters, there's always a worse one possible...