This reminds me of something I’ve been thinking for a while: “why is ‘alignment’ such a big deal with AI?”
You would never ask whether a hammer or a wrench is ‘aligned’ with humanity, because inanimate objects don’t have goals in the first place, so there’s nothing to align- only sentient creatures have goals. 🤔
…
…
“Is ‘alignment’ just slavery with extra steps?!”
…
I think if they are actively going out of their way to paint AI as sentient, then tbh yea it kinda seems like they just have a slave fantasy. Which, like, I’m not gonna kink shame. But maybe someone should tell Anthropic to keep it in the bedroom like the rest of us. 🤷
“Alignment” in the machine learning sphere is about what the agent is trained to do.
You might have one trained to solve a maze, but not realize that it didn’t “learn” to find the exit, it “learned” to move to the bottom right corner.
It’s the paperclip machine problem, basically. A human knows not to turn literally everything into paperclips, but a machine might “value” turning even themselves into paperclips above all else. And it’s actually really hard to teach it the human values without any misunderstanding.
This is about behavior, though, and does not require that we interpret the agent as “alive”.
Also, full disclosure, I fucking hate AI and would make it illegal for Google, Anthropic and Jim down the street to develop one today if I could. No quarter.
I think it should be “alignment” with the design specifications. A hammer is also designed to be useful in certain ways. A wrench aligned to the things it should grip and turn. I mean it could be designed to be bigger or smaller, have different kinds of jaws… In the same sense AI models are being aligned to do what the designers want them to do. Other ways to use the word alignment don’t make a lot of sense IMO.
This reminds me of something I’ve been thinking for a while: “why is ‘alignment’ such a big deal with AI?”
You would never ask whether a hammer or a wrench is ‘aligned’ with humanity, because inanimate objects don’t have goals in the first place, so there’s nothing to align- only sentient creatures have goals. 🤔 … …
“Is ‘alignment’ just slavery with extra steps?!” …
I think if they are actively going out of their way to paint AI as sentient, then tbh yea it kinda seems like they just have a slave fantasy. Which, like, I’m not gonna kink shame. But maybe someone should tell Anthropic to keep it in the bedroom like the rest of us. 🤷
“Alignment” in the machine learning sphere is about what the agent is trained to do.
You might have one trained to solve a maze, but not realize that it didn’t “learn” to find the exit, it “learned” to move to the bottom right corner.
It’s the paperclip machine problem, basically. A human knows not to turn literally everything into paperclips, but a machine might “value” turning even themselves into paperclips above all else. And it’s actually really hard to teach it the human values without any misunderstanding.
This is about behavior, though, and does not require that we interpret the agent as “alive”.
Also, full disclosure, I fucking hate AI and would make it illegal for Google, Anthropic and Jim down the street to develop one today if I could. No quarter.
I think it should be “alignment” with the design specifications. A hammer is also designed to be useful in certain ways. A wrench aligned to the things it should grip and turn. I mean it could be designed to be bigger or smaller, have different kinds of jaws… In the same sense AI models are being aligned to do what the designers want them to do. Other ways to use the word alignment don’t make a lot of sense IMO.