A post argues AI 'alignment' is incoherent and would be inadequate even if it were possible.
One post argues AI 'alignment' is infeasible and incoherent, and would be inadequate even if it were possible. It says replacing that framework should be a high priority for AI safety and links to what it calls a short summary by Jon Stokes.
15.7K
4 Sources, first seen ago
@carl_feynmanHe's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI safety. Not what I expected to be doing this morning.
@sebkrierRT @Meaningness: 🧵 AI "alignment" is infeasible, incoherent, and would be inadequate if it were possible. Replacing this unworkable concept…
@DanielleFongRT @carl_feynman: He's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI sa…@carl_feynmanHe's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI safety. Not what I expected to be doing this morning.@sebkrierRT @Meaningness: 🧵 AI "alignment" is infeasible, incoherent, and would be inadequate if it were possible. Replacing this unworkable concept…@DanielleFongRT @carl_feynman: He's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI sa…@carl_feynmanHe's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI safety. Not what I expected to be doing this morning.@sebkrierRT @Meaningness: 🧵 AI "alignment" is infeasible, incoherent, and would be inadequate if it were possible. Replacing this unworkable concept…@DanielleFongRT @carl_feynman: He's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI sa…@carl_feynmanHe's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI safety. Not what I expected to be doing this morning.@sebkrierRT @Meaningness: 🧵 AI "alignment" is infeasible, incoherent, and would be inadequate if it were possible. Replacing this unworkable concept…@DanielleFongRT @carl_feynman: He's right. I've been thinking about alignment wrong for decades. I have to rethink a lot of what I believe about AI sa…—
Not ranked yet
—
Not ranked yet
