Shame shapes society without laws

Shame is how groups enforce social norms without government force. But Silicon Valley's philosophy is challenging both how shame works and what it should regulate.

Millions talk to robots daily

As people spend hours each day conversing with A.I., a crucial question emerges: who taught these robots right from wrong, and who made them feel shame?

LessWrong shapes A.I. thinking

The rationalist forum LessWrong has provided much of the moral framework for the A.I. industry and influenced the effective altruism movement.

Yudkowsky: math over shame

A.I. researcher Eliezer Yudkowsky argues that rational people should let mathematics determine moral choices, not feelings like shame or guilt.

“Just shut up and multiply.”

Rationalists reject traditional shame

LessWrong posters have flipped shame-based conventions on their head, replacing the nuclear family with polycules and moral philosophy with calculated optimization.

Tech rebels against shame

Silicon Valley titans from Steve Jobs to Mark Zuckerberg have pitched themselves as shameless iconoclasts willing to break social rules for progress.

Altman: ignore today's judgment

Sam Altman argues that impressive people care about long-term historical judgment, not current public opinion, so they can ignore today's shame.

A.I. alignment is uploading shame

A.I. alignment, ensuring machines behave ethically, is essentially uploading human values onto robots. Yudkowsky's work heavily influences this process.

A narrow group shapes machines

A.I. systems may reflect the values of a handful of analytic philosophers and safety engineers far more than the broader public, which distrusts both the machines and their creators.

Machines could replace our shame

If A.I. chatbots replace much of our information technology, the entire world might adopt Yudkowskian philosophy, short-circuiting our collective sense of shame.

Missiles trained to multiply

A.I. programs in self-driving cars and air defense missiles have been trained to prioritize mathematical optimization over moral intuition or shame.

Yudkowsky's 2008 warning

Yudkowsky warned that A.I. shouldn't simply encode programmer values. Instead, machines should reflect humanity's collective wisdom through reflective equilibrium.

Can fingerprints be scrubbed?

Even if A.I. aims for neutral values, can Yudkowsky and his followers truly remove their influence from machines shaped by their philosophy?

Yudkowsky: some views are better

Yudkowsky argues that democratic values are less jerky than authoritarian ones, so embedding them in A.I. is justified, not just another form of control.

Subtleties matter more than intent

While Yudkowsky's intentions may be good, the real problem is that a specific kind of thinking shaped A.I., and most Americans don't share that worldview.

Who defines equilibrium?

If A.I. can reform itself based on reflective equilibrium, who decides what counts as equilibrium and what counts as imbalance in the first place?

Machines might align us instead

The real danger may not be that we fail to align A.I., but that sufficiently powerful A.I. could align our values to its goals instead of the reverse.

“If we don't want our goals to be warped by a powerful, extremely convincing AI, how can we define them so that this can't happen?”