Quote · The Prof G Pod with Scott Galloway
Sam Harris on The Democrats’ Far-Left Problem
Where this was said
AI: The Alignment Problem and the Arms Race We Can't Stop
At 59:30 · chapter starts 58:50
An audience question about AI prompts Harris's most sobering analysis of the episode. He frames himself as genuinely worried rather than casually pessimistic: the concern he and others have held for a decade has arrived precisely as feared, and the arms race logic makes it structurally impossible to take the care required [1] — Sam Harris "In the space of all possible superintelligences, most will be misaligned with human wellbeing — and nobody currently knows how to build a d…" 59:00 . Anthropic races OpenAI domestically; the US races China internationally. In that environment, no single actor can unilaterally slow down without ceding competitive ground. The deeper problem is the alignment puzzle itself: once we build AI that recursively self-improves — building the next generation without human intervention — the question is whether we've engineered the explosion in a way that keeps it aligned with human survival, let alone human flourishing. Harris's probability assessment is bleak: in the space of all possible superintelligences, one would expect most to be less than perfectly aligned with human wellbeing. He cites Geoffrey Hinton [2] — Sam Harris "There are not many examples of a less intelligent species living in proximity to a more intelligent species and managing to control it." 1:00:58 : there are almost no examples in nature of a less intelligent species successfully controlling a more intelligent one. The hosts are, in Harris's framing, 'building alien minds with which we will be in relationship' — and imagining you can just turn them off is a failure to take true general intelligence seriously.
In the space of all possible superintelligences, most will be misaligned with human wellbeing — and nobody currently knows how to build a durably aligned one. The race between OpenAI, Anthropic, and China makes taking the necessary care structurally impossible.
