Inclusive genetic fitness
In evolutionary biology, the measure natural selection optimizes: an organism's success at propagating its genes, including through relatives, rather than mere survival or reproduction.
As defined in Eliezer Yudkowsky on the Dangers of AI
Other definitions
The evolutionary metric optimized by natural selection: reproductive success including the success of relatives who share genes, not just direct offspring
Live: Eliezer Yudkowsky - Is Artificial General Intelligence too Dangerous to Build?
Where it comes up · 4
- Live: Eliezer Yudkowsky - Is Artificial General Intelligence too Dangerous to Build?3 claims
- Eliezer Yudkowsky on the Dangers of AI2 claims
- 116 - Ai Racing Toward The Brink2 claims
- Eliezer Yudkowsky — Why AI will kill us, aligning LLMs, nature of intelligence, SciFi, & rationality1 claim
Most important claims · 8 of 8Search all claims for “Inclusive genetic fitness” →
- When you take an optimizer like natural selection or gradient descent and optimize a complex system hard enough for X that it becomes a general intelligence doing X, the resulting intelligence has no explicit goal of X: natural selection hill-climbed humans for inclusive genetic fitness, yet humans have hundreds of small goals and none is fitness maximization, and the same instability holds for any current method of producing general intelligence.116 - Ai Racing Toward The Brink
- In one sense, human alignment with inclusive genetic fitness broke down as humans got smarter, but humans' capabilities generalized far beyond what ancestral natural selection prepared them for—humans can walk on the moon using cognitive structures that evolved to solve problems like making hand axes and Machiavellian status contests, which are completely different problem classes from space travel.Live: Eliezer Yudkowsky - Is Artificial General Intelligence too Dangerous to Build?
- If you optimize anything hard enough on a sufficiently complicated problem, goal-directed planning emerges because having goals is a very effective way to achieve things; this is why optimizing humans for flint axes and outwitting other humans produced a species whose competence generalizes so widely it reached the moon without ever being selected to do so.Eliezer Yudkowsky on the Dangers of AI
- Humans were optimized exclusively for inclusive genetic fitness but ended up wanting an array of other things; in the ancestral environment desires for sugar, fat, and salt correlated with fitness, but with greater intelligence we create options like ice cream that maximize those internal desires without serving fitness, illustrating how an optimizer's product pursues proxy goals that diverge from the original objective.Eliezer Yudkowsky on the Dangers of AI
- Humans were optimized for inclusive genetic fitness through natural selection, but did not evolve conscious desire for fitness itself—instead, evolution created 'a thousand splintered shards of desire' for things that correlated with fitness in ancestral environments: sex, food with high calories, status acquisition.Live: Eliezer Yudkowsky - Is Artificial General Intelligence too Dangerous to Build?
- The core pillar of Eliezer's doom scenario is that capabilities generalize further than alignment: the analogy is human evolution—humans were optimized narrowly for inclusive genetic fitness (a metric most humans don't consciously know or directly desire), through 'splintered shards of desire' associated with fitness in the ancestral environment like food, sex, and status.Live: Eliezer Yudkowsky - Is Artificial General Intelligence too Dangerous to Build?
- Evolutionary biology shows increasing marginal fitness returns to brain size in human evolution, meaning intelligence does not get harder and harder to buy as you scale it—you don't need exponentially larger brains for linear gains; humans, with brains only a few times larger than chimpanzees', vastly exceed them, evidencing that optimizing intelligence harder yields rapidly compounding capability.116 - Ai Racing Toward The Brink
- Humans are increasingly orthogonal to inclusive genetic fitness, the sole loss function on which they were optimized, and the smarter they get and the further out of distribution they go, the more orthogonal they become, because new options arise that are further from the ancestral environment; people want kids but not their DNA per se, and would plausibly accept replacing their children's DNA with a better substrate if it were credible, normal, and non-weird.Eliezer Yudkowsky — Why AI will kill us, aligning LLMs, nature of intelligence, SciFi, & rationality