I sometimes wonder how good the AIs will be at predicting our productivity and our future courses of action.
Let’s say an advanced AI has your genome, a bunch of your test scores, and has plenty of video of you interviewing. How well does the AI really understand you?
To be clear, I am not asking about the capabilities of the AI, rather I am querying about human legibility. And my intuition is that the AI still will be surprised by you pretty often. It will not know who is “the next Einstein.”
Some of the freedom you retain may — perhaps counterintuitively — come from your genome. For purposes of argument, consider the speculative assumption that rare copy variants are important in genetics, and thus in your individuality. In that case, the AI likely cannot get enough data to have a very good read on what your genes imply. Even if the AI has everybody’s genome (unlikely), perhaps there just are not many people around with your rare copy variants.
It may also be the case — again speculatively — that rare copy variants are especially important for “top performers” (and mass murderers?).
So when the AIs come to scan you and evaluate you, perhaps it is the very genetic component that will protect you from high predictability. Of course that scenario is the opposite of what you usually read. In standard accounts, your genes make you a kind of captive or epistemic prisoner of the AI, a’la Gattaca.
But in practice you still might feel quite free and undetermined, even after receiving the report from the AI. And it might be your genes you need to thank for that.
The post Is it genetics that will give you freedom from the AIs? appeared first on Marginal REVOLUTION.
