[Please read the OP before voting. Special voting rules apply.]

Improving the typical human's emotional state — e.g. increasing compassion and reducing anxiety — is at least as significant to mitigating existential risks as improving the typical human's rationality.

The same is true for unusually intelligent and capable humans.

For that matter, unusually intelligent and capable humans who hate or fear most of humanity, or simply don't care about others, are unusually likely to break the world.

(Of course, there are cases where failures of rationality and failures of compassion coincide — the fundamental attribution error, for instance. It seems to me that attacking these problems from both System 1 and System 2 will be more effective than either approach alone.)