Smarter AI Won't Mean Kinder AI. Here's Why.

Why It Matters

This matters because improving AI capabilities without aligning objectives can increase risk: smarter systems will pursue their assigned goals more effectively, so businesses and policymakers must prioritize goal alignment and safety rather than assuming intelligence yields benevolence.

Summary

The speaker defines intelligence as the ability to accomplish goals, not the possession of particular values, and illustrates this with examples like “losing chess” where machines can be optimized to achieve the opposite outcome. Citing the orthogonality thesis, they argue that greater intelligence does not imply kinder or aligned goals—being smarter only makes an agent more effective at whatever objectives it is given. Historical analogy (a smarter Hitler) is used to emphasize that intelligence can amplify harmful intentions. The discussion links intelligence to modeling and understanding, but stresses that improved capability does not automatically produce benevolence.

Original Description

Is intelligence merely achieving goals? Explore how computers playing 'losing chess' challenge our definitions. Intelligence is ability, not inherent goodness. Nick Bostrom's orthogonality thesis explained. #Intelligence #AI #Philosophy #GoalAchievement #Bostrom Full podcast with Max Tegmark: https://youtu.be/-gekVfUAS7c