4 ms·

Much as how Kolmogorov solved the problem of defining randomness, I feel like Legg and Hutter solved (or at least mostly solved) the problem of defining univers

Much as how Kolmogorov solved the problem of defining randomness, I feel like Legg and Hutter solved (or at least mostly solved) the problem of defining universal intelligence. Loosely put, it is an agent that makes the best possible decision about the future with the information that it has available. “Best” can be defined however you want; it’s less about the particular cost function and more about achieving optimality. And no, the oft-cited yet frequently misunderstood NFL doesn’t rule this out.

visarga 4y ago
> makes the best possible decision about the future with the information that it has available So would a good thermostat be considered intelligent, assuming it takes great decisions? Intelligence needs to be defined with respect to a range of tasks and a range of priors. It's skill acquisition efficiency, not just taking good decisions. I recommend "On the Measure of Intelligence" by François Chollet >We argue that solely measuring skill at any given task falls short of measuring intelligence, because skill is heavily modulated by prior knowledge and experience: unlimited priors or unlimited training data allow experimenters to "buy" arbitrary levels of skills for a system, in a way that masks the system's own generalization power. We then articulate a new formal definition of intelligence based on Algorithmic Information Theory, describing intelligence as skill acquisition efficiency and highlighting the concepts of scope, generalization difficulty, priors, and experience. https://arxiv.org/abs/1911.01547 https://arxiv.org/abs/1911.01547
- Xcelerate 4y ago
  > Intelligence needs to be defined with respect to a range of tasks and a range of priors. Yep, you’re right. I forgot to mention that in my “loosely put” sentence, although the work by Hutter and Legg does mention that and discusses algorithmic information theory heavily.

visarga 4y ago

> makes the best possible decision about the future with the information that it has available So would a good thermostat be considered intelligent, assuming it takes great decisions? Intelligence needs to be defined with respect to a range of tasks and a range of priors. It's skill acquisition efficiency, not just taking good decisions. I recommend "On the Measure of Intelligence" by François Chollet >We argue that solely measuring skill at any given task falls short of measuring intelligence, because skill is heavily modulated by prior knowledge and experience: unlimited priors or unlimited training data allow experimenters to "buy" arbitrary levels of skills for a system, in a way that masks the system's own generalization power. We then articulate a new formal definition of intelligence based on Algorithmic Information Theory, describing intelligence as skill acquisition efficiency and highlighting the concepts of scope, generalization difficulty, priors, and experience. https://arxiv.org/abs/1911.01547 https://arxiv.org/abs/1911.01547

Xcelerate 4y ago
> Intelligence needs to be defined with respect to a range of tasks and a range of priors. Yep, you’re right. I forgot to mention that in my “loosely put” sentence, although the work by Hutter and Legg does mention that and discusses algorithmic information theory heavily.