Stefan,

Biases fall into all these categories. Certainly some biases were
useful in the ancestral environment, and even today.

The key difference  however is that an optical illusion is relatively easy
to recognize where a  cognitive bias is not.

I'm not so sure. I'd say that optical illusions can be hard to
recognize, and often reveal some visual biases. For example, we are
biased to assume we live in a three dimensional world with objects and
people moving around. Certainly a good assumption in the ancestral
environment, and true also today, But when we watch television, and
think we see objects and people, we are wrong; we are actually seeing
pixels on a 2d screen rapidly turning on and off. Even if we are not
fooled, our ability to even follow this electronic trickery
demonstrates the bias in the visual parts of the brain.

(Steven Pinker's  _How the Mind Works_ is a great popularization of all this.)

Ben Goertzel has mentioned this "3d space bias" as an example of a
useful one. It certainly help improve the efficiency of the AGI if it
does not have to figure out by itself that it lives in space-time.


What's wrong with being biased?

Some biases, however, are downright dangerous, especially those which
help us as we pursue our self-interest at the expense of others.

Studying bias is useful because we have to avoid passing on our
dangerous bias to our AGIs; because biases may be useful in improving
the efficiency of the initial AGI; because, as design quirks often
help us understand a system, biases may provide insight into how the
mind works;  and because AGI builders, like anyone else, must avoid
dangerous biases in their own thinking.

Joshua

-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e

Reply via email to