4. It may be possible to use purely trainable AI’s if we can created world model

4. It may be possible to use purely trainable AI’s if we can created world model and train against it, but at present my suspicion is that we will use AI for auto association as man does and programming (like we use laws, traditions, norms, manners, and habits) to make choices.


Source date (UTC): 2020-11-17 17:12:26 UTC

Original post: https://twitter.com/i/web/status/1328747840986894337

Reply addressees: @attyzachmiller @bierlingm

Replying to: https://twitter.com/i/web/status/1328747416724574214


IN REPLY TO:

Unknown author

@attyzachmiller @bierlingm 3. The model I propose is adversarial, in that any AI relies on two processes, one to auto-associate and calculate, and another to limit solutions (paths) to the moral. This mirrors predator/prey hemispheric division (bias) of labor in our brains, without limiting hypotheses.

Original post: https://x.com/i/web/status/1328747416724574214

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *