Responsible AI License

Live in a few minutes talking about this license:

Here is the 1-hr investigation of the license. I wanted to bring this up on the forum as well. After reading through the software license (we skipped the EULA), I decided not to suggest it as a potential license for HTM projects created by Numenta. You can watch the video to get my “in the moment” interpretation of this license, but I’m including a summary here:

This license, if applied as I read it, would prevent the development of AGI systems in many ways.

Restricts Learning About Humanity

The current “surveillance” and “Criminal” sections restrict the classification of aspects of humans, which is something AGI systems will be doing constantly as they interact and learn about humans. To restrict this would be to disallow these systems to learn about humanity and reality.

Restrictions should be applied now to human organizations

All these restrictions have good intentions, and are trying to make life better for disadvantaged people. If we as a society deem them valuable and worthy, they should be applied to existing human organizations first and foremost. This was particularly relevant in the “Heath” section.

Bias is already in the data

Bias is when given information does not match reality. It’s when internal reality doesn’t match external reality. The bias of intelligent agents (humans) exists in the data we collect about ourselves. Nothing in this license addresses this, and none of the restrictions in the license will prevent it.