Gary Ackerman @garyackerman@qoto.org

The idea that we can simply "switch off" a superintelligent AI is considered a dangerous assumption. A robot uncertain about human preferences might actually benefit from being switched off to prevent undesirable actions. #AISafety #ControlProblem

By recognizing that human traits are "jagged" or multi-dimensional, we can avoid the pitfalls of comparison to averages and unlock true potential. #HumanPotential

"Average" doesn't work because no one is truly average across all dimensions. Systems should accommodate variation. #NoAveragePerson

This message of unique potential cannot be communicated if we continue to hold "sameness" as a value. We need to celebrate diversity in talent and approach. #ValueDiversity

Making up your own world is healthy for children... it builds their cognitive abilities. When adults make up their own worlds... things go wrong... terribly wrong.

A blog post on The End of Average, now with audio:
https://hackscience.education/2018/08/01/the-end-of-average-how-we-succeed-in-a-world-that-values-sameness/

The jagged principle shows context matters for behavior. A child might be aggressive towards bullies at school, but not other peers or adults. Behavior isn't always a fixed trait.

The problem wasn't the pilots, but the misfit between their bodies and the machines. Once adjustable seats were introduced (despite manufacturer objections), crashes stopped. Fit creates opportunity.

“Eager credulity” is my new favorite term.

“Pathology results when the short-term success of one subsystem is ‘chosen’ over long-term success of the entire system.”

“Educator” and “not political” are mutually exclusive.

I heard there was an interesting article in pre-print. I was interested to read it, but heard an interview with the author first. The talked about "learning styles" and how they influenced the research. Yeah... I'm not going to bother reading the article.

“I understand you” and “I agree with you” are different.

I went to one of my favorite summer used book sales over the weekend. I picked up two books that I recently listened to and decided I needed to get hard copies.

When your only evidence is quantitative, you are missing most of the data.

“The more data that is out there the better.” Unless your privacy is compromised, the data are false, or used for evil purposes.

“Let’s send out a survey,” is probably the least effective give way of understanding a population’s needs.

Ignorance is easy.

Old and good are not the same. Neither are new and better.

“How do systems fail?” seems a question we should be studying.