Looking backward, looking forward.

• We are in the final push to the end of the semester now.
• There are just four lectures with content and one in-class review session left.

Looking backward, looking forward.

• We may look back on this period as an inflection point in the history of education for two reasons:

1. Post-COVID return to in-person teaching.
2. Beginning the A.I. Conversation.

• So I want to start the talk today by looking back at what I think is important.

First, Simplicity.

A topic that we have not covered explicitly in this course.

1. Simple systems, that you can clearly introspect and explain are critical for decision making.
2. Complexity is not a virtue, but it is a good way to hide from responsibility!
3. Our task is to find simple expression of complex systems.

• Note that simplicity need not be reductive; the sufficient number of dimensions of controls for a train is one, two for a car, three for an airplane.

Second, Transparency.

1. Being able to show your work is, at this stage, a form of communication. To your future self and to others who may consume your analyses after the fact.
2. If you understand the system you will have some understanding of what happens when you change the inputs.
3. Mistrust opaque systems (this is my most devastating critique of AI) understanding why a decision process came to the conclusion that it did is as important as understanding the conclusion.
4. If you want to find a bad decision, just look for the ones that are least well understood.

Third, Doubt.

This we have talked about at every turn.

Always ask:

1. how is the decision process tricking us?
2. How are the data tricking us?
3. What is the real goal, what are the data trying to tell us?
4. What is the goal of the people who made the data?

Third, Doubt.

• Are we inheriting the system that gives us this answer form another process that has different needs and thus might trick us?
• This is the point of the current discussion of decision making.

Fourth, The answers are easy, it’s the questions that are hard.

• You’ve been trained to answer questions, but that is the easy part, the part that can be automated.
• Programming easy things is often about as hard as programming hard things. Remember systems of equations in Python!
• Deciding what information to use and how to structure your question has always been the most important issue. Even more so now, as we can often trivially solve any well structured problem.

Fourth, The answers are easy, it’s the questions that are hard.

• We have inherited an education system from a world where computers were people (and not nearly as well paid as they should have been).
• We (educators) used ability to solve complex problems as a measure of ability.
• You will need to unlearn complexity, this is something that we all get trained for incorrectly.

The minds that did that work and that will do the work you need to do are not different, but the training, and habits of mind that that you need in order to succeed at it are very different.