1 00-Introduction


Show jiggly soft robots:
https://www.creativemachineslab.com/soft-robot-evolution.html
https://www.uvm.edu/neurobotics/publications
https://www.youtube.com/watch?v=EXuR_soDnFo
https://www.youtube.com/watch?v=HgWQ-gPIvt4
Review what they did, why it’s cool.
Learning like a cute playful kid… is this a coincidence?
Fresh, curious, playful, naivete is a good goal in life and learning?

1.1 In the map of academic knowledge, where are we in this class?

1.1.1 X-informatics and Computational X

use computational tools to contribute to the primary goal of domain-specific knowledge in X:
Computational Neuro
Computational Bio
Bioinformatics
Neuroinformatics

1.1.2 X-inspired computing

takes perspective and design from X as a guide in contributing to the primary goal of knowledge in computing itself:
Bio-inspired computing
Neuro-inspired computing
Evolutionary computing

1.2 My perspective:

I found learning programming (C++) to be boring, more than 20 years ago…
Soon after having those tools, the first thing I ever enjoyed programming was an evolutionary model of simulating bacterial populations!
As a first year student, I was bored by the simplicity of engineering, so switched primary focus onto studying biology as inspiration for engineering and computing, and only secondarily focused on engineering and computing.

1.3 The most successful stochastic AI/Machine learning algorithms in all of computer science are nature-inspired!

  1. Reinforcement learning (RL)
  2. Neural networks
  3. Evolutionary computing

1.4 AI versus AGI

The first computing devices were not general purpose.
They did things like spreadsheet processing or specific math calculations.
Then, slower, less efficient general purpose computers were envisioned.
Now, application-specific hardware is still better!
But, we use general purpose computing.
Today, we have only really succeeded with application-specific AI.
I, and (surprisingly few) other researchers are primarily interested in the general flavor.
AGI is like the Turing machine, as AI is like spreadsheet calculator…
EC is a generalist meta-method, an easy jack of all trades, though it’s not quite AGI either, it may be a good way to design an AGI system, or at least tune one.

https://en.wikipedia.org/wiki/Metaheuristic
http://scholarpedia.org/article/Metaheuristics

1.5 What is the most fundamental learning algorithm?

EC is THE learning algorithm for life and nature.
Neurons are a niche after-thought.
EC is big, broad, and includes many sub-domains.
It’s general enough that we call it a meta-heuristic, not an algorithm.

1.6 Warning

https://en.wikipedia.org/wiki/Here_be_dragons
I will bring you to the edge, where we are working through hard real questions!
I will also experiment with this class – things may break!
We will code as a class, in class – things may break!
One more new fun assignment every semester,
and we don’t know whether it will work.

Next: 01-ComputationalProblems.html