1. Evolutionary Algorithms and Artificial Intelligence Paul Grouchy PhD Candidate University of Toronto Institute for Aerospace Studies pgrouchy@gmail.com

2. Intro to Evolutionary Algorithms (EAs) Program flow of a Genetic Algorithm (GA): 1.Randomly initialize population of “genomes” 2.Evaluate “fitness” of all genomes 3.Select high-fitness genomes to become “parents” 4.Produce new population of “offspring” genomes from “parent” genomes 5.End of a single “generation”

3. Intro to Evolutionary Algorithms (EAs) Toy problem: Maximize the sum of 5 bits Genome: 01100 Fitness (sum of bits) 2

4. Intro to Evolutionary Algorithms (EAs) Toy problem: 1 generation 01100 010000100100010 fitness: 1 fitness: 2 fitness: 1

5. Intro to Evolutionary Algorithms (EAs) Toy problem: 1 generation 01100 00010 Parents 0111000000 Offspring mutation 01000 crossover point

6. evaluate fitness of each genome using fitness function select and reproduce parents based on fitness values Generation t Generation t+1 Mutation Crossover 01101 01001 01100 01011 01111 Intro to Evolutionary Algorithms (EAs)

8. Evolutionary Computation: A Unified Approach (2006) Kenneth De Jong

9. EAs as AIs http://boxcar2d.com/

10. https://xkcd.com/720/

11. EAs as AIs Eureqa (http://creativemachines.cornell.edu/eureqa) – Based on Genetic Programming (GP):

12. EAs as AIs Eureqa (http://creativemachines.cornell.edu/eureqa)

13. EAs as AIs http://www.gp-field-guide.org.uk/ (FREE!)

14. EAs are Embarrassingly Parallelizable

15. AI vs. AGI AI:

16. AI vs. AGI Artificial General Intelligence (AGI):

17. AI vs. AGI Artificial General Intelligence (AGI):

18. EAs to evolve AIs

19. evaluate fitness of each genome using fitness function select and reproduce parents based on fitness values Generation t Generation t+1

20. EAs to evolve AIs InputsOutputs

21. EAs to evolve AIs evaluate fitness of each genome using fitness function select and reproduce parents based on fitness values 0.321.10-0.21…0.11 =

22. EAs to evolve AIs

25. Learning and Generalizability [Urzelai & Floreano, 2001]

26. Learning and Generalizability [Soltoggio et al., 2007]

27. EAs to evolve AIs

28. Can we evolve an abstraction of a brain? 0D3v0 Ordinary Differential Equation Evolution

29. Learning Capabilities Simulation environmentTypical evolved forage path Typical evolved “eat” output

30. https://xkcd.com/534/

31. ALife/Evolution of Communication Sim (x,y)(x,y) c in (Δx,Δy)(Δx,Δy) c out

32. ALife/Evolution of Communication Sim

39. THANK YOU!!! Paul Grouchy PhD Candidate University of Toronto Institute for Aerospace Studies pgrouchy@gmail.com