Assignment #16
Provide a short discussion of each of the assigned papers (listed under Course Materials). Below are questions to think about (you should discuss at least some of these but
you don't have to address them all):
Evolving 3D Morphology and Behavior by Competition
Questions
- Sims evaluates different design choices based on subjective judgment (how "interesting" is the resulting behavior). Can you think of more objective metrics? How could we decide if one design choice (e.g., tournament pairings) is better than another (e.g., random pairings)?
Hide and Seek
Questions
- The paper notes that randomization of the environment, during training, is helpful. Why do you think this is?
- Most of the time in RL, we have a specific task in mind, and we directly train agents to solve that task. A goal of emergent intelligence, in contrast, is for agents to
develop skills that surprise us, solving tasks that were not directly given to them by human designers. How successful is this paper in that respect? In what ways
did the environment design shape which kinds of skills might emerge?
- What do you think would be a good competitive game that would lead to the emergence of more general skills or intelligence?
Lenia
This is a long paper; try to understand how the simulation dynamics work (section 2), and then feel free to skim the many pages of results that show all the different kinds of "life" that emerge.
Questions
- Lenia is a rather bottom-up way of understanding life/intelligence: it defines the basic dynamics of the environment and then looks at what abilities emerge. This is in contrast to the more
top-down approach that most of AI is devoted to: define a task or objective and optimize to solve it. What are some advantages/disadvantages of the bottom-up approach
compared to the top-down approach. How could you add top-down signals to Lenia (e.g., how could you optimize the lifeforms to solve a task)?
General questions
- Compare and contrast the methods in today's readings with the previous papers we read on curiosity and novelty search. How are they related?
- In light of the discussion in class on April 9th: are these algorithms creative? What is missing in order for them to develop solutions or skills that we would call truly creative?
Upload a single PDF file through Canvas by
April 11th at 1 pm.