Conclusion
- Robots fit in.
- Working in Nest and SpiNNaker simultaneously helps as one
system often helps understand the other. I expect using
BrainScales and using different neural models will also help.
- Take Home Points:
- We need to move from toy agents to reasonably sophisticated agents
that learn about their environments.
- We need a suite of agents and agent components.
- We need to consider how to put those components together.
- Agents need to persist for increasing lengths of time.
- We need to move to a more stochastic and robust approach (than
the binary CA approach I'm currently using).
- This will enable us to build agents that are beyond the cababilities
of current non-neural techniques.
- This will be a good step toward real AI.