Develop Interesting Systems
- Use Human-like components (simulated neurons)
- Keep in touch with reality by
- Solve problems like humans do (cognitive models and architecture)
- Develop interesting systems (AI programs)
- Integrate Learning
- Take advantage of emergent algorithms
- Repeat
- Another key is to build functioning systems that aren't cognitive
models.
- These can be used for a range of AI behaviours like categorisation,
and modelling symbolic behaviours (like rules).
- It is also good to build more sophisticated systems like
agents.
- Our own CABot agents sit in a virtual world, view it, interact
with it, and have their own plans.
- To some extent, they start to address the symbol grounding problem.
- AIs are going to need to be situated in some environment to understand
us, so we might as well start putting them in environments now.
- They also provide us with direct evidence of things that are
needed by such systems.
- Improving these systems, extending their capabilities, and
putting them in new domains, gives:
- industrial applicability, enthusiasm, and more evidence about
what does and doesn't work.
