You all have probably heard (many times) about my work with simulated neurons. We working with both the Human Brain Project's systems, and our own simulator.
We'd like to build more capable systems.
We'd like to build systems that are more biologically accurate. This means the neurons are like human neurons, and the topology is like the human topology. We're comfortable with animal like topologies too; we're making the assumption that mammalian neurons are like human neurons.
To build larger systems, it really simplifies things if the system learns.
To build realy human like systems, they're going to have to learn.
This talk will give you a flavour of that, particularly focusing on learning categories.
This extends the bridges over the gap between learning and cognitive modelling, all done with simulated neurons.