Research Targets
1) Research and Development of Machine Consciousness
Develop a large-scale spiking neural network as a candidate system of machine consciousness.
Utilize discrete synaptic connections obtained from invasive connectome projects as initial weights.
As a working hypothesis, determine a system architecture (e.g. generative model) and train the network according to the loss functions defined by it.
Validate the final product using objective measures such as network function (e.g. visual recognition) and network dynamics (e.g. similarity to noisy neural dynamics observed in mammal brains).
2) A Test for Machine Consciousness: integration of biological and mechanical brain hemispheres
Since it is impossible to test machine consciousness through external observation or investigation of internal mechanisms using objective measures, only one method remains: making use of subjectivity.
We need to connect our own brains to the machine and “see” for ourselves whether consciousness resides within. By making use of the “master-master” configuration of visual consciousness in our two biological hemispheres, we replace one of our hemispheres with a mechanical hemisphere.
If we subjectively experience the machine visual field, we must conclude that a stream of visual consciousness has genuinely emerged in the mechanical hemisphere.
3) A Radically New Type of Brain-Machine Interface that Reads and Writes from Nerve Bundle Sections:
The key to the approach is a totally new type of BMI that enables reading and writing information with unprecedented precision.
The best part of the brain to tap into is the three neural fibres that connect the two cortical hemispheres, namely, the corpus callosum, the anterior commissure, and the posterior commissure. I proposes dissecting the three neural fibres and inserting a CMOS based, double-sided two-dimensional electrode array. Importantly, it would be coated with biological tissue to provide a target for he dissected axons to regenerate onto and firmly attach to the interface surface.
The objective is to read from and write into all of the axons required to integrate the two potentially independent streams of consciousness. Importantly, due to the critical problems suggested by Histed, Bonin and Reid (Neuron, 2009), regarding writing information with conventional electrodes placed in the grey matter, the proposed brain–machine interface is likely the only plausible method for the brain to sufficiently communicate with artificial devices.
