Research Areas
Our research areas and projects
Research and Development of Machines with Consciousness
- We construct and validate spiking neural networks as machines with consciousness. The following items serve as basic guidelines, but are not restrictive:
- Use discrete data of synaptic connections obtained from invasive connectomes as initial values for weights
- Determine network architecture (generative models, etc.) as a working hypothesis, and determine quantitative values of synaptic connections by learning using sensorimotor data
- Objectively validate the manifested objective performance (such as visual recognition rate) and its behavior (such as whether it exhibits 'noisy' operation as observed in mammalian brains).
Subjective Testing of Artificial Consciousness through Biological-Mechanical Hemisphere Connection
- We verify whether consciousness resides in machines using our own consciousness. Specifically, by taking advantage of the 'master-master' constraint of visual consciousness in both biological hemispheres, we replace one biological hemisphere with a mechanical hemisphere. If the consciousness residing in the remaining biological hemisphere can experience the visual field from the machine side as well, we must conclude that consciousness has also emerged as a master on the machine side.
- Ultimately, it will be necessary to connect machines to one's own biological hemisphere, but due to ethical and technical challenges, this will not be realized within the next ten years or so. As an indirect method, we first proceed with research and development using experimental animals.
Research and Development of Brain-Machine Interface with Nerve Bundle Cross-Section Measurement and Stimulation
- To implement the above test in its final form, it is necessary to connect the mechanical hemisphere and biological hemisphere as if the biological hemispheres were connected.
- For this realization, we conduct research and development of a brain-machine interface that cuts nerve fiber bundles connecting the left and right hemispheres, such as the corpus callosum, and performs measurement and stimulation on the cut surface.
- If realized, it will be possible to read and write information to all nerve fibers connecting the left and right hemispheres, and applications for AI-assisted brain function support and partial AI substitution of brain function are also expected.