We interviewed Dr. Yukiyasu Kamitani of Kyoto University’s Department of Intelligent Informatics in the Graduate School Informatics. He is a leading expert in the development of brain decoding.
Dr. Kamitani received his undergraduate degree from the University of Tokyo and his PhD from California Institute of Technology. He conducted research at Harvard Medical School, Princeton University and the Advanced Telecommunications Research Institute International (ATR) Brain Information Laboratory between 2001 and 2004. In 2015, he joined the Graduate School of Informatics at Kyoto University where he is now a faculty member.
First, What is Brain Decoding?
Brain signals can be regarded as code that represents our state of mind. The act of decoding those signals is called brain decoding.
Currently, the Kamitani Lab is using machine learning and other techniques to model the brain's information processing and information representation, and is conducting research on decoding visual images and dreams from the brain.
Video: Deep image reconstruction: Natural images
Motivation to Pursue Brain Decoding
Dr. Kamitani, now a pioneer in the development of brain decoding technology, began this line of research due to disappointment with methods used at the time.
I developed brain decoding when I was at Princeton in 2003-4. At that time, around 1990, fMRI came out and spread rapidly, but to be honest, there was no research that I thought was that great. The standard method at the time, SPM(GLM), was a very boring model, but everyone believed in it. I was so disappointed with the situation that I decided to do it myself, and started to rebuild the methodology from scratch. Brain signals also appear as patterns in images, and I think the natural way to analyze them is to decipher what is encoded in them.
When I was given new experimental methods, I would approach it with a playful mindset. With this approach, using GLM to "map the brain" was not an obvious thought. I began with an obvious approach; because brain signals are patterns, it would be a good idea to consider pattern recognition models. Although this approach had been around, it was not yet mainstream. *SPM:Statistical Parametric Mapping GLM:Generalized Linear Model
Challenges in Today’s Brain Decoding Research
Almost 20 years have passed since Dr. Kamitani first began brain decoding research. From his pioneering perspective, what are the prominent challenges in brain decoding research today?
The brain imaging and brain measurement techniques are currently not the bottleneck. I believe that we will unlock more with current imaging techniques by working on the analysis. That is not to say that advances in brain imaging will not improve brain decoding. If the resolution of MRI improves or if more electrodes can be implanted in the brain, it will allow more information to be decoded. However, much of the work in analysis remains to unlock this potential.
My hope (for future BCI) is to be able to measure the brain in a natural state. Having a implanted electrodes in the brain with wireless communication would be ideal.
The Kamitani Laboratory is currently conducting research on intracardiac imagery and dream decoding. The lab is focused on developing better decoding techniques with higher accuracy.
Communicating via the Brain Directly
In "Brain Science Guru 2017 "Brain Decoding: Technology for Reading the Mind from the Brain," he spoke about the realization of communicating directly through the brain using brain decoding.
Video: "Brain Science Guru 2017 "Brain Decoding: The Art of Reading the Mind from the Brain"
Is it currently possible to perform brain-to-brain telepathy? And what stage is the technology at now?
Currently, the BMIs are those that restore motor function in people with physical disabilities. Furthermore, for patients with paralysis, it is still likely that restoring the motor function of the arm to use a tablet is more user-friendly than directly manipulating the tablet with brain signals. There is still work to be done here.
I believe there are a ton of interesting basic technologies, and there is high potential in measurement and analysis techniques, but I think one of the biggest problems in BMI is that there are no applications that truly meet the needs of the market. I think that in a few years, Neuralink and other companies will be able to realize simple operations of prosthetic hands, such as moving a cursor or inputting text.
In fact, in the research I am doing with Osaka University, we have been studying the use of electrodes implanted in the head to move a prosthetic hand, but I myself do not think that the output of such simple motor commands is the true potential of the brain-machine interface.
Generally speaking, in order to provide a service or product, it is necessary to satisfy the needs of a third party. BMI/BCI is no exception, and because it is a developing technology, it is difficult and important to know "for whom" the technology will be used.
The Neuroverse
Dr. Kamitani sees potential in "neuroverse" as a need for BMI/BCI.
I think of the brain-machine interface in my mind as something like the Matrix or Inception, where the world inside the brain is taken out and shared. The world inside the brain, which I call the neuroverse, will eventually be intertwined with the metaverse, and sharing such a world with others is the direction I see in the long run. I am interested in creating the metaverse out of the world extracted from the brain.
Industrial Relations with the Kamitani Lab
Industrial applications are also important for the development and generalization of technology. Are there any plans to increase the involvement of industry in future research?
At one time I was preparing to start my own business, but the technology was still in its infancy, and I was disappointed with the VC industry, where money is made by mentioning the keyword "brain." I am now focusing more on basic research and art production. I am an advisor to a metaverse company called cluster, but I am not doing any applied research at the moment. Mr. Kato, the president of Cluster, and I have been talking about how we should aim for the future of the brain-machine interface, such as extracting the world from the brain, and how we should work toward that goal. Because this is not something that will immediately lead to industry, we will continue to do more basic research.
Brain Science in the Future
Dr. Kamitani has been at the forefront of brain science for several decades, but what are his thoughts on the future of brain science?
It is becoming increasingly clear that much of the research that was touted when I was a student was not good at all. The poor reproducibility of psychological research is well known, but I believe that cognitive neuroscience research is just as bad or worse. Brain imaging research has been spectacularly self-promoting, but there are not many reproducible results.
We have turned a blind eye to the fact that science was being conducted with loose discipline, saying that it was a young field, but I think the current situation is that the reality is gradually being exposed. So, the challenge for the future is how to rebuild that and make it into a mature science.
Conclusion
In this interview, we asked Dr. Kamitani of Kyoto University about the impetus for the development of brain decoding and the future of brain science research. The neurotech industry is gradually gaining momentum, but we felt that it is important to have a stricter attitude toward science and not be misled by ideals and deceptions. We would like to pay attention to the future growth of decoding technology.
