I was reviewing today this really interesting piece of cognitive neuroscience research. Why Neuroscience? Well, when one looks at AI, there are two components to it. 1) The computer science part that is capable of intelligent behavior and 2) the application of the cognitive science or human brain part.
Paul Allen, a philanthropist, investor, and innovator, best known as the
co-founder of Microsoft alongside Bill Gates, donated $500 million of his own money and then made a $300 million pledge to the Allen Institute for Brain Science, making it his single largest philanthropic recipient; this is to identify which genes are expressed the most strongly, or not, in the human brain.
Whilst the above fact is interesting, the reason why I stopped and listened to the podcast was because there was an angle on AI Machine Learning used to identify and learn key brain signals that assess how related things are from one another, and therefore identify genetic (and other) clues that can affect our behavior patterns. Particularly knowing that in 2013, Allen moved the institute’s research into AI.
Bradley Voytek, the speaker, explains that his interest was to understand how the brain talks to itself to organize meaningful behaviors and how relevant the “brain” work, he is undertaking with his team and students, was to our everyday lives.
Listening to the podcast, it sounded like that the key cognitive problem that was solved, was that given a pattern of activation within the brain, one could capture what a person could be really doing or thinking using new AI methods. This means really new ways of addressing a lot of things from crime to buying behavior, to understanding basic emotions or reaction to medicines.
Still, such approach could enable researchers to avoid the interpretation and variability that often exist in their findings by individualizing the real key components that explain how things work, so that to come up with superior conclusions that can be extrapolated to other fields of work.
The next data-driven approach that we are taking, is taking huge amounts of brain data from large groups of people and seeing what kind of variability we get, how can we use machine learning to decompose the signals into their core components and then match that up with other databases to try and understand where these oscillations are coming from and what their roles are.