SM: Context is a key element of my Web 3.0 formula.
JJ: At the time, I did not know the importance of context. That came later at Stanford. I just knew that it what was in people’s minds. Since I could not get into people’s minds, the closest thing I could do is judge what was in their minds by their behavior. I realized that I was tapping into a discipline that I was not trained to be a part of, which is cognitive science. I did not want to build a company based on my intuition, so I knew I had to go somewhere to study this. I started on the computer science side looking for research in this area. There was very little of it.
One thing led to another, and I realized that the true solution lay on the other side of the campus in the cognitive side. I still maintained advisors on the computer science side, but I went to the communication department to an interactive media lab. I met a professor named Cliff Nass. He understood everything about human behavior that I could ever hope to understand, even though that is not his primary area of study. He gave be great insight on where I should focus and and what I should go after.
I then found tons of other research surrounding explicit behavior and dealing with understanding why people like things. It is effectively mind- reading and trying to figure out what you want. I repeatedly ran across research which indicated that context was king.
SM: Unconstrained AI is a very difficult problem. If you can constrain it within a context, then the vocabulary is limited and relationships between vocabularies are easier to realize.
JJ: I was also trying to understand, from a human biology perspective, why context is king. Why are people animals of context? Basically what you will find is that our brain structure is a contextual machine. Physical columns store individual information patterns. The brain is essentially a memory prediction machine. When you push a chair and see if fall, that is a cause and effect pattern. The brain stores that. A part of the brain will burn a scar and record that pattern. When you learn something new, the brain will emit chemicals and ask you to repeat that process. It wants to reinforce that impression.
Physically, biologically, neurons are burning through that tissue. You don’t want to burn it too many times because you will burn a hole and that would not be good. The brain then starts to emit chemicals saying “boring, boring” to protect itself. That forces you to learn other things. Whenever there is an exception, say you push the chair and it does not fall, then a neighboring area will record another behavior pattern.
Knowledge and intelligence occurs when that process is repeated millions and billions of times. The neurons that control the interest of Person A and Person B are unique. We all have our unique experiences and upbringings. Nobody in the world is alike. There is no way we can predict, holistically, what somebody will like or dislike. Even if you know their profile, they are still influenced by their current context. Peers can help understand context. If you have affinity with peers then biologically you have parts of the brain which are nearly identical.
SM: This is something I have been thinking about for a long time. Collaborative filtering, wisdom of the crowds if you will, misses the notion of ‘people who share my interest’; it has a very superficial understanding.
JJ: Collaborative filtering is overly simplistic. Just because you and I bought this device does not mean that everything you like is something that I will like as well. It seems to work well for books. Books have context, which is why Amazon does well with that piece. However, I bought a baby gift for my sister’s child and I still receive recommendations to buy diapers and things. They miss the mark there. They don’t get the context.
This segment is part 5 in the series : Simulating The Brain: Baynote CEO Jack Jia
1 2 3 4 5 6 7