[Note from PP, March 28, 2017: The following article is a guest post by long-time blog commenter Animekitty / Illuminati Cat and does not necessarily reflect the views of Pumpkin Person. I did very little editing of this article, other than fix two typos and redact personal information I felt the very candid and honest author might regret sharing in retrospect. The picture was selected by me. 🙂 Out of respect for the author, please try to keep the comments on topic. Although I understand that conversations naturally evolve in new directions, please start on-topic.]
Hello, my name is Animekitty / Illuminati Cat.
I have been reading pumpkins blog for 2 years. I have learned a lot so far from him. [Personal details redacted by PP, March 28, 2017] He has said that the parallelism of the brain means that intelligence doubles every 5 IQ points. [Comment from PP, March 28, 2017: This theory is not my own, but based on a Promethean Society member’s observation that speed of complex learning and problem solving seems to double every 5-10 IQ points][More personal details redacted by PP, March 28, 2017] What I want to show in this guest post are my ideas I have regarding Artificial Intelligence. The reason I am interested in IQ is because I think that it is key to developing smart computers that could one day be our friends.
I have been studying intelligence since I was 12 years old when I first got interested in Artificial Intelligence while reading books at my middle school library. I went to a magnet middle school for science because I had written an essay about science in 5th grade. Every lunch period I would go to the library and read books on lasers and computers. I would draw diagrams of my ideas on my windows 95 computer. I saw characters on my Gameboy and thought how could I make them smart. Here is a diagram I made in 7th grade from all the books I read:
In 9th grade in high school I read this book:
Artificial Life: A Report from the Frontier Where Computers Meet Biology Paperback – July 27, 1993 by Steven Levy
I became fascinated with evolving systems and digital DNA.
Later on in 10th grade, I discovered that loops exist in the brain. Because of this I now understood that patterns could be stored and recognized by and arrangement of loops. I made this diagram:
Once I was in 11 grade I was put into and advanced computer class. In this class, we had to choose a project and I chose to do A.I. based on Jeff Hawkin’s book On Intelligence. The book said the brain was a hierarchy of pattern recognizers and that understood the world by predicting which patterns came next. The teacher in my class confused this with back propagation but a hierarchy is not just about error correction. It has to do with prediction and temporal patterns.
In back propagation, you have a picture, like a dog or cat and then you categorize it. This is completely a feed forward network. The picture is categorized and if a signal goes into the wrong category an error signal is propagated backward to get the signals into the right category. In the hierarchy system I was making, there were no categories. What it did was to predict signals with a compression algorithm that would send signals forward only when they became unpredictable. As the unpredictable signals moved forward they would be stored based on redundant temporal patterns. The program would also predict the consequences of its own actions when interacting with a teacher through reinforcement learning. This is my final report on my high School A.I. Project [report containing personal info redacted by PP, March 28, 2017].
After high school, I made little progress in my A.I. ideas up until recently. The main reason is that I did not understand how programming works and even in my high school project most of the code I did had nothing to do with higher level programming. My current idea is about recurrent neural networks and selective attention. Selective attention is the basis for learning in a parallel fashion. Recurrence is when output is feedback into the system. The results of your actions are fed back into your perceptions. And thinking is just the ability to feedback imagined perceptions. The decision process in the frontal lobes determines whether two ideas should be connected or not. We imagine if an idea fits with another idea just as in the real world we use tools that fit the task that needs to be done.
A main component of the brain is called the limbic system. This part of the brain gives us our motivations. It is what drives us forward to what we want and away from what we don’t want. The brain evolved to help us move in our environment and the limbic system is vital to that processes. Intelligence is at its core a way of learning how things move in the environment. Pumpkin calls it the mental ability to adapt. So there are two components to intelligence. Learning how other things move in patterns and then moving yourself based on your motivations. Imagination and creativity are the results of temporal patterns that the limbic system feels must go together. New ideas come together in the associative regions of the brain.
Recently I made a program in Java that rewires itself based on selective attention. A node will randomize what other nodes it is connected to if the sum of the signals it receives from those nodes does not match its main input signal. This is the motivation that a node wants its collective inputs to match its main input signal.
I know that I am probably not smart enough to create A.I. but I do want to understand how it will work someday. A person I know online who is 170 IQ said that he created A.I. in the 90’s but that he had no environment for it to learn and grow in. There are several brain models I have seen at online university lectures that demonstrate rudimentary intelligence. One of them is called BabyX that is a simulated baby. I think that because software is a hyperplastic medium, that future A.I. will most definitely have IQ’s exceding 170. Their virtual brains will not get tired nor need sleep. They will see patterns we cannot because the way their brain is connected it will coordinate selective attention much better than we can. Parallelism will make them smarter than humans