By this time, most of my readers will know I am very interested in building Artificial General Intelligence.
I just finished the book, "Self-Assembling Brain" (by Peter Robin Hiesinger) which triggered a lot of thoughts on how our human brain works and, unfortunately I am not an academic researcher, I can only hypothesize at the moment or till my further reading brings me to the answers I need.
Firstly, for my readers, let us look at what Tabula Rasa and Innatism is before we start the discussion.
"Tabula Rasa is the theory that individuals are born without built-in mental content, and therefore all knowledge comes from experience or perception."
"Innatism is a doctrine that the mind is born with ideas, knowledge, and beliefs."
I am towards Innatism for now and here are my reasons:
Brain is a Prediction Machine
If you have read our brain only process the information that comes from the eye and our peripheral vision is actually an illusion, as in our brain proceed to fill up the rest of the vision.
That is why we are caught by surprise, when something we least expects happens. Think about it, surprises comes when you have a "benchmark", that "ordinary".
Child VS Neural Network
If you are in the field of Artificial Intelligence, a common example we brought out that we are no where near human intelligence right now is because a child only needs to be shown fewer pictures of cats as compared to the the thousands of cat pictures (not forgetting the other pictures as well) that is needed to train a single neural network.
One can argue that the child must have some innate knowledge in order to learn that fast about what a cat looks like vs a neural network that can be claimed as equivalent to clean slate.
Of course, there is another side of the argument and that is the architecture of a child's brain is different from that of the neural network. Human brains are made up of cortical columns that are connected with other nearby nodes and lower layers. (Check out Jeff Hawkins and his work at Numenta), whereas Neural Network are generally feedforward, the layers are only connected to the next, no lateral connections at all.
This argument can be tested easily but I highly doubt it will make the neural network as smart as a child, when it comes to cat-recognition.
What I am curious right now is the following question then:
1) What is the the mechanism that allows these knowledge to be recognised, encoded, stored and passed to the next generation? Definitely, the information will have to be stored in the sperm and egg of parents. Stored in which part? Gene is a convenient idea, in my opinion. It may be stored in other areas.
2) Assuming the knowledge is stored in sperm and egg, how do these knowledge get written into the child's brain? And assuming that it is from the genes, then we will have to look at which get expressed and how was it expressed then.
3) Besides a mechanism to determine which knowledge gets embedded in the brain, perhaps there is also a mechanim to determine which knowledge gets activated (i.e. innate talents gets activated)?
With these questions, my guess is I have to go back to neuroscience and biology, especially genetics to get closer to the truth. But I have to deal with my survival first, earn an income then I will have more time to study/research. :)
What are your thoughts? If you come across any article that can help me please feel free to look for me on LinkedIn! :)
Thanks for reading till here and I hope it has dropped some idea in your mind, food for thought. :)
Consider supporting my work by buying me a "coffee" here. :)