This week I finally have a rough first draft which is mostly a detailed look and summary of Spaun and Deep Mind. Writing this helped me come up with a thesis, and after a meeting with Justin, my thesis seems more clear.
My original thesis that I presented to Justin was focused on how Spaun under-performs when compared to the brain while Deep Mind out-performs the brain. Spaun can only recognize 94% of numbers from 0-9 and write the out, while Deep Mind constantly updates its neurons to make better predictions than the human brain can. However, since I was basing the performance of Spaun on just those percentages and determined it under-performs the brain, Justin pointed out that the reasoning may seem a bit thin. Therefore, we decided to change the thesis to focus on how Spaun under-performs the brain and even Deep Mind based on the restrictions and constraints used to build Spaun.
Since Spaun is designed to mimic the brain, its neural networks reflects certain regions of the brain such as the hippocampus, etc. However, certain restrictions cause it to not perform as well as the human brain can. An example of this is that Spaun has 40 times less neurons than the brain and does not have neural networks for every part of the brain. Besides pointing out certain restrictions and constraints that seem to limit the ability of Spaun, I will use the Q-learning algorithm as a foil or benchmark since it allows Deep Mind to perform superior to the human brain.
In terms of timeline, I aim to have a final edit of my paper by the end of next week. I will finalize my thesis and further my draft by Tuesday so that I can have a poster ready to be turned in by the end of Tuesday.
- Unknown Type