March 22, 2013. Recently DARPA released a bulletin that was titled “DARPA envisions the future of machine learning”. They went on to explain what Machine Learning means in their view; “the ability of computers to understand data, manage results, and infer insights from uncertain information”.
It seems strange that something as basic as learning needs an explanation. Learning is certainly not unique to certain computer programs, and it is not the ability to understand data. We learn every minute of every day. In human terms, learning is the ability to acquire knowledge from the world around us through our senses. I described how biological learning works in my book “Higher Intelligence, how to create a Functional Artificial Brain”. Synaptic Time Dependent Plasticity ‘programs’ the synapses to match temporal-spatial pulse trains. Brains are much better at learning than any computer program. Compact intelligent machines can be created by copying the brain’s method of computing in a synthetic neuro-anatomy. Bayesian probability is an 18th century statistical method for determining the likeliness of a belief, which is calculated from evidence. Some people assume that this is how the brain works. However, the brain does not use data. Data is a description of knowledge, not the knowledge itself. Knowledge is such an intricate thing that a whole science is devoted to it. Epistemology is a branch of philosophy that seeks to explain knowledge. Knowledge is highly relational, and the relationships between objects are not logical and personal. The smell of a BBQ for instance makes one person hungry, while it repels another.
Knowledge is formed through learning. Learning is the prime function of the brain. Any neuron model that does not incorporate learning is therefore flawed. Computer scientists have been trying to reinvent intelligence since 1943, with the result that the word has been given a new meaning, very different from the meaning of human intelligence.
The Synthetic Anatomy chip faithfully represents the functions of synapses, glia and neural cells in a digital form. Like the brain, the chip is not programmed, but learns. It does not have an address bus, nor a memory module. Its data bus is thousands of bits wide. It is a building block for a synthetic brain. To understand how it works it is necessary to understand the brain. That is why the book spends much time explaining how the brain accomplishes its cognitive functions. Synthetic neuro-anatomy can be applied in everything from toys, to robotics.