{image source: Pixar Wall-E}
Artificial Intelligence (AI) often feeds our cravings for a science-fiction. Flying cars, a trip to moon and back, self-adjusting autonomous homes to tune the lights, music, temperature, movie choices, lunch menu in the kitchen based on our mood of the day and what not.
However, the technology is not sufficiently developed yet such that it can make our life fully autonomous. In our day to day lives, its impact is still limited to merely watching our (consumer pattern) shopping behavior, observing market trends, promoting products, and performing silent vigilance on users.
AI is neither short of vision nor it is lacking economic support to get you all to that place where everything and anything is taken care of in our lives. So, why does it feels that technology is stagnated and not moving at same swift pace as AI marketing. Until AI is commonplace, it feels a far-fetched gadget from sci-fi movie.
AI aims to model the behavior of the human brain, which is known to process 11 million bits of information every second. Artificial intelligence, as the names implies, aims to artificially mimic the human intelligence. The computers that run Machine learning algorithms function as simple I/O systems and are very much one directional and deterministic.
Whereas the human nervous system is multi-way model. What does this mean? Take an example of touching a hot object. The automatic reflex is to take the hand away from heat source and this reflex arc functions as two way. First, sensory receptor cells detect the stimulus of high temperature. The stimulus is encoded by the joint firing of spikes (known as population encoding) by a network of sensory neurons in the dermis. Action potential is sent to relay neurons in the grey matter of the spinal cord. The impulse is not sent to the brain because the reaction would be too slow. Instead, the relay neurons transmit the impulse directly to motor neurons near the triceps and biceps. The activation of the motor neurons causes a quick muscle motion and is processed in the same neural network, where each output layer also is capable of processing as the input layer.
Furthermore, the brain operates without any central clock . Brain processes emerge and dissolve in parallel in different parts of the brain at different frequency bands: theta (5–8 Hz), alpha (9–12 Hz), beta (14–28 Hz) and gamma (40–80 Hz). It is lot slower than computing speed of machines. So Intelligence is not about how fast we can compute but rather it appears to be collective computation is what works in brain. Its like a system made of numerous small systems of intellects those work in sync.
Biological neuron model works in Feed forward propagation way only, there is no back propagation. Whereas in Artificial Neural networks, Back propagation (a learning algorithm used to train artificial neural networks by working backwards from output nodes to input nodes layer) is used to inductively optimize the weights in internal layers so as output could be more aligned to expected output, in Artificial Neural Network. This makes the Artificial intelligence models like a trial-and-error models.
In my view, it is important to upgrade the artificial neural networks models to function multi way as well as to behave a group of sub systems, more like a collective brain function.
Commentaires