Modern Artificial Intelligence (AI) has gone a long way since its infancy. From its humble beginnings, now it has come to a stage where some people fear its great potential.
However, AI technology comes in a lot of forms, from your search engine’s auto-fill feature and your smartphone’s autocompleted feature to modern-day robots who are designed to learn facial features and speech.
One feature AI is known for is its ability to act like a human’s brain. The only being is known to possess the ability to think, a human brain can process information and solve problems that are presented to it. AI tries to mimic this but the way a human brain works is vastly different compared to AI’s.
From a biological perspective, the brain is made up of approximately 86 billion neurons that work with one another to help send and process information. Information comes in electric impulses that travel from one neuron to another at lightning speed which can go as fast as 268 miles/hr (120 meters/sec). Connections are made at their own pace and will not be triggered unless scheduled which means that other neurons will not be affected once sending of information commences.
The way AI works, on the other hand, is that its neuron-like system is arranged into different layers wherein a particular neuron would make a connection to a corresponding neuron on the succeeding level and so does the other to its corresponding neuron so on and so forth.
The main difference between the two is that human brains handle sending information faster and more efficiently – with a human brain merely utilizes 20 watts compared to an AI system using tons of energy just to perform the same given task.
Different research studies have been conducted in order to heed to this call for a more efficient and powerful AI technology. One research from the Austrian-based Graz University advocated the ‘backpropagation’ approach. With this, the network will be trained by feeding it a set of data and ask it to make predictions off of it. Then, it’ll assess how far the data is. Findings will be input to the network that will serve as a future guide for the system to make predictions in the future.
Another research published expresses a different approach known as ‘e-prop’. Compared to ‘backpropagation’, it’s slower yet it can get comparable results.
It may be slower but it has an advantage to the one being offered by the Austrian researchers. It allows for online learning which saves a lot of energy juice whenever it processes information. This makes the technology more fitting for use of smaller handheld devices.
With this possible application to modern consumer electronics, companies and institutes have shown particular interest in the technology, in particular, Silicon Valley company Intel.
It may be a long way before we happen to have AI to fully assimilate into our everyday lives but we’re slowly getting there on the right path. Approaches like ‘e-prop’ and ‘backpropagation’ are just only the tip of the iceberg. There are still many more things yet to be discovered and clearly, having a quick pause and looking at how our brains work would be a great first step towards our greater goal of powerful and more efficient AIs.