Artificial Intelligence in the future; For years, Jeff Hawkins has been saying that brain science will change computing and the way we think about Artificial Intelligence (AI). However, he advances that the vision of intelligent robot-humanoids interacting with us, as we see it in science fiction, will not happen, at least not anytime soon or for practical purposes.
Hawkins published his new book “A Thousand Brains: A New Theory of Intelligence,” in which he proposes a revolution in the understanding of intelligence, but also a disruption in the development of Artificial Intelligence.
Recently, Dr. Sabrina Neuman, a graduate in Computer Science and AI from the Massachusetts Institute of Technology, warned that, despite the fact that robots were becoming stronger, the challenge was to make them more efficient in detecting the environment through sensors, and capable of “making decisions”, so she proposed a new custom hardware design that could save time in data processing, which would mean improving the performance of existing and future robots. This breakthrough, which will be presented in the coming weeks, shows the current bifurcation regarding the development of AI by scientists.
Hawkins believes that the future of AI must be based on understanding how the brain works. He is the founder of Palm Computing, established the Redwood Center for Theoretical Neuroscience, and runs the neuroscience and AI research company, Numenta, which is based in Silicon Valley. In his many presentations he accepts that most people know him from his work at Palm and Nadspring, where he developed different products including the personal computer, but he highlights his work in theoretical neuroscience and the study of the functioning of the neocortex.
He mentions that his interest as a neuroscientist arose after reading an article in the Scientific American magazine, dedicated to the brain, where topics about neurons, their development, diseases, vision, among others, were addressed. In this sense, and for more than a decade, Hawkins raised the need to have a theory about the brain.
In 2003, during a TED presentation, he explained that, in order to have access to knowledge about intelligence, it would first be necessary to define what an intelligent system is, and to know how the brain works based on a theory. He argued that everything is complex until it is understood: “We are our brains. My brain talks to your brain. Our bodies are like passengers, and if we want to understand who we are and how we feel and perceive, we really have to understand what brains are.”
To reinforce the need for such a theory, he even exemplified with the heliocentric theory of the solar system developed by Copernicus, which at the time faced strong resistance and yet now we do not even question it.
He warned that science is based on theory and experiments, and in neuroscience we have a lot of data regarding anatomy, physiology, behavior, but no theory to begin to analyze them. For this reason, he developed one that now translates into the proposal of four minimum attributes of intelligence:
The first, learning by moving to build a mental model of things. It is argued that one cannot sense everything around us at once; therefore, we have to move to build a mental model of things, which he calls “embodiment.”
Second, there are tens of thousands of cortical columns that capture this sensory information with a partial picture of the world to, together, generate an overall view. From this arises the idea of a thousand brains, which in an AI system would imply that a machine controls different sensors (vision, touch, radar, etc.) in order to obtain a more complete model of the world.
Third, continuous learning where you master new things without forgetting previous ones; it is clarified that in current AI systems this cannot be done.
Fourth, frames of reference, which means that knowledge of the world is relative to the point of view.
Hawkins accepts that his lab has already moved from neuroscience to focus on Artificial Intelligence development, based on brain function, where he claims to have very significant advances involving up to 50x acceleration in existing networks, as well as more robust and less power-hungry networks.
He is optimistic, but sees it as unlikely that the first things to be developed with this new AI will be classic robots; intelligent machines will not be humanoids like those we have seen so far in science fiction, but perhaps truly intelligent vehicles that understand what traffic is and what driving is; security systems, where the brain is used as a response, but not much mechanics.
Moreover, he believes that many intelligent machines in the future will not do anything like what humans do now, and even argues that eventually the interest in generating a human-like machine will be lost.
Hawkins accepts that he does not know how all this scientific progress will end, but shares that he knows of many people who invented the microprocessor and that, although they knew that what they were doing was very significant, they did not know what was going to happen with that invention. That is, at that time they could not anticipate the internet, smartphones, and all those tasks that are now so common in our daily lives as a result of such technology.
Therefore, he reiterates that his neuroscience-based AI theory will bring incredible changes in the next century that will benefit society as a whole, since it will help preserve knowledge: “We’re not going to be here forever, but our machines could be. It’s a way of essentially preserving us for a time and place we don’t know yet. The world is our limit.”
Thus, Hawkins’ theory opens the debate around AI.
¡Stay tune! Go and visit our blog. New articles every week.