The Future Of Intelligence
The entire concept of Artificial Intelligence (AI) has its basis on the workings of the human brain. The human brain is a wonder. Humans have been trying to understand its workings for centuries. Moreover, since the conception of computational tasks, humans have been trying to dabble in the field of AI. On Intelligence by Jeff Hawkins and Sandra Blakeslee delve into the workings of the human brain, its ability to compare new experiences to old memories, why machines are still unable to mimic this capability, and what lies in store for intelligent machines in the future.
More Power Doesn’t Translate To More Intelligent
As smaller computers became more powerful than their huge ancestors, scientists and researchers have been dreaming about a computer that can think like the human brain. Yet, humanity is miles away from developing machines as intelligent and capable of creative thought and understanding, as the human brain is. This is due to the fact that computers and the human brain are based on totally different principles.
Computers are made for the storage of information. This information programs a computer to perform certain tasks based on the command or code it is given. Fundamentally, a computer cannot use the stored information later on and use it to understand and process other information, and neither can it learn new information.
The human brain, however, has the ability to learn, apply previous experiences to new ones, understand and apply the information it acquires, via observation. This is what makes it intelligent.
When the supercomputer Deep Blue beat the world chess champion, Garry Kasparov, it was able to do so merely due to the fact that it was able to calculate the probabilities of winning by running the numbers of each and every move and counter-move, and not because it was more intelligent than Kasparov.
Increasing power, therefore, doesn’t translate into being more intelligent, because more power will only increase a computer’s ability to compute faster. Computers will never be able to think as the human brain does. For that, a computer will have to have the ability to understand how the human brain works.
The human brain processes information brilliantly. It combines the information received by the senses with existing memories, making humans capable of experiencing their surroundings.
The neocortex is responsible for sensory perception and conscious thought processes. It is made of a number of layers that combine and relate new information to stored memories. The new raw sensory information passes through these layers, to which previously experienced details are added.
For example, when one sees the face of a school friend, the eyes pass on this visual information to the brain. The lower layers of the neocortex further pass the information to a higher-layer that combines stored memory information, links it to the new visual, and thus helps us recognize the friend. This process is so fast that it enables us to experience our surroundings seamlessly and fluidly.
When the information sent to the brain is absolutely new and the neocortex has no data to reference it with, the information is sent to the higher layer, where it is stored as a new memory. Thus the brain builds its ever-growing database of memories.
Connecting The Past And Future
How do we know for a fact that when we turn the knob on a door, it will open the door? How do we, so easily, predict this future?
The brain stores different memories in different regions that get activated in a sequence or a pattern, when it receives familiar information. If we take listening to a favorite song as an example, one region of the brain recognizes the music notes from a previous memory, while another region references previously-stored memories of the lyrics, and yet another region links the two. Thus we recognize the song every time we hear it.
These sequences and patterns essentially help us predict the future. Every time we experience something, the brain combs through different regions, looking for similar memories, activating the same nerve cells that were activated then. In this manner, the brain can dig up former reactions and reactions after that, enabling us to know what reactions to expect this time around. For example, the brain is able to predict that cars begin to move when the signal lights turn green, purely based on past experiences and stored memory.
The human brain therefore, makes predictions and adapts those predictions every time there is a new experience. This is the process of learning.
The Brains Unique Complexity
After scientists realized that Artificial Intelligence lacks intelligence because it is based on traditional machines, they started to develop models based on the human brain – neural networks. Neural networks, like the human brain, use pathways of artificial neurons to pass information.
Unlike the central storage unit in a computer, there is a vast network of neurons in the brain, in which information is stored. Scientists are trying to emulate this network to pass information.
In a neural network, one neuron causes the reception of information to activate other connected neurons. Each neuron, in turn, sends the signal of the input to other selective neurons, triggering a wave of input signals. Based on how sensitive a neuron is to a particular information input, the strength of the signal that is passed on increases or decreases. The strongest signals end up being the output of the wave.
Thus, if one inputs the alphabets ‘a’ and ‘n’ in a neuron, it will send the signal to other connected neurons and so on. Thus the neurons that are related to ‘a’, ‘n’, and ‘an’ will be the strongest signals that will be the output.
Yet, these artificial neural networks are no closer to the sophistication and complexity of the human brain. They are limited by a one-sided flow of information, unlike the brain, where information can be affected by feedback loops from the higher parts of the brain, thereby affecting signals. It is these feedback loops that help us remember a mathematical formula learned when we were in school after trying hard to recollect.
Neural networks also do not have the ability to build memory banks as the brain does. Thus previously received inputs cannot be stored for later use.
Will Computers Be Intelligent?
While humanity is still far from creating truly intelligent machines, we might be closer than expected. The biggest obstacle that scientists still have to overcome is to provide a computer a memory capacity as large as the human brain. This would mean having a memory of about 8 trillion bytes to replicate the brain’s synapses. Today, a computer has only 100 billion bytes.
Despite the size of memory, it could be a feasible prospect today, if not for the second obstacle that scientists face – to make a computer with such a large memory in a feasible size to enable practical use.
The answer to this problem could be solved by silicon chips that not only consume less power but also fast and robust. Silicon chips that could exceed the memory capacity of the human brain are not too distant a possibility!
However, the third obstacle that challenges is that in the human brain, one nerve cell is connected to a thousand others. In AI, scientists still have not discovered how to achieve this kind of byte connectivity in silicon chips.
The possible solution? Single fiber optic cables. These cables are used in telecommunications to transmit more than a million concurrent conversations over a single cable.
If scientists are able to over come these three obstacles, more intelligent computers might as well be on their way!
Intelligence Machines: Threat Or Benefit?
The world of entertainment has already cast AI as a threat to humanity. Many movies and novels about killer, self-aware sentient beings have made greens at the box office. However, thankfully reality will be a far cry away from these!
AI machines will have their intelligence modeled around the neocortex of the brain. The neocortex does not process emotions and feelings such as love, hate, anger, fear, and desire. These emotions are generated and processed in a more primitive, older part of the brain.
As long as this part does not feature in the making of AI, the machines will remain unemotional, only capable of emulating the brain’s ability to learn, understand and link new information to stored memory. Essentially, they will be able to think, but not feel.
To think of it, the benefits of AI as intelligent as the human brain will eventually far outweigh the risks and threats imagined. Considering the fact that a machine will never ‘die’, it could end up have a memory that would exceed that of humans, and be able to accumulate even more knowledge than humans do in a lifespan. While it is possible that AI could surpass the human ability to think and process knowledge, it would never become a threat.
Today the human brain outweighs computers in knowledge and the ability to learn, experience, and think due to the way the neocortex works. Technological development and neural networks bring the possibility of machines being smarter than humans, better at the memory, and more knowledgeable in the future. However, the perception that AI could bring about the extermination of the human race is merely a good movie concept.