In my last article, I talked about topics such as form and function, processing and memory capacity, connectivity, power consumption, artificial intelligence architectures, network intelligence, working speed, learning techniques and data-based decision resolution. Now let's come to the 10 differences between artificial intelligence and human intelligence: Let's get started then.
1. FORM AND FUNCTION WITH ARTIFICIAL INTELLIGENCE
Neural networks are software that runs on a computer. You can delete them from the computer and reinstall them. These neurons are codes inside software. However, in the human brain, the software is coded directly into the hardware, and software and hardware are an inseparable whole. For example, your memories and the information you learn are encoded into protein chains in the trillions of neural network connections between your neurons.
So much so that your brain does not save files like a computer and does not capture the world on video like a camera. You may have a photographic memory, but your brain does not take pictures. Instead, it records impressions and associations (similar to correlations in neural networks). That's why you can't even remember your mother exactly. Each time you recreate and reconstruct the memory (you remember the color of the sweater wrong, etc.).
Only your weight values allow you to approximate the birthday you spent with your mother; because weight values create dispositions that generate coherent memories (fidelity memories).
2. SIZE OF ARTIFICIAL INTELLIGENCE
The human brain, along with other cells, consists of 100 billion neurons. There are 20-30 billion neurons in the cerebral cortex, which produce the upper functions that make us human, such as consciousness, emotions, and ethical values. Neural networks today usually consist of 300 virtual neurons.
3. CONNECTIVITY WITH ARTIFICIAL INTELLIGENCE
In neural networks, each layer is linked to an upper and lower layer in the hierarchy; that is, neural networks operate sequentially, and although they collect correlations associatively from multiple channels (weight values of virtual neuron connections), they essentially learn sequentially. He establishes the correlations of virtual neuron connections by analogy, that is, with pattern recognition units, in accordance with the philosopher Plato's like-like-knows principle:
In short, virtual neurons in each layer are also pattern recognition units. There are about 300 million pattern recognition units in the human brain, each consisting of 100 neurons. We learn new things with these units. As we get older, we learn more difficult and slower as our pattern recognition units fill up over time. You can read the article How fast do we think to see how the human brain performs big data analysis with pattern recognition units. However, let me point out:
The human brain is not divided into layers. Even specialized departments such as the visual and auditory centers do not consist of different tissues in the cerebral cortex. In fact, an unrelated part of the brain can take over the function of the visual center in those who have suffered a brain injury. Indeed, not all brain regions are equally connected. Nor is there a separate region, tissue, or physiological mechanism that makes up human consciousness.
4. POWER CONSUMPTION WITH ARTIFICIAL INTELLIGENCE
The human brain consumes about 20 Watts of power when operating, which is equivalent to a standard notebook. However, the brain runs a million times more neurons than the processor cores with 20 Watts. So much so that you can call the human brain, which contains 100 billion neurons, a processor with 100 billion cores. In addition, our brain uses energy very efficiently. It does not overheat and explode like smartphone batteries when you work hard or stay in the sun.
5. STRUCTURAL ARCHITECTURE
As I explained in the quantum consciousness article, the human brain is not a quantum computer that performs many parallel operations with a single core processor. On the other hand, it is a parallel processing organ, taking advantage of the power of multi-core processors with at least 300 million pattern recognition units within 10 billion neurons. However, it does not perform parallel operations in any particular order or sequence.
Think of it this way: When reading a book, of course, you read the lines in order. However, sometimes something catches your eye on the page and you jump to the next sentence a few paragraphs and go back to where you left off, or you quickly flip through the pages of the first book you bought and take a look.
6. ACTIVATION POTENTIAL WITH ARTIFICIAL INTELLIGENCE
Not all neurons in the human brain fire at the same time and with the same intensity. This is one reason why the brain uses energy efficiently. If all neurons sent signals at the same time, we would have an epileptic attack. The firing potential of our neurons depends on the chemical reactions and electrical potential in the cell and brain fluid. However, in artificial intelligence learning, virtual neurons in neural networks work continuously with a continuous flow of data. Thanks to their weight values, they make continuous data analysis.
7. SPEED RACE WITH ARTIFICIAL INTELLIGENCE
The human brain is not a computer! The main purpose of the brain is not to solve problems, build morally advanced civilizations, or solve the mysteries of the universe. The main purpose of our brain is to make it easier for us to survive and pass on our genes to our descendants, while managing our vital functions.
On the other hand, our brain encourages creativity and curiosity in order to provide an additional adaptive advantage in the process of evolution. We pay the price for this by making mistakes, forgetting and getting tired. Frankly, the human brain works very slowly. While a standard laptop performs 10 billion operations per second, the neurons in our brains perform at most 1000 operations.
However, if you say that its function, not its speed, is important, you can see that the human brain can do many things with a capacity of only 1000 Hz that computers with a capacity of 10 billion Hz cannot. In short, it is meaningless to compare speed with artificial intelligence. Artificial intelligence can perform the same operation millions of times without errors; but we get sleepy every day. However, AI cannot (for now) write War and Peace. It is written by Tolstoy :))
8. LEARNING TECHNIQUES WITH ARTIFICIAL INTELLIGENCE
At the beginning of the article, I said that I will explain the feedback in neural networks about artificial intelligence and machine learning: Neural networks always focus on the best solution when routing to remove traffic congestion. When insufficient solutions appear in the weights, they reset the virtual neuron connection correlation values that lead to these initial solutions during feedback.
In short, while learning something new, artificial intelligence forgets the old one, and this has huge disadvantages! For example, by showing 1 million pictures of dogs to neural networks, you enable the artificial intelligence to learn to recognize dog photos on the internet. However, if you start showing cat pictures as dogs, unconscious and forgetful artificial intelligence with fish memory learns cat pictures as dog pictures.
The human brain, on the other hand, learns new things based on what it has learned before; so there is no such thing as irrelevant data in associative thinking. Even the most irrelevant data can become necessary and useful over time. Indeed, that is what creativity is. Know that the mouse and touch screen that created e-commerce and digital marketing, and even the world wide web (www) itself, were developed by physicists whose only job was to collide protons head-on at near-light speeds in a CERN particle accelerator. This is creativity and digital transformation.
9. INFRASTRUCTURE PROBLEM WITH ARTIFICIAL INTELLIGENCE
The human brain is made up of living soft tissue, and we call it wet hardware, as I mentioned in the Ghost in the Machine article. However, when we say that the software is coded into the hardware in the human brain, we do not mean the truth. This is a flawed analogy; because our brain is neither digital nor analog. It does not consist of two separate parts, software and hardware.
As a matter of fact, the human brain learns new things by establishing new neural networks between neurons and creating intracellular chemical predispositions that change the activation (firing) potential of neurons. In short, the human brain changes its structural architecture as it learns.
10. LEARNING RESOLUTION
The final difference between artificial intelligence and human intelligence, learning resolution, can be divided into two: accuracy and precision. Accuracy is, for example, accurately zooming in on a single atom with a scanning electron microscope. Precision is how clearly you can see that atom by zooming in high. It's also like the difference between shooting zombies in the body or head in video games.
Clearly, the human brain has evolved to survive by rapidly adapting to environmental conditions through making sense of the world. Therefore, it brings together different potential risks and opportunities with associative awareness and quickly; but it does not work with microscope precision.
To sum things up, human intelligence creates models of the world. Neural networks that learn with artificial intelligence make pattern recognition classifications. For this reason, they are very sensitive to inaccurate data and are easily mistaken for even the smallest corrupted data.
