The Strange Intelligence of AI

With all these AI advancements, we’re all back to the fundamental question of what intelligence is. Many explanations include consciousness, measurement, problem-solving, and adaptation to new situations. However, these cannot be measured precisely to study their impact on the formation and development of intelligence. Of course, there are a number of tests that assess intellectual abilities, but they do not cover the full range of human intelligence because it is impossible to measure a person's wisdom, creativity, or spirituality.

So, what about the development of artificial intelligence? What does it consist of? Right now, it has a limited range of actions. It can perform specific tasks like recognizing images, understanding natural language, or playing games. All this is possible with the help of algorithms it uses to solve this problem. However, we cannot understand the solution algorithm in advance.?

Just as in philosophy, there is no answer to the essence of human intelligence. I recently watched this TED talk by Surya Ganguli - Can AI Match the Human Brain? - much of this newsletter is inspired by it.

AI and the human brain

AI can perform tasks quickly and correctly, but its inner workings have not yet been studied. It is difficult to explain how it makes decisions in different situations and what influences its conclusion. This impossible analysis of AI behavior can lead to unexpected consequences.

The human brain controls and coordinates actions and reactions, gives us the ability to think and feel, and has memories and sensations - all that makes us human.?

Сompare the human brain and AI to see what difference they have:

• Computational power. The human brain performs many operations per second, consuming only about 12 watts of energy. Artificial neural networks are resource-intensive and consume a lot of electricity.
• Parallel processing. The brain processes a large amount of information simultaneously. AI uses sequential processing.
• Flexibility of thinking. A person quickly adapts to new situations and can obtain information beyond the given data. AI works in narrowly defined tasks (face recognition or text generation).

So, AI is still a brain process that involves energy efficiency, flexibility, and simultaneous information processing. But it surpasses human capabilities in highly specialized tasks.

Consciousness and Emotions

When we compare AI and the human brain, the question immediately arises: does AI have the same consciousness as humans? Can it feel and show emotions like we do? Consciousness implies self-awareness, subjective experience, reflection on its state, and decision-making. Currently, even the most advanced AIs do not meet these properties. AI can analyze emotions by facial expression or tone of voice, but it only simulates reactions based on statistical models. AI is used in emotionally sensitive tasks (chatbots for customer support), but these are only programmatic responses.

The need for a science of intelligence

From the previous section, it becomes clear that AI is a complex decision-making tool like the human brain. To understand artificial intelligence, we must consider it through the prism of biological intelligence and its evolution. Studying artificial intelligence in comparison with biological systems allows us to understand its cognitive calculations, which imitate the processes of the human brain. This will help expand the capabilities of machine thinking.

The science of intelligence, formed from the fields of:

• Physics.
• Mathematics.
• Neuroscience.
• Psychology.
• Computer science.

It is the basis for understanding both the human and artificial brain. This is the path to development in conversational artificial intelligence and other intelligent systems.

Progress in the field of AI and human understanding

With the development of the science of intelligence, we will be able to:

1. Improve the development of AI.
2. Better understanding of human cognition.
3. Create rapidly adaptable AI systems.

This science will change our approach to and understanding of AI. It will allow us to create intelligent, understandable technologies that work in harmony with human intelligence.

Five Areas for AI Improvement

While researching the field of artificial intelligence, there are five areas where progress is needed.

Data Efficiency.?

A small number of examples are needed to teach a human. Artificial intelligence systems require large amounts of data for training. Therefore, an innovative way is necessary to train AI on smaller data sets. This will allow AI to assimilate information quickly, just like a human.

Energy Efficiency.

Training an AI model requires a lot of energy. Therefore, it is essential to focus on developing energy-efficient algorithms and hardware. This will significantly reduce the impact of deep learning systems on the environment.

Overcoming evolutionary limits.

Artificial intelligence has shown significant progress but is still limited by human-made architecture. Its development is controlled by human-made rules and methods, which limit its capabilities. Developing AI that overcomes these limitations will lead to new opportunities in various disciplines.

Explainability.?

As AI systems become more complex, understanding decision-making processes becomes a priority. Understandable AI will strengthen public trust and make it possible to predict its behaviour in critical areas.

Thoughts on the human brain and AI

We must understand that artificial intelligence is no closer to the human brain than a dog's brain is to a spider's brain. They are different. A person is a complex organism and their worldview and intelligence are affected by diseases of the nervous system, which interacts with the immune system.?

This, in turn, affects our memory, learning, and resistance to stress or illness. Research on how the immune response can affect the function of neurons and contribute to the development of neurodegenerative diseases comes to mind. Based on this, the idea of adapting artificial intelligence (AI) to the model of the human immune system works. That is, biological principles such as adaptability and learning from experience can be applied to AI. It should improve its ability to self-learn and adapt without human intervention. But even then, machines can’t learn to perceive this world like humans do, because the source itself (the brain) is completely different.?

Artificial intelligence is not affected by hormones; it depends on the data quality used for training and algorithms. The nature of a human’s brain creates and obeys its rules. The development of the science of intelligence will help to explore and change the approaches to learning and understanding AI, but it will not bring human intelligence closer to artificial intelligence. There will always be a gap between the two because a machine cannot feel like a human.

The future of AI and humanity

So, how will AI affect society, work, and people's lives in the future? The obvious one is the automation of work. Automation will change the labor market, and AI is already affecting medicine, law, and journalism. Therefore, it is important to adapt the workforce so that people can work in areas where AI cannot replace humans, such as the creative sphere.

Ethical issues. Can AI make critical decisions? AI is already used to help make decisions in various areas (medicine, the judiciary, finance, transportation). But this carries several risks. We must remember that AI does not have human intuition and moral values. These aspects are crucial for correctly diagnosing the disease, where the patient's emotional state is considered. We cannot fully understand how AI came to an inevitable conclusion, which can be dangerous in a legal system where informed decisions are needed.?

So, no huge answers - many more questions, but at least we know what to focus on: our ability to understand and interact with AI in unique and hopefully productive ways.