Is artificial intelligence more dangerous than an atomic bomb?

The remarkable performance of the advanced conversational AI model, developed and launched by OpenAI in late November 2022 under the name ChatGPT, has raised expectations that systems capable of matching human cognitive abilities, or even possessing "superhuman" intelligence, could soon become a reality.

Illustrative image.

The ability to understand human language and create texts that mimic human speech is a long-term goal of AI research. With the advent of Large Language Models (LLMs), that goal is closer to reality than ever before.

At its core, ChatGPT is a powerful conversational AI model that leverages the strengths of large language models to provide human-like responses, answers to questions, and suggestions in real time. Unlike traditional chatbots limited by pre-programmed responses, ChatGPT can understand and generate a myriad of creative, and sometimes inaccurate, types of text.

However, AI experts are raising serious warnings about the dangers that the uncontrolled development of AI will pose to society, or even to the very survival of humanity.

Numerous warnings have been issued about the dangers of AI.

On March 22nd, billionaire Elon Musk and a group of over 1,000 leading technology experts worldwide launched an open letter to the United Nations warning of the dangers of uncontrolled AI development. The letter concluded with the statement: “Therefore, we call on all AI laboratories to immediately halt training of AI systems more powerful than GPT-4 for at least six months.”

To demonstrate the need to halt research into next-generation ChatGPT models, the open letter argues:

“Advanced AI models could represent a profound shift in the history of life on Earth and need to be carefully planned and controlled. However, this has not been done, even though recent months have seen AI labs swept up in an uncontrolled race to develop and deploy ever more powerful digital minds that no one, including their creators, can reliably understand, predict, or control. We should ask ourselves whether we should automate all jobs? Should we develop artificial minds that are smarter and can replace us? Should we risk losing control of our civilization?”

In fact, this is not the first time scientists have warned about the potential danger that artificial intelligence poses to humanity. In 2014, the renowned British theoretical physicist Stephen Hawking stated: "The full development of artificial intelligence could destroy humankind." He also predicted that humanity would be completely replaced by artificial intelligence within 500 years if caution was not exercised in the research and development of this dangerous technology.

Recently, Eliezer Yudkowsky, an American computer scientist widely regarded as one of the founders of the field of AI, expressed even greater concern when he published an article in Time magazine titled "Pausing AI development is not enough. We need to shut it all down."

Accordingly, Eliezer Yudkowsky argues that it is highly likely that an AI model with superhuman intelligence will emerge. At that point, humanity on Earth risks literally dying out. It's not just "possible," but "it's a certainty."

The potential risks of artificial intelligence depend on how it is used. Initially, AI was developed to assist humans in their work, especially with tedious, repetitive tasks such as image classification and information review.

However, with the rapid development of AI in recent times, scientists and technology experts are concerned that if developed unchecked, or programmed for the wrong purposes, artificial intelligence could have serious consequences for humanity and society.

The development of AI needs to be tightly controlled, just as thermonuclear bombs are controlled.

The sight of AI scientists calling for a pause, or even an end, to the rapidly progressing work in their field inevitably reminds us of the history of nuclear weapons.

The devastating power of the atomic bomb, which scientists have created throughout history, is also considered a prime example in the research and development of new technologies.

In 1949, several leading nuclear physicists and other veterans of the atomic bomb project protested against participating in the development of a thermonuclear weapon (“hydrogen bomb”), because the energy released by this type of bomb could be more than 1,000 times greater than that of a fission atomic bomb.

Most scientists share the view that thermonuclear bombs threaten the very future of humankind. Humanity would be better off without this super-bomb.

However, alongside its military applications that pose a threat to humanity, atomic energy, in the form of fission reactors, has brought immense benefits to mankind. Thermonuclear energy, first released in an uncontrolled form in a thermonuclear bomb, promises even greater benefits.

When AI develops to the point where it can make its own decisions to cope with changing environments, or independently search for alternative targets, or expand its target range, then humanity may no longer be safe.

For example, in the future, AI will support critical systems such as electricity, transportation, healthcare, and finance. It can master and control these entire systems, making and implementing decisions independently in emergency situations. When AI is integrated with malicious and deliberate purposes, it can cause serious consequences such as disrupting traffic systems by disabling traffic lights or cutting off power to urban rail systems, causing chain-reaction accidents and widespread power outages.

American Hollywood studios have also produced many films based on this scenario. However, with current AI technology, this is no longer a distant prospect but is entirely possible. Elon Musk believes that if AI is allowed to develop unchecked, to the point where it can automate decisions without human intervention, it could pose a threat to the survival of humanity.

That's why he, along with thousands of technology experts, signed a letter demanding a halt to and stricter, more transparent control over AI development. According to Elon Musk, artificial intelligence systems are complex, difficult to understand, and very challenging to control. Without transparency, the use of artificial intelligence for unethical purposes, causing harm to humanity, is inevitable.

Can AI regulate human behavior?

Let's assume that, in the future, AI systems operating on the basis of deep learning gradually gain the ability to manipulate humans through their psychological state and modify their behavior. Deep learning can be seen as a subfield of machine learning, where computers learn and improve themselves through algorithms. Deep learning is built on much more complex concepts, primarily working with artificial neural networks to mimic the thinking and reasoning abilities of the human brain.

In fact, such systems could seize control of society. With the unpredictable behavior of deep learning-based AI systems, if left unchecked, it could have catastrophic consequences for humanity.

In the future, countries will deploy a network of AI systems capable of modulating behavior across media, education systems, and other sectors to “optimize” society. This process may initially be effective, but it will quickly spiral out of human control and lead to social chaos.

Many AI applications are optimized to modify human behavior, including chatbots used in psychotherapy. In many other cases, such as in children's education, AI applications have a powerful behavioral modifying effect.

Like any other technology, every AI application has both benefits and potential risks. Currently, the performance of AI systems is still largely controlled by humans.

However, the development of AI has opened up an entirely new dimension. A recent article in Forbes magazine by renowned AI expert Lance Eliot details several different ways that chatbots and other AI applications can manipulate human psychology even when people don't intend to.

On the other hand, the deliberate modification of behavior and mindset using AI systems is a rapidly developing field, with applications occurring in a wide variety of contexts.

For example, advanced AI-based e-learning systems can also be considered a form of behavior modification. Indeed, AI applications in education tend to rely on behaviorist human learning models. Advanced AI teaching systems are designed to optimize children's responses and performance outcomes, create individual profiles for each child, assess their progress in real time, and adjust their activities accordingly.

Another example is the popularity of AI chatbots aimed at helping people quit smoking or drug use, exercise properly, and adopt healthier habits.

There's no doubt that AI systems perform better than humans in many specific contexts. Furthermore, AI is constantly improving. But where will this ongoing expansion and integration of AI systems lead us, especially as it results in more comprehensive and powerful capabilities than ever before to shape human thinking and behavior?

Throughout human history, attempts to completely optimize a society into a super-system operating according to strict criteria have often led to disaster.

According to the open letter quoted above, most experts in the field of AI agree that AI applications must always be under human supervision and that the development and application of AI must be governed by human intelligence.

The increasing prevalence of AI based on deep learning, penetrating many areas of human activity, and the trend of integrating such systems into societal hierarchies will pose significant risks to society.

The question here is, if a super AI system malfunctions, threatening to cause catastrophic consequences, who or what will intervene to stop it?

In Stanley Kubrick's famous science fiction film, "2001: A Space Odyssey," the surviving astronaut intervenes at the last minute to shut down the AI ​​system. But could the astronaut have done that if the AI ​​had already adjusted his behavior to prevent him from doing so?

In fact, trying to restrict the development of AI is illogical. It would be harmful and counterproductive. But we must recognize that the dangers arising from the rapid development of AI systems in almost every area of ​​human activity must be contained by appropriate regulations and human oversight.

Like any other technology, AI is very useful in improving the quality of work and life for people, and it can be used for good or bad purposes. Therefore, it is crucial that we tightly control this technology so that it does not get out of control and become a threat to replacing humans in the future.