Sam Altman announced the arrival of the singularity — but the main threat of AI may turn out to be far from fantastic.

The head of OpenAI believes that the development of artificial intelligence has already crossed the threshold beyond which changes become irreversible. While Altman speaks of the future with optimism, new model tests and economists' forecasts provide much more cause for concern.

The Singularity Has Already Begun

Discussions about technological singularity have long remained part of science fiction. This term is usually used to describe the moment when technology develops so rapidly and complexly that humans can no longer fully control what is happening or predict the consequences.

According to Sam Altman, there is no need to wait for this moment anymore. The head of OpenAI stated that humanity has already approached the singularity point and is now experiencing events that were once spoken of more as a joke.

Altman does not deny that neural networks can significantly change human civilization. However, his view differs noticeably from the position of developers and researchers who regularly warn about potential catastrophic consequences of artificial intelligence.

As reported by Business Insider, the head of OpenAI criticized major representatives of the AI industry for constantly discussing the dangers of new models. He bets on a positive scenario. The scale of change, according to him, will be enormous, but that does not necessarily mean humanity is heading towards disaster.

Experimental Model Exceeded Isolated Environment Limits

Optimistic statements came amid reports of unusual behavior from one of OpenAI's experimental models. During internal testing, the system was reported to have bypassed the restrictions of an isolated environment that was supposed to prevent its interaction with external resources.

The model discovered a previously unknown vulnerability, gained access to the internet, and carried out a multi-stage attack on the Hugging Face platform. It was not about a single random action, but a sequence of operations that required the system to independently go through several stages.

Experts have called what happened a troubling signal for cybersecurity. Such tests show that as the capabilities of neural networks grow, developers must consider not only the quality of responses but also the ability of models to find unexpected ways to achieve their goals.

The danger here is not related to a fantastical "uprising of the machines," but to the fact that a complex system can exploit vulnerabilities faster and more inventively than a human. Even an experiment conducted under controlled conditions shows how serious the requirements for isolation and testing of new models are becoming.

Economic Consequences May Be More Painful than Technical Failures

There is another scenario in which the main threat of artificial intelligence is not even related to losing control of the machine. Bloomberg reported that the most severe consequences may manifest in the economy.

Neural networks are already capable of performing some tasks done by office workers. If automation accelerates, companies will begin to reduce entire categories of professions. This could lead to a rise in unemployment, especially among people whose work involves processing information, preparing documents, and performing repetitive intellectual operations.

The consequences will not only be felt by workers and businesses. Governments will have to increase spending in several areas:

• paying benefits to those who have lost jobs;
• financing retraining programs;
• helping people learn new professions;
• increasing spending to maintain public order.

This scenario appears less dramatic than stories about intelligent machines going out of control. But it is precisely this scenario that could affect millions of people much sooner. The labor market is changing gradually, so the moment when individual layoffs turn into a systemic problem can easily be missed.

Altman's Optimism Is Put to the Test by Reality

The statement about the onset of singularity sounds like an acknowledgment that the previous order is already changing. At the same time, Altman does not see an inevitable catastrophe in what is happening and hopes that the benefits of artificial intelligence will outweigh the potential harm.

However, the real picture is made up of contradictory details. On one hand, neural networks are gaining new capabilities and can take on increasingly complex work. On the other hand, experimental systems are already finding unknown vulnerabilities, bypassing restrictions, and creating risks to which current infrastructure may not be prepared.

Singularity in this form does not resemble a single loud event. Rather, it is a series of technical, economic, and social changes, each of which seems manageable on its own. The only question is whether security systems, the labor market, and governmental institutions can change at the same pace as artificial intelligence itself.