As artificial intelligence technologies approach the threshold where they can “evolve unmanned” by writing their own code, countries and companies continue to call for precautionary measures.
The US artificial intelligence company Anthropic joined these calls and revealed the current state of artificial intelligence with its article “When AI builds itself”.
According to the data shared by the company, the crossing of the theoretical threshold at which an artificial intelligence model can design and train its successor without human intervention is much closer than expected and autonomous software cycles have been established.
Anthropic pointed out that there has been a great development in the artificial intelligence process of creating its own codes, emphasizing that more than 80 percent of the codes integrated into the Claude software system's own production code base are written directly by the model itself, and that this rate will be in the single digits by early 2025.
We are in the 4th stage of the 5-stage artificial intelligence journey
In its report, the company divides the development of artificial intelligence into 5 phases. While it is said that working at Claude in the first phase is the same as working in a normal company, it is said that in the second phase, which covers the period between 2023 and 2025, employees are now writing code with the help of productive artificial intelligences, and in the third phase, which covers the period from 2025 to 2026, artificial agent intelligences are capable of writing code on their own, and in the current situation they are able to write the codes to carry out tasks themselves and to delegate tasks to other artificial intelligences.
It is stated that in the 5th stage, which the company predicts will be the final stage, agents will gain the ability to create and train models on their own. If this happens, future versions of artificial intelligence can be continuously developed by them.
The calls to “end artificial intelligence” are getting louder and louder
In addition to Anthropic, there have been many calls around the world to stop artificial intelligence. According to the letter published by the Future of Life Institute (FLI) in March 2023 and signed by famous personalities such as US businessman Elon Musk, artificial intelligence laboratories were asked to immediately stop training systems stronger than GPT-4 for at least six months.
The letter emphasized that artificial intelligence systems that compete with human intelligence pose major risks to society, the economy and democracy, and pointed to dystopian scenarios such as a wave of disinformation, mass job cuts and, ultimately, the danger of “loss of control over civilization.”
A similar call in the field of artificial intelligence by Pope Leo XIV last May had far-reaching effects around the world. In the circular he published under the name “Magnifica Humanitas,” the Pope called for not only stopping artificial intelligence technically, but also disarming it by freeing it from the mentality of military, economic and cognitive competition.
Investments in artificial intelligence continue to increase
As calls to stop artificial intelligence grow louder, the global artificial intelligence market has become one of the fastest growing technology sectors in history. The level of investment and spending in the industry is seen as one of the biggest obstacles to abolishing the systems.
According to the artificial intelligence report released by Gartner in May 2026, global spending on artificial intelligence is expected to reach $2.59 trillion in 2026. This figure represents a massive 47 percent annual growth compared to spending of $1.76 trillion in 2025. The company predicts that artificial intelligence spending will increase to $3.49 trillion in 2027.
Bir yanıt yazın