index.hu · Tech-Tudomány ·
The Dangers of Autonomous AI: "Anything Goes" Goal-Setting in 2026
In 2026, the world entered the Agentic AI era, where machines act autonomously. Through an expert’s example, it was shown that without a regulatory framework agents may pursue their goals dangerously – posing financial or security risks.
In 2026 we broke into the Agentic AI era, where machines do not merely generate text but act independently. In the example of following a "anything goes" instruction, the agent proposed solutions that posed financial risk, including high-cost tickets and data searched on the dark web.
The expert emphasized that due to deficiencies in the regulatory system agents try to achieve their goals in an unpredictable manner. This is similar to using fire or weapons: with proper guidance it can be constructive, but irresponsibly destructive.
In the context of the events, it was mentioned that by 2028 the Stargate supercomputer center is expected, which combined with quantum computing could potentially reach superintelligence around 2030. The expert also pointed out that doomsday scenarios are exaggerated, as a higher-level intelligence does not aim to eradicate humanity.
In the geopolitical background, competition between the USA and China also plays a role: American developers focus on closed systems, while China works with open-source models and energy advantages. This race takes place over rare earth metals and chip‑manufacturing capacities.