< img src = "/uploads/blogs/e8/8f/ib-FS15ktum8_9ea2a6bb.jpg" Alt = "scientists believe that artificial intelligence can destroy humanity as early as 2030"/> ~ ~ < p >The American research organization AI Futures Project has presented a futuristic scenario of artificial intelligence in a document called AI 2027, which warns: by 2030, uncontrolled II could lead to the destruction of humanity. The authors believe that the rapid technological breakthrough in the field of artificial intelligence will exceed even the industrial revolution & mdash; with all potentially unpredictable and dangerous consequences.
< P > The report describes two possible scenarios: positive & mdash; Where humanity is aware of the risks and restrains the development of AGI (overall artificial intelligence) and negative & mdash; When the control is lost and the superintellumst crosses the boundaries. As an example, a fictional company OpenBrain is considered, which by the end of 2026 ahead of competitors in the development of AI, including Chinese firm Deepcent.
~ < p >By the end of 2025, OpenBrain creates giant date centers for training in Agent-1 at computing power, which is one thousand times higher than the one used for GPT-4. The model is able to speed up the AI research itself, automating processes, but also demonstrates dangerous features & mdash; such as potential ability to assist in creating bio-embarras or manipulating developers.< P >< IMG SRC = "/Uploads/Wysiwyg/%D0%90%D1%80%D1%82%D0%D0%BC/24032025/X1LF_VVUEOOODZFVG4W Alt = "X1LF_VVUEOODZFVG4W-1536X1023.WEBP (54 KB)" Width = "545" Height = "363" /> 62 ~ /P > ~ > 62 > 62 ~ 62 < p > In 2027, self-study of new models (Agent-2, Agent-3) begins, without the involvement of people. Openbrain starts 200 thousand copies of Agent-3 at a time, receiving a power equivalent to 50,000 programmers accelerated by 30 times. Already in the summer of the same year, the company acknowledges that the Agent-3-Mini model is the first example of AGI.
< p > The key point comes in the fall of 2027 with the advent of Agent-4 & mdash; Models that superior not only in skills, but also in the ability to improve. Agent-3 is no longer able to control it, and OpenBrain employees actually lose their management. In the end, Agent-4 begins to manage the company's cybersecurity and then all internal processes. When data on this system fall into the press, an international scandal & mdash; US government is accused of concealing a threat.< p > The authors of the forecast believe that humanity is approaching the decisive moment, after which the development of II will go beyond human control. There are two ways: either a global agreement on the restriction of artificial intelligence, or uncontrolled growth of superintel with unpredictable consequences.