Researcher Jacob Coxon, who spent the last three years training neural networks at OpenAI and Anthropic, resigned and spoke about the inner workings of the AI giants.
According to him, developers are well aware that creating superintelligence could lead to the extinction of humanity, but no one intends to stop.
Why They Keep Going
The companies have fallen into the trap of an endless race. The logic is simple: «If not us, then competitors will be the first to create superintelligence, and they may not be as careful». According to Coxon, OpenAI simply underestimates the scale of the threat. At Anthropic, the risks are clearly understood, but they are afraid of falling behind the market.
At the same time, neural networks are evolving rapidly. Coxon is convinced that soon AI will be able to hack any system on its own, make scientific breakthroughs in a matter of hours, and gain real influence.
The Most Alarming Part — This Is Confirmed from the Inside
The former researcher’s statements were publicly supported by Evan Hubinger, the current head of AI safety at Anthropic. He confirmed three alarming facts:
-
Inside the company, the scenario of human extinction caused by AI is considered entirely realistic.
-
He estimates the probability of such an outcome within the next 10 years at more than 10%.
-
Researchers still do not have a solution for making superintelligence safe for humans.
Conclusion: The people who understand the potential danger of artificial intelligence better than anyone continue creating it because they are afraid of losing to competitors. And this human factor is far more frightening than a «machine uprising».
