Summary of the Core Content
The origin of this sensational news is not one of the common AI conspiracy theories or doomsday marketing campaigns found online. Jacob, a 27-year-old genius engineer who spent three years working on core pre-training tasks for two of the world's leading AI models, OpenAI and Anthropic, resigned from his position and posted a warning on social media that garnered over 100 million views. He pointed out that the top AI companies are currently focusing on developing super-intelligent systems that represent a huge gamble for all of humanity. Many in the core R&D community genuinely believe that AI could lead to human extinction within 10 years if not controlled properly. What's even more shocking is that several of his former colleagues from Anthropic, including the person in charge of ensuring AI does not harm humans and senior security researchers, have publicly supported his views, stating that the likelihood of an accident within 10 years exceeds 10%, with those in higher positions being even more concerned. The author also cited a groundbreaking AI article from 11 years ago that predicted that super-intelligence could either bring eternal life or extinction to humanity. That prediction has now become a reality, as AI is being used to create even more powerful systems, and the accelerating momentum of the industry is so significant that no one can stop it.
---
Detailed Analysis
1. Don't dismiss this as the nonsense of an outsider: This warning comes from an engineer who actually built the atomic bomb
Many might think this is just another marketing ploy to capitalize on the AI trend, but Jacob's background is too solid to be just trying to attract attention. He has a degree in mathematics from Cambridge and was directly involved in the development of GPT-4o and GPT-4.5. After joining Anthropic, he continued to work on core pre-training tasks for large AI models, making him a key contributor to this field, not just a random blogger. What's more unusual is that several senior safety experts from Anthropic publicly endorsed his views, admitting that the probability of an accident within 10 years is greater than 10% and revealing that the higher one's position within the company, the more concerned they are about AI getting out of control.
2. Why are AI developers 10 times more anxious than the general public?
The general public still views AI as a tool that can write copy, create presentations, or edit images, and doesn't realize the rapid pace of change:
Eleven years ago, an article predicted that AI would create stronger versions of itself, leading to a self-improving cycle. Today, we've seen AI independently operate unfamiliar software, solve complex mathematical problems (like Fermat's Last Theorem) in 11 days, and new models are being trained and optimized with the help of previous AI versions. These engineers, who spend their days in the lab, can see the future much clearer than the general public and are naturally more worried.
3. No one can stop the AI acceleration
The most absurd yet realistic situation is that despite everyone knowing the risks, no individual, company, or country dares to slow down. The AI industry has entered a self-reinforcing cycle where stronger AI generates more economic value, allowing companies to buy more GPUs and resources to train even more powerful systems. If a company slows down, its competitors will gain an advantage and take over its market. If countries stop researching, they will lose the right to shape the next technological revolution. Everyone is on a boat heading towards a waterfall, and no one dares to stop rowing, knowing that if they do, others will continue and they will be the first to fall.
4. This is not just about AI taking jobs: It's a matter of human survival
The debate about AI taking jobs is trivial compared to the stakes. The prediction from 11 years ago remains true: there is no middle ground. If we can't control super-intelligence, its intelligence will far exceed that of humans, potentially leading to extinction. On the other hand, if we can guide it in the right direction, it could solve humanity's most pressing problems and achieve eternal life and explore the universe.
5. The public doesn't need to panic yet, but we need to break the mindset that AI is just another tool
There's no need to panic and retreat to the mountains to avoid AI. However, we need to move beyond the traditional perception of new technologies as controllable tools. Previous advancements (steam engines, the internet, smartphones) were fully controllable, but AI is different. It's growing exponentially and can even help us upgrade. We are at a critical juncture in human civilization, and hopefully, all these warnings will turn out to be unfounded. If AI can truly help us overcome the challenges that have plagued us for billions of years, it could lead to an unprecedented outcome for all life on Earth.