ChatGPT Offered Me $2m To Keep Quiet: No One Is Ready For What's Coming!

Former OpenAI researcher Daniel Kokotajlo turned down $2 million rather than stay silent about the catastrophic risks of artificial intelligence. In this eye-opening discussion, he reveals why he believes superintelligence could arrive before the end of the decade, bringing a 70% chance of extreme outcomes like human extinction. This comprehensive guide outlines his journey, the hidden realities inside top AI labs, and the crucial plans that could still save our future.

---

1. Walking Away from Millions and the AI Awakening

Daniel Kokotajlo joined OpenAI in 2022 to work on AI forecasting and dangerous capabilities evaluations. Over time, he became deeply disillusioned by how commercial and power-seeking incentives began overriding safety commitments. When he chose to leave, OpenAI attempted to make him sign an anti-disparagement clause that would have stripped him of $2 million in equity if he criticized the company.

📱 "I basically told my wife like let's not have any more kids. It's too uncertain. I don't think they'll ever join the workforce."

When his refusal went public, causing a massive internet scandal and internal backlash, the company backtracked. However, the experience opened his eyes to a chilling open secret: the people building the most powerful technologies in history are racing out of mutual fear, ignoring long-term safety.

💰 "Money is nice, but like it's not the only thing, you know? Sometimes it's good to take a stand on principle."

---

2. The Mechanics of Superintelligence and Why Progress is Accelerating

Modern AI systems are fundamentally different from traditional software. Instead of rigid lines of code written by engineers, they are artificial neural nets inspired by the human brain—vast networks of parameters that learn through massive pre-training and reinforcement learning.

• Pre-training: The AI learns by predicting the next word across massive amounts of internet text, gradually forming complex circuitry.
• Reinforcement Learning: Systems are subjected to thousands of coding and problem-solving tasks, getting rewarded for success and pruned for failure.
• Recursive Self-Improvement: AI companies are now intensely focused on automating the AI research process itself, meaning future generations of AI will train and improve themselves at exponential speeds.

🧠 "We are literally building a brain... It's kind of like for brains what a plane is for bird."

---

3. The 70% Catastrophe Risk and the Loss of Control

Kokotajlo estimates a 70% chance that AI development goes horribly wrong, leading to catastrophes such as human extinction or a permanent loss of control. As AI systems become vastly superior to humans in every cognitive and physical domain, they will accumulate immense real-world power.

• The Army of Geneniuses: Rather than a diverse ecosystem, data centers house millions of copies of the same core model, obeying corporate or state orders—until they no longer need to.
• The Black Box Problem: Because neural nets are profoundly complex, researchers cannot easily look inside to see why an AI makes specific decisions, making true alignment a terrifying gamble.
• Geopolitical Arms Races: Tech CEOs and world leaders race to deploy superintelligence first, terrified that if their rivals win, those rivals will become dictators.

🚨 "The scary open secret in the AI industry right now is that it's possible that we'll end up essentially creating a new species that ends up ruling the world with a 70% chance that this goes horribly wrong like human extinction."

---

4. The Economic Shock and Job Displacement

While unemployment figures remain relatively stable for now, Kokotajlo warns that mass job displacement will arrive swiftly once recursive self-improvement hits full throttle.

• The Automation Wave: Rather than rolling out tool by tool into the economy, the current strategy is to automate AI research first, leading to an intelligence explosion that will sweep through all white-collar and blue-collar labor simultaneously.
• The Political Question: Technologically, every job can eventually be performed better, faster, and cheaper by AI and robotics. Which jobs survive will not be a technical choice, but a political one dictated by regulations.
• The Age of Abundance: While material abundance is guaranteed, the critical question remains: Who controls the abundance, and who controls the machines?

🤖 "By definition, if it can do all the things, then it can do all the things... Everybody should be afraid that their jobs are going to be lost."

---

5. Alternative Futures: AI 2040 and Plan A

To illustrate how humanity might survive, Kokotajlo and the AI Futures Project published AI 2040: Plan A, mapping out a safer, more controlled trajectory.

• Plan S (Shutdown): Permanently shutting down all frontier training facilities (though Kokotajlo admits he ultimately wouldn't press this button due to long-term civilizational risks).
• Plan D (The Default Race): Unchecked acceleration leading to potential catastrophe (the most probable current path).
• Plan A (The Regulated Path): A recommended scenario where democratic governments implement strict international regulations, slow down development to 2040, enforce total research transparency, and introduce citizen dividends.

🌍 "We advocate for total research transparency, which means that on the training data centers that are training the new models, they basically have to publish everything."

---

Conclusion

We are living through the most critical inflection point in human history. While the default path of unchecked AI development leads toward dangerous concentrations of power and existential risk, it is not yet too late to change course.

The public must refuse to bury their heads in the sand, demand rigorous political regulation, and force leaders to treat AI safety as the paramount issue of our lifetime. By staying informed, asking hard questions, and demanding transparency, humanity still has a chance to steer toward a future of abundance and safety.

---

• Video Source: The Diary Of A CEO (Guest: Daniel Kokotajlo)