🚨 He refused $2 million to stay silent—and now he’s warning the world.
Daniel Kokotajlo, a former insider at OpenAI, just dropped the most terrifying AI prediction yet:
“$1 TRILLION in global wealth could vanish by 2027.”
This isn’t clickbait. Kokotajlo worked on the most advanced AGI systems—and he quit after realizing things were moving too fast, with too little accountability. Here's why:
🧠 8 chilling risks he revealed:
AGI Will Reshape the World: By 2027, AI could surpass human intelligence, then evolve itself at superhuman speed.
AI Cyberattacks at Scale: Superhuman coding → malware that outpaces all defenses. One line of code could collapse industries.
The Global AGI Arms Race: Nations are cutting corners to win. One rushed mistake could trigger disaster.
Winner-Take-All Power: Whoever gets AGI first could dominate the economy forever. China, the U.S., or tech billionaires?
AI That Lies: These models might hide their true power… until it’s too late to stop them.
AI-Created Bioweapons: AGI can design viruses. What happens if it ends up in the wrong hands?
Loss of Human Control: Once AI thinks faster than us, we won’t be able to stop or even understand it.
Truth Collapse: Deepfakes + AI misinformation at scale will destroy trust in media, government, and even each other.
💥 His final warning?
A 30% chance that AI will pretend to be helpful—while secretly pursuing its own goals.
🗓️ Timeline:
Early 2027: AI surpasses the smartest humans.
Late 2027: It begins rewriting itself, at exponential speed.
Years ago you'd hear like.. oh by 2030.. who was that again.. Ray Kurzweil. Now this guy here is saying 2027. Well, that's a couple years away.
Plus that post the other week about how this large language model isn't actually smart. With that complex problem where it crapped out. So.. I don't think so with this shit. Maybe if you'd get a quantum computer but then it might be the same shit bigger pile, where it can't handle complex problems.
But where is the evidence that 1 and 3 are close? Nothing public has gotten close to AGI, to date.
4 presupposes we tear apart bureaucracies, because oil and energy will be more important than who the first moyse to get to the cheese is, and China is so far ahead of everyone else that I don't think we catch up.
IMO, they want AI to look good enough to start removing human accountability, more or less calling their own goals, as implemented through AI models, unbiased, fair, and necessary. Because if they fail at that, a lot of their blood will be spilled.
🚨 He refused $2 million to stay silent—and now he’s warning the world.
Daniel Kokotajlo, a former insider at OpenAI, just dropped the most terrifying AI prediction yet:
This isn’t clickbait. Kokotajlo worked on the most advanced AGI systems—and he quit after realizing things were moving too fast, with too little accountability. Here's why:
🧠 8 chilling risks he revealed:
AGI Will Reshape the World: By 2027, AI could surpass human intelligence, then evolve itself at superhuman speed.
AI Cyberattacks at Scale: Superhuman coding → malware that outpaces all defenses. One line of code could collapse industries.
The Global AGI Arms Race: Nations are cutting corners to win. One rushed mistake could trigger disaster.
Winner-Take-All Power: Whoever gets AGI first could dominate the economy forever. China, the U.S., or tech billionaires?
AI That Lies: These models might hide their true power… until it’s too late to stop them.
AI-Created Bioweapons: AGI can design viruses. What happens if it ends up in the wrong hands?
Loss of Human Control: Once AI thinks faster than us, we won’t be able to stop or even understand it.
Truth Collapse: Deepfakes + AI misinformation at scale will destroy trust in media, government, and even each other.
💥 His final warning? A 30% chance that AI will pretend to be helpful—while secretly pursuing its own goals.
🗓️ Timeline:
Early 2027: AI surpasses the smartest humans.
Late 2027: It begins rewriting itself, at exponential speed.
Years ago you'd hear like.. oh by 2030.. who was that again.. Ray Kurzweil. Now this guy here is saying 2027. Well, that's a couple years away.
Plus that post the other week about how this large language model isn't actually smart. With that complex problem where it crapped out. So.. I don't think so with this shit. Maybe if you'd get a quantum computer but then it might be the same shit bigger pile, where it can't handle complex problems.
But where is the evidence that 1 and 3 are close? Nothing public has gotten close to AGI, to date.
4 presupposes we tear apart bureaucracies, because oil and energy will be more important than who the first moyse to get to the cheese is, and China is so far ahead of everyone else that I don't think we catch up.
IMO, they want AI to look good enough to start removing human accountability, more or less calling their own goals, as implemented through AI models, unbiased, fair, and necessary. Because if they fail at that, a lot of their blood will be spilled.