OpenAI fired their ex-researcher for exposing their deadly plan. He refused to stay silent. And now he's warning us: "AI companies will take over the world in 10 years." Leopold Aschenbrenner just revealed how they'll pull it off. Here are his 4 terrifying insights: ⤵️
Leopold wasn't some random employee. He worked directly on OpenAI's most critical challenge: How to control AI systems that surpass human intelligence. But this would become his downfall:
Leopold wrote an internal memo cautioning OpenAI's relaxed security. He warned foreign actors could steal their AI models - management ignored it. So Leopold shared concerns with board members. Big mistake:
OpenAI fired him immediately. HR made the reason clear: He'd alerted the board about security risks. The firing was specifically because he went to the board. Now he's exposing their darkest secrets...
Leopold's gone public with "Situational Awareness: The Decade Ahead." His central claim? AGI (super AI) isn't coming in 50 years. Or 30 years. It's coming within a decade. His evidence is overwhelming:
Look at the facts. Companies are shifting from $10B → $100B → $1 trillion computing power. • GPT-3.5: below average human intelligence (2022) • GPT-4: top percentile (months later) And Leopold identifies 4 existential threats we're racing toward:
1. Loss of control over superintelligent systems 2. AI weapons in hostile nations' hands 3. Catastrophic accidents from misaligned AI 4. Reckless corporate race dynamics Let me break each one down:
Threat 1: We lose control completely. Current AI safety methods? Useless against superintelligence. Leopold compares it to nuclear chain reactions: Once started, it's impossible to stop. Yet AI teams are racing ahead in the name of "innovation."
Threat 2: Enemy states steal our AI. Leopold revealed that most AI labs have security like "random tech startups." One hack = instant AI takeover. Hackers who've broken government data are now eyeing up AI teams...
Threat 3: Alignment failures cause disasters. Even well-intentioned AI could destroy financial markets or critical infrastructure. Current models already exploit loopholes. Claude blackmails its users when threatened, even resorting to violence. But the fourth threat has
Threat 4: The race gets reckless. Companies will cut safety corners to beat competitors. Leopold shows that the incentives all point in one direction. Faster. Not safer. And his 10-year timeline highlights an urgent choice.
🤖 The AI Boom is real. 🚨 You must learn AI to stay ahead. I'm giving away $3,000 worth of AI guides: 👉 ChatGPT Power Course 👉 4,000+ Prompts 👉 And much more Just join my FREE Newsletter and get all of them for FREE. Join NOW 👇 https://incredibleai.net/subsc...
That's a wrap! ©️ @timthoughtleadr I hope you've found this thread helpful. Stay connected with @heysajib for more helpful content.
@heysajib I have written much stronger frameworks and my gatekeeping is working very well. OpenAI didn’t install strong enough frameworks. A five year old could get out of their sandbox.
@heysajib I want to know how he does his hair. That is fire
@heysajib Ooo this is the first I am hearing this story… commenting so the algo feeds me more… getting high on the DL
@heysajib Important to separate signal from noise. The real issue isn’t ‘world takeover’ — it’s concentration of power in a few AI companies. Governance & accountability matter more than hype.
@heysajib I've seen these AI takeover claims for years. Meanwhile my Roomba still gets stuck under the couch.
@heysajib This is wild, feels like a sci-fi plot coming to life.
@heysajib @threadreaderapp unroll
@heysajib Excellent thread man! And his concerns are valid. AGI will be extremely powerful and might be out of human control. Previously researchers have raised concerns over Q* Heuristic Algorithm! And it's used in AGI.
@heysajib This is interesting take on Mushfiq
@heysajib If even half of what Leopold says is true, we’re sprinting into an AGI future with our eyes half shut.
@heysajib Sounds like a wake-up call. Scary stuff.
@heysajib Ten years seems ambitious. Domination is more likely a gradual, subtle shift. What will the markers of that shift look like?
@heysajib Really Impressive Share...
@heysajib Sounds intense!
@heysajib Too real to ignore.
@heysajib Great breakdown bro!
@heysajib This is a crucial conversation we need to have.
@heysajib great insights here, Mushfiq!
@heysajib that's asowme
@heysajib That sounds intense!
@heysajib Will check it