OpenAI co-founder wanted a 'doomsday bunker,' for the ChatGPT team and why CEO Sam Altman is the reason behind it
Former
OpenAI
chief scientist and co-founder
Ilya Sutskever
told his research team in 2023 that the company would need to build a protective bunker, often known as 'doomsday bunker,' before releasing
artificial general intelligence
(AGI), according to new revelations from an upcoming book about the AI company's internal turmoil.
"We're definitely going to build a bunker before we release AGI," Sutskever declared during a 2023 meeting with OpenAI scientists, months before his departure from the company. When pressed about the seriousness of his proposal, he assured colleagues that bunker entry would be "optional."
The startling disclosure comes from excerpts of "Empire of AI," a forthcoming book by former Wall Street Journal correspondent Karen Hao based on interviews with 90 current and former OpenAI employees. The book details the dramatic November 2023 boardroom coup that briefly ousted CEO
Sam Altman
, with Sutskever playing a central role in the failed takeover.
Sutskever, who co-created the groundbreaking AlexNet in 2012 alongside AI pioneer Geoff Hinton, believed his fellow researchers would require protection once AGI was achieved. He reasoned that such powerful technology would inevitably become "an object of intense desire for governments globally."
What made OpenAI co-founder want a 'doomsday bunker'
Sutskever and others worried that CEO Altman's focus on commercial success was compromising the company's commitment to developing AI safely. These tensions were exacerbated by
ChatGPT
's unexpected success, which unleashed a "funding gold rush" that safety-minded Sutskever could no longer control.
"There is a group of people—Ilya being one of them—who believe that building AGI will bring about a rapture," one researcher told Hao. "Literally a rapture."
This apocalyptic mindset partially motivated Sutskever's participation in the board revolt against Altman. However, the coup collapsed within a week, leading to Altman's return and the eventual departure of Sutskever and other safety-focused researchers. The failed takeover, now called "The Blip" by insiders, left Altman more powerful than before while driving out many of OpenAI's safety experts who were aligned with Sutskever's cautious approach.
Since leaving OpenAI, Sutskever has founded Safe Superintelligence Inc., though he has declined to comment on his previous bunker proposals. His departure represents a broader exodus of safety-focused researchers who felt the company had abandoned its original mission of developing AI that benefits humanity broadly, rather than pursuing rapid commercialization.
The timing of AGI remains hotly debated across the industry. While Altman recently claimed AGI is possible with current hardware,
Microsoft AI
CEO Mustafa Suleyman disagrees, predicting it could take up to 10 years to achieve. Google leaders
Sergey Brin
and DeepMind CEO Demis Hassabis see AGI arriving around 2030. However, AI pioneer Geoffrey Hinton warns there's no consensus on what AGI actually means, calling it "a serious, though ill-defined, concept."
Despite disagreements over definitions and timelines, most industry leaders now view AGI as an inevitability rather than a possibility.
AI Masterclass for Students. Upskill Young Ones Today!– Join Now
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
23 minutes ago
- Time of India
Welcome to campus, here's your ChatGPT
OpenAI , the maker of ChatGPT , has a plan to overhaul college education -- by embedding its artificial intelligence tools in every facet of campus life. If the company's strategy succeeds, universities would give students AI assistants to help guide and tutor them from orientation day through graduation. Professors would provide customized AI study bots for each class. Career services would offer recruiter chatbots for students to practice job interviews. And undergrads could turn on a chatbot's voice mode to be quizzed aloud before a test. OpenAI dubs its sales pitch "AI-native universities." Play Video Pause Skip Backward Skip Forward Unmute Current Time 0:00 / Duration 0:00 Loaded : 0% 0:00 Stream Type LIVE Seek to live, currently behind live LIVE Remaining Time - 0:00 1x Playback Rate Chapters Chapters Descriptions descriptions off , selected Captions captions settings , opens captions settings dialog captions off , selected Audio Track default , selected Picture-in-Picture Fullscreen This is a modal window. Beginning of dialog window. Escape will cancel and close the window. Text Color White Black Red Green Blue Yellow Magenta Cyan Opacity Opaque Semi-Transparent Text Background Color Black White Red Green Blue Yellow Magenta Cyan Opacity Opaque Semi-Transparent Transparent Caption Area Background Color Black White Red Green Blue Yellow Magenta Cyan Opacity Transparent Semi-Transparent Opaque Font Size 50% 75% 100% 125% 150% 175% 200% 300% 400% Text Edge Style None Raised Depressed Uniform Drop shadow Font Family Proportional Sans-Serif Monospace Sans-Serif Proportional Serif Monospace Serif Casual Script Small Caps Reset restore all settings to the default values Done Close Modal Dialog End of dialog window. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like How Much Money Should You Have Before Hiring a Financial Advisor? SmartAsset Learn More Undo "Our vision is that, over time, AI would become part of the core infrastructure of higher education," Leah Belsky, OpenAI's vice president of education, said in an interview. In the same way that colleges give students school email accounts, she said, soon "every student who comes to campus would have access to their personalized AI account." To spread chatbots on campuses, OpenAI is selling premium AI services to universities for faculty and student use. It is also running marketing campaigns aimed at getting students who have never used chatbots to try ChatGPT. Live Events Some universities, including the University of Maryland and California State University, are already working to make AI tools part of students' everyday experiences. In early June, Duke University began offering unlimited ChatGPT access to students, faculty and staff. The school also introduced a university platform, called DukeGPT, with AI tools developed by Duke. Discover the stories of your interest Blockchain 5 Stories Cyber-safety 7 Stories Fintech 9 Stories E-comm 9 Stories ML 8 Stories Edtech 6 Stories OpenAI's campaign is part of an escalating AI arms race among tech giants to win over universities and students with their chatbots. The company is following in the footsteps of rivals like Google and Microsoft that have for years pushed to get their computers and software into schools, and court students as future customers. The competition is so heated that Sam Altman, OpenAI's CEO, and Elon Musk, who founded the rival xAI, posted dueling announcements on social media this spring offering free premium AI services for college students during exam period. Then Google upped the ante, announcing free student access to its premium chatbot service "through finals 2026." OpenAI ignited the recent AI education trend. In late 2022, the company's rollout of ChatGPT, which can produce human-sounding essays and term papers, helped set off a wave of chatbot-fueled cheating. Generative AI tools like ChatGPT, which are trained on large databases of texts, also make stuff up, which can mislead students. Less than three years later, millions of college students regularly use AI chatbots as research, writing, computer programming and idea-generating aides. Now OpenAI is capitalizing on ChatGPT's popularity to promote the company's AI services to universities as the new infrastructure for college education. OpenAI's service for universities, ChatGPT Edu, offers more features, including certain privacy protections, than the company's free chatbot. ChatGPT Edu also enables faculty and staff to create custom chatbots for university use. (OpenAI offers consumers premium versions of its chatbot for a monthly fee.) OpenAI's push to AI-ify college education amounts to a national experiment on millions of students. The use of these chatbots in schools is so new that their potential long-term educational benefits, and possible side effects, are not yet established. A few early studies have found that outsourcing tasks like research and writing to chatbots can diminish skills like critical thinking. And some critics argue that colleges going all-in on chatbots are glossing over issues like societal risks, AI labor exploitation and environmental costs. OpenAI's campus marketing effort comes as unemployment has increased among recent college graduates -- particularly in fields like software engineering, where AI is now automating some tasks previously done by humans. In hopes of boosting students' career prospects, some universities are racing to provide AI tools and training. California State University announced this year that it was making ChatGPT available to more than 460,000 students across its 23 campuses to help prepare them for "California's future AI-driven economy." Cal State said the effort would help make the school "the nation's first and largest AI-empowered university system." Some universities say they are embracing the new AI tools in part because they want their schools to help guide, and develop guardrails for, the technologies. " You're worried about the ecological concerns. You're worried about misinformation and bias," Edmund Clark, the chief information officer of California State University, said at a recent education conference in San Diego. "Well, join in. Help us shape the future." Last spring, OpenAI introduced ChatGPT Edu, its first product for universities, which offers access to the company's latest AI. Paying clients like universities also get more privacy: OpenAI says it does not use the information that students, faculty and administrators enter into ChatGPT Edu to train its AI. (The New York Times has sued OpenAI and its partner, Microsoft, over copyright infringement. Both companies have denied wrongdoing.) Last fall, OpenAI hired Belsky to oversee its education efforts. An ed tech startup veteran, she previously worked at Coursera, which offers college and professional training courses. She is pursuing a two-pronged strategy: marketing OpenAI's premium services to universities for a fee while advertising free ChatGPT directly to students. OpenAI also convened a panel of college students recently to help get their peers to start using the tech. Among those students are power users like Delphine Tai-Beauchamp, a computer science major at the University of California, Irvine. She has used the chatbot to explain complicated course concepts, as well as help explain coding errors and make charts diagraming the connections between ideas. "I wouldn't recommend students use AI to avoid the hard parts of learning," Tai-Beauchamp said. She did recommend students try AI as a study aid. "Ask it to explain something five different ways." Belsky said these kinds of suggestions helped the company create its first billboard campaign aimed at college students. "Can you quiz me on the muscles of the leg?" asked one ChatGPT billboard, posted this spring in Chicago. "Give me a guide for mastering this Calc 101 syllabus," another said. Belsky said OpenAI had also begun funding research into the educational effects of its chatbots. "The challenge is, how do you actually identify what are the use cases for AI in the university that are most impactful?" Belsky said during a December AI event at Cornell Tech in New York City. "And then how do you replicate those best practices across the ecosystem?" Some faculty members have already built custom chatbots for their students by uploading course materials like their lecture notes, slides, videos and quizzes into ChatGPT. Jared DeForest, the chair of environmental and plant biology at Ohio University, created his own tutoring bot, called SoilSage, which can answer students' questions based on his published research papers and science knowledge. Limiting the chatbot to trusted information sources has improved its accuracy, he said. "The curated chatbot allows me to control the information in there to get the product that I want at the college level," DeForest said. But even when trained on specific course materials, AI can make mistakes. In a new study -- "Can AI Hold Office Hours?" -- law school professors uploaded a patent law casebook into AI models from OpenAI, Google and Anthropic. Then they asked dozens of patent law questions based on the casebook and found that all three AI chatbots made "significant" legal errors that could be "harmful for learning." "This is a good way to lead students astray," said Jonathan S. Masur, a professor at the University of Chicago Law School and a co-author of the study. "So I think that everyone needs to take a little bit of a deep breath and slow down." OpenAI said the 250,000-word casebook used for the study was more than twice the length of text that its GPT-4o model can process at once. Anthropic said the study had limited usefulness because it did not compare the AI with human performance. Google said its model accuracy had improved since the study was conducted. Belsky said a new "memory" feature, which retains and can refer to previous interactions with a user, would help ChatGPT tailor its responses to students over time and make the AI "more valuable as you grow and learn." Privacy experts warn that this kind of tracking feature raises concerns about long-term tech company surveillance. In the same way that many students today convert their school-issued Gmail accounts into personal accounts when they graduate, Belsky envisions graduating students bringing their AI chatbots into their workplaces and using them for life. "It would be their gateway to learning -- and career life thereafter," Belsky said.
Time of India
an hour ago
- Time of India
H&M owners quietly buying its shares
Hennes & Mauritz (H&M), the fast-fashion retailer that's been listed on the Swedish stock market since 1974, is steadily moving back towards private ownership. The founding family has stepped up purchases of H&M shares, spending more than $6.6 billion since 2016 to amass nearly two-thirds control and fuelling speculation it could take the Stockholm-based company back into private hands - despite denials from family members. The Perssons, one of Sweden's wealthiest families, has ramped up insider buying by reinvesting dividends, boosting its H&M stake to almost 64% from 35.5% over the past nine years through a holding company Ramsbury Invest, saying little about their intentions other than that they "believe" in H&M, founded in 1947 by Erling Persson. Stay informed with the latest business news, updates on bank holidays and public holidays . AI Masterclass for Students. Upskill Young Ones Today!– Join Now
Time of India
2 hours ago
- Time of India
Govt appoints RBI DG T Rabi Sankar as 16th Finance Commission member
NEW DELHI: Reserve Bank of India (RBI) deputy governor T Rabi Sankar was Saturday appointed as a part time member of the 16th Finance Commission, a finance ministry statement said. He will hold office from the date of assuming charge until the submission of the commission's report or Oct 31, 2025, whichever is earlier. Sankar's appointment is to fill the vacancy created by the resignation of one of the full-time members of the commission, Ajay Narayan Jha, on personal grounds. The commission was constituted on Dec 31, 2023. The commission is mandated to submit its recommendations by Oct 31, 2025, for the award period of five years, commencing from April 1, 2026. Stay informed with the latest business news, updates on bank holidays and public holidays . AI Masterclass for Students. Upskill Young Ones Today!– Join Now
