Late-night chaos! OpenAI CEO personally reveals 72-hour palace intrigue: AGI will reprice $BTC, 8 billion people each with an AI agent—are you still chasing memes?

Have you heard? OpenAI’s second-in-command, Greg Brockman, recently spilled all the secrets of his ten-year journey on a podcast. No pleasantries, no PR spin, just pure substance.

He mentioned that in 2015, when he resigned from Stripe, everyone thought he was crazy. At that time, DeepMind was the AI industry’s giant gorilla, monopolizing funding, talent, and data. But he and Otman exchanged a look and decided, “We have to do this,” and the next day, he committed full-time.

The initial core team was small. Dario Amodei went to Google Brain, leaving only Sutskever, Shulman, and himself. Ten top researchers were watching, asking, “Who else will join?” To break the deadlock, Otman suggested an off-site event. Brockman organized a meetup in Napa, printed T-shirts in advance, and brainstormed a roadmap that continues to this day: first reinforcement learning, then unsupervised learning, and finally gradually tackling complex tasks.

A key milestone was 2017. They projected the costs of AGI and found it required an unprecedented scale of data centers. As a non-profit, their fundraising cap was too low. Musk, Otman, Sutskever, and he reached a consensus: establishing a for-profit entity was the only way out.

The real technological breakthrough came with the paper “Unsupervised Sentiment Neurons.” Just by training a model to predict the next character, it spontaneously understood sentiment polarity. Brockman said that was the first time he realized—machines are not just learning grammar but understanding semantics. When testing GPT-4, someone asked, “Why isn’t this AGI yet?” Using standards from two months ago, GPT-4 already fully met them. It can fluently discuss any topic, though some traits are still missing, but the economic transformation driven by computing power is already real.

Regarding the relationship between “predicting the next word” and reasoning, Brockman straightforwardly said they are essentially the same. The core of prediction isn’t about repeating known data but about inferring the future in unseen situations. Intelligence, prediction, and compression are the same thing.

When discussing the internal power struggles of 2023, he publicly revealed details for the first time. On the day Otman was fired, he received a message requesting a video call. All board members were online besides Otman. They informed him Otman was dismissed, and he was also removed from the board, but they hoped he would stay. He said he wasn’t angry at the time, just felt the logic was off.

After resigning, he received massive support. Core members like Jacob, Simone, Alexander, and others resigned one after another. Along with Otman, they quickly drafted a new company blueprint. He and Otman designed a backup plan—codenamed “Phoenix.” Satya Nadella from Microsoft got deeply involved, discussing funding new projects and onboarding everyone. Before Thanksgiving, employees who were supposed to fly home canceled their flights, and the office was packed with people.

The petition demanding the board’s resignation was signed by so many that Google Docs crashed. It wasn’t until Monday morning that Sutskever publicly supported unity on Twitter, and Brockman felt OpenAI could get back on track. He and Sutskever later spent a lot of time talking openly, airing out all the pent-up issues, ultimately repairing their relationship.

Throughout the weekend, competitors circled like vultures, trying to poach top talent with high salaries. But surprisingly, no one left. Brockman quoted football coach Bill Belichick: top players don’t play for money but for “the person next to them.” That was the true diamond moment—under extreme pressure, the team凝erged into the toughest whole.

On the technical front, he firmly believes AI evolution is a victory of “massive compute + simple algorithms.” From Dota to GPT-4, this logic has been repeatedly validated. He revealed an astonishing fact: “Almost all of OpenAI’s code is now written by AI.” Human experts still excel in architecture and modular design, but the underlying code is mostly handled by AI. AI can even generate novel ideas humans haven’t thought of—optimizing circuits in chip design, deriving elegant formulas in quantum physics contrary to academic expectations.

Compute power is the real bottleneck, and it’s also the most ridiculed yet most correct bet for OpenAI. While everyone debates products, they quietly build data centers. He predicts that super-large data centers dedicated to solving single problems (like cancer) could appear “as early as this year.”

Regarding the global race, Brockman prefers to call it a “global AI renaissance.” The US must balance export controls with maintaining leadership. True leadership isn’t just about leading the pack but about guiding the world to build consensus. He said that attempts by competitors to “distill” and steal成果 overlook a core point: AI development is exponential. Every time we are analyzed, we have already moved on to the next generation of more powerful models. The real moat isn’t any single model but the machine that continuously produces top-tier models.

The core goal of regulation must be “benefiting humanity.” When traditional career paths become unstable, technology must improve everyone’s quality of life. He clarified a common misconception: claims that data centers consume excessive water are mostly false. They use closed-loop cooling systems, with minimal and recycled water, and have committed not to raise residents’ electricity prices.

Brockman’s ultimate goal is to enable 8 billion people worldwide to have their own personal AGI. It understands your background, works for you around the clock, and comprehends your long-term goals. You still set the objectives, but it is a deeply aligned agent system focused on your long-term well-being.

We are entering the “compute economy” era. Software engineering is being redefined, and humans are shifting from “operators” to “vision managers.” As compute becomes a scarce resource, whoever controls data centers controls the future. For Bitcoin and the entire crypto market, AI-driven compute demand could reshape the value logic of energy and digital assets—this isn’t a fantasy but a reality already unfolding.

Follow me: for more real-time analysis and insights into the crypto market! $BTC $ETH $SOL

#WCTC交易王PK #Crypto market volatility continues #rsETH attack aftermath