Written by: Techub News
At the recent Stripe Sessions conference held in San Francisco, OpenAI co-founder and CEO Sam Altman had a deep conversation lasting an hour with Stripe co-founder and CEO Patrick Collison. This conversation took place at a crucial moment when the development of AI technology seemed to be entering a "parabolic growth" phase. The two tech leaders not only reflected on the ups and downs of OpenAI over the past decade but also focused on how AI has evolved from a cutting-edge research area into a fundamental force that is about to reshape global infrastructure, business logic, and scientific discovery. Altman's insights provided a precious perspective for understanding OpenAI's strategic intentions and the future direction of the AI industry.
Perception of Singularity: From Codex Eruption to Ubiquitous AI Assistants
Patrick Collison opened with an interesting assertion: defining January 1 of this year as the beginning of the "singularity," which he noted was the 119th day. Sam Altman agreed, believing that we are currently in a technological "take-off" phase. This perception is not baseless; several metrics within OpenAI have experienced a "parabolic" inflection from late last year to early this year. Altman attributed this change to the capabilities of the model, particularly its coding abilities, which crossed a critical threshold at the end of last year and the beginning of this year.
"Codex is entering its shining moment," Altman said. He shared two personally perceived milestones: one came with GPT-5.2, and the past few weeks saw a giant leap that made him feel this is about to become his primary interface for interacting with computers. While the most steadfast users currently use AI for programming, a "wave" is driving more users to engage with it for various non-coding tasks. OpenAI's ambitions extend far beyond programming, aiming to cover all work done at computers. Altman estimated that in non-coding areas, they may have only completed 10% of the journey, but progress will be rapid with a real user base.
What will be the next area fully unlocked by AI after programming? Altman believes that programming may be a special case due to its immense societal demand and natural fit with the model. But the more fundamental change lies in people's realization of how much time they waste on tedious computer operations—switching between messaging apps, copying and pasting, responding to tedious emails, and so on. AI will take over these "drudgeries," thereby greatly enhancing people's work happiness and quality of life. The improvement in this subjective experience will be profound.
When discussing specific AI application experiences, Altman cited OpenAI's recently launched agent product, OpenClaw, sharing several "magic moments." As a home automation enthusiast, his attempts over the years to build better home automation systems had always failed, and OpenClaw is the first solution that satisfied him. He even created a messaging application he had always wanted to automate the unpleasant morning influx of messages. Even more surprisingly, while testing the new product, he had his AI agent use a disposable virtual card to buy himself a gift under $20 online, and the agent chose to purchase an HTTP status code design graphic on Gumroad. These experiences blurred the lines between tools and entities with "will," creating a wonderful and slightly eerie sensation. Altman even revealed that while planning a celebratory party for GPT-5.5, he asked the model for its thoughts, and the model provided a detailed, thoughtful, and humorous proposal, leading him to feel a kind of real "moral pressure."
The Evolution of OpenAI: From Research Lab to Global Intelligent Utility
The conversation naturally turned to OpenAI itself. This now globally renowned organization has been in existence for 11 years. Altman shared a previously untold "crazy" story from OpenAI's history: during the eight-month window between the completion of GPT-4's training and its official release, everyone within the company was using the model they believed would change the world, while the outside world knew almost nothing about it. This state of internal collective awareness and external feedback vacuum led them to doubt whether they were experiencing a "collective mental illness." Though not as dramatic as the later boardroom antics or lawsuits, those who experienced it felt it was extraordinarily peculiar.
Regarding management style, Sam Altman described himself as far from a micromanager. His philosophy is to find excellent people, give them a high-level direction, and let things happen naturally. He divided OpenAI's development into three stages: the first stage was a pure research company, aiming to explore the road to AGI that seemed "completely crazy" at the time; the second stage involved continuing research while learning how to build a product company; now they are entering the third stage, which is learning how to build a global, large-scale "intelligent token factory" based on the previous two.
"I believe what we are doing is building a new utility," Altman stated, "People will want to use large amounts of tokens and intelligence in various ways. We need to make these intelligences as smart, cheap, abundant, and usable as possible." This requires deep full-stack integration and massive infrastructure building. He admitted that during the transition from the first stage to the second, he did not fully realize how much his management style needed to change, and that the third stage will be entirely different, not naturally fitting for him. He is contemplating how to change himself, or to find the right person, or even building an AI that can manage this new venture.
In response to concerns from the outside about AI labs "consuming the entire value chain" and forming hegemonic power, Altman provided a clear response. He praised Stripe as an infrastructure provider whose success is highly aligned with customer success. He hopes that OpenAI can achieve a similar model: to become an infrastructure provider, a business with "perpetually low profit margins" but massive scale and rapid growth, providing a "smart metering device" for businesses to purchase for automating processes, accelerating internal operations, or building products. He believes that as AI becomes increasingly intelligent, the switching costs will decrease, making it difficult to maintain high profit margins. OpenAI's goal is to deeply align with the success of the globally distributed economic engine.
This naturally leads to another hot topic: the enormous investment in computing power. The scale of infrastructure building that Altman proposed two to three years ago, which sounded "absurd" at the time, is now increasingly becoming a reality. He acknowledged that this is likely to become the most expensive infrastructure project in human history. Fortunately, revenues are rapidly growing to match expenditures, and productivity improvements far exceed expectations. The key is that when the price of intelligent units is low enough, demand is almost limitless. "We won’t build a Dyson sphere and fill it with data centers... but maybe we will?" he said half-jokingly. He believes that we are not currently in a capital expenditure bubble, as the demand scale is reasonable compared to the investment scale. As for how to identify a true bubble, the former investor admitted he has never found a reliable framework, and economists "successfully predicted three out of the last eight recessions."
Talent, Collaboration, and the Shift in the Startup Ecosystem
In the AI field, which heavily relies on top talent, how to manage those charismatic yet possibly strong-willed "star" employees? Altman revealed that someone once summarized the "secret" to OpenAI's success for him: he succeeded in getting a group of individuals who all considered themselves unique or the most capable, and insisted on doing things their own way, to work together long enough to achieve breakthroughs. The secret is "a lot of pain." OpenAI relies on a few profound shared beliefs: believing in scaling, believing in concentrating resources, and believing that the endeavor being pursued is significant enough that people are willing to set aside personal grievances. For instance, during the training of GPT-3, the company devoted nearly all its computational power to a single research project, which was distinctly different from the culture of other labs, like DeepMind, that emphasized balanced distribution, but they chose to believe firmly and eliminate distractions.
Regarding the nearly ten years of successful collaboration with co-founder Greg Brockman, Altman believes that shared history, shared values, deep mutual respect, and complementary skill sets are key. He observed that at Y Combinator (YC), the length of time a founding team knows each other is an important predictor of success. He feels grateful to have such a profoundly trusted partner to share this intense entrepreneurial journey.
As one of the world's most successful venture investors, Altman was also asked whether the qualities of successful entrepreneurs in the AI era have changed. He humorously mentioned that the industry used to mock "idea kings" who had ideas without technical capability, but now the situation is reversing—"idea kings" are experiencing a "moment of revenge." While technical ability is still important, he is now very willing to invest in founders who deeply understand users but cannot code at all. This is a significant shift.
Facing the potential emergence of AGI or "singularity" in the coming years, how to approach venture capital with a ten-year cycle? Altman believes that it indeed requires a "true suspension of disbelief," but this may be the correct way to live. You cannot stop doing everything just because you believe the singularity is coming; you must live and plan as though the world will continue to operate in an understandable way for a long time. OpenAI’s own planning reflects this: they have signed a 20-year power and land agreement, with a clear vision for products within the next two years, but it becomes much fuzzier beyond that.
Regarding the persistence of startups built on AI (formerly known as "GPT wrappers," now mostly referred to as "AI applications" or "agent platforms"), Altman's viewpoint has remained consistent: a company should stand on the side of wishing for AI to get smarter. If your business is about patching a clear weakness in the current model, then you'll feel disheartened when the next, more powerful model is released. If your business thrives as the model's intelligence increases, then you'll be happy. Whether they are "wrappers" or "platforms," the principle is the same.
Empowering the Future: Science, Society, and OpenAI's Ultimate Vision
When asked which organizations are most capable of effectively utilizing AI, Altman cited a few examples. First is Shopify's CEO Toby Lutke, who was one of the first to announce that the company would "fully invest in AI" and took it upon himself to automate everything with AI, demanding that the team do the same. This kind of top-down, hands-on push from the highest level is remarkably effective. OpenAI even plans to attempt sending "technical experts" to work directly with company CEOs to automate their work, generating a "fractal effect." Second are those companies that are "uncomfortably permissive" with data access, allowing AI access to code repositories, meeting notes, Slack messages, emails, and all kinds of data. This poses a huge challenge for large companies with sensitive data, involving trade-offs between privacy and efficiency, which may require regulatory adjustments, but its power has been vividly demonstrated in small teams. Altman cited the blockchain project Tempo incubated by Stripe, where the team set up an agent tool in Slack that can nearly orchestrate everything in the company—reading documents, creating tasks, writing code, deploying, testing, all operations of a small organization are carried out in a single Slack channel, creating a stunning scene.
Regarding the future of open-source AI, Altman confidently stated, "Of course there is a future." Most demand is currently focused on more intelligent, faster, and cheaper cutting-edge models, but there is also significant demand for open-source, and he anticipates that its relative importance will increase over time.
Science is a field that excites Altman greatly and is a key focus for the OpenAI Foundation. He believes the most critical contribution of AI to human quality of life may lie in accelerating scientific discovery. Starting from models developed a few months ago, particularly GPT-5.5, these models are intelligent enough to help outstanding scientists generate better ideas and even make some small but important discoveries. In the future, combined with automated laboratories and robotics, the scientific process will accelerate significantly. If it is possible to complete what once took ten years of scientific work in a year, the compounding effect would be enormous. The OpenAI Foundation will invest funding, expertise, and technology to accelerate science, believing this will benefit the world in wonderful ways. Altman revealed that this could likely become one of the largest foundations in the world. They have already provided funding to the Arc Institute, which is committed to utilizing technologies like CRISPR and AI, with the goal of achieving the first cures for complex diseases such as Alzheimer's.
Finally, Patrick Collison asked, in a context where the development of AI seems to have a certain "inevitability," how Sam Altman hopes his personal involvement will change the world's trajectory. Altman's answer returned to one of OpenAI's core principles: iterative deployment. He recalled a mainstream viewpoint in the industry before the release of ChatGPT, which suggested that AI was too dangerous to be controlled by anything other than a small group of people thinking about AI safety, and that it was "an information hazard" that must be locked in an ivory tower. The prospect of such concentrated power made him deeply uneasy. He firmly believes that avoiding such concentration of power will enable the world to use and build upon this technology widely; although the process may be a bit messy, it will ultimately offer a gift to the world, upon which a larger gift will be built for everyone. He believes in entrepreneurial spirit, innovation, and that most people are kind, capable of doing amazing things with tools. Thus, he considers his greatest contribution to be pushing this technology to become a democratized technology for people to use and build upon. This may be the deepest interpretation of the name "OpenAI" in the new era.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
