Arrangement: New
On June 19, OpenAI officially released its first podcast, in which CEO Sam Altman systematically responded for the first time to a series of questions about the pace of GPT-5 advancement, the Stargate project, the development of next-generation AI terminal devices, the controversy over model memory capabilities, and the evolution of social structure after the arrival of AGI.
Altman talked about his real experience of using AI in parenting and education as a "new father". He also revealed the core choice OpenAI is facing from the perspective of a corporate decision maker: how to maintain a balance between technological leaps, privacy boundaries and trust structures.
"My children will never be smarter than AI, but they will grow up to be much stronger than our generation." Altman admitted on the show that this generation of children will grow up in a world where AI is fully infiltrated, and their dependence on, understanding and interaction with intelligent systems will be as natural as the previous generation's habit of smartphones. The new role of models such as ChatGPT in family companionship and knowledge enlightenment has opened up a new paradigm for parenting, education, work and creativity development.
AI is becoming the growth environment for the next generation
Altman mentioned that although society has not yet formed a unified definition, "every year more and more people believe that we have reached an AGI system." In his view, the public's demand for hardware and software is changing extremely rapidly, and current computing power is far from meeting potential needs.
When the conversation turned to Altman's new fatherhood, he admitted that ChatGPT provided a huge help in the early stages of parenting. "Although many people were able to raise their children well without ChatGPT, I'm not sure I can do it." After the first few weeks of "asking about everything", he gradually focused his questions on the baby's development rhythm and behavioral habits. He pointed out that this type of AI tool has begun to take on the role of "information intermediary" and "confidence enabler" in parenting.
Not only that, Altman is also thinking about the impact of AI on the growth path of the next generation. He said bluntly, "My children will never be smarter than AI, but they will grow up to be much stronger than our generation." He emphasized that this generation of children will naturally grow up in an environment where AI is everywhere, and their dependence on and interaction with AI will be as natural as smartphones in the past decade.
Altman shared a story that was circulated on social media: In order to avoid repeating the plot of "Thomas the Tank Engine" to his child, a father imported the character into the voice mode of ChatGPT, and the child talked to it for more than an hour. This phenomenon caused Altman to worry deeply: the extension of AI in companion roles may cause the alienation of "quasi-social relationships", which will in turn pose new challenges to social structure. He emphasized that society needs to reset its boundaries, but also pointed out that society has always found ways to deal with the impact of new technologies in history.
In the field of education, Altman observed the positive potential of ChatGPT in the classroom. "With good teachers and good courses, ChatGPT performs very well," but he also admitted that when students use it alone to do homework, it is easy to degenerate into "Google-style copying." He cited his own experience as an example, pointing out that people were also worried that "he only knows Google", but in the end they found that both children and schools can quickly adapt to the changes brought about by the new tool.
When asked about what ChatGPT will look like in five years, Altman said, "ChatGPT in five years will become something completely different." Although the name may remain, its capabilities, interaction methods, and positioning will change fundamentally.
AGI is a dynamic definition, and Deep Research’s capabilities are leaping forward
When talking about the industry buzzword "AGI", Sam Altman gave a more dynamic explanation. He pointed out, "If you asked me or someone else five years ago to define AGI based on the cognitive capabilities of the software at the time, the definition given at that time has been far surpassed today." As the intelligence of the model continues to increase, the standard of AGI is constantly being raised, showing a state of "dynamic advancement".
He emphasized that there are systems that can significantly improve human work efficiency and perform tasks with economic value. What is really worth asking is: what kind of system can be called "super intelligence"? In his opinion, systems that have the ability to make autonomous scientific discoveries or can greatly improve the efficiency of human scientific discoveries are close to this standard. "This will be a very good thing for the world."
This judgment has also been reflected within OpenAI. Andrew Mane recalled that when they tried GPT-4, they felt that "a decade of exploration space had been opened up." In particular, the moment when the model was able to call itself and demonstrate preliminary reasoning capabilities made people realize the possibility of a new stage.
Altman agreed with this and further pointed out: "I have always believed that the core driving force for improving the quality of human life is the speed of scientific progress." The slowness of scientific discovery is the fundamental factor limiting human development, and the potential of AI in this regard has not yet been fully released. Although he admitted that he has not yet mastered the complete path of "AI automatic scientific research", the research team's confidence in the direction of progress is rapidly increasing. He shared that from GPT-4.0.1 to GPT-4.0.3, a new key idea can be proposed every few weeks, and almost all of them work. This rhythm is exciting and confirms the belief that "breakthroughs will come suddenly."
Andrew Mane added that OpenAI recently switched the default model to GPT-4.0.3, and the most important update was the introduction of Operator mode. In his opinion, many Agentic systems in the past, despite their high promise, were not "anti-fragile" enough and crashed when they encountered anomalies. The performance of GPT-4.0.3 is very different. Altman responded that "many people told me that they felt the breakthrough moment of AGI was the Operator mode of GPT-4.0.3." Although he himself did not have a particularly strong feeling, the feedback from external users is worth paying attention to.
The two further discussed the new capabilities brought by "Deep Research". Andrew said that when he used this tool to research Marshall McLuhan, AI could search, filter, and organize materials online and generate a complete data package, which was more efficient than manual research. He also developed an app to generate audio files from questions to meet the needs of "limited memory but strong curiosity".
Altman then shared another extreme usage scenario: a "learning addict" used Deep Research to generate complete reports on various topics of interest, sitting there all day reading, questioning, and iterating, completely immersed in the AI-driven learning cycle.
Although Altman claims that he is unable to fully utilize these tools due to time constraints, he is still willing to prioritize reading the content generated by Deep Research in his limited time.
As functions continue to be enhanced and user scenarios become increasingly diverse, the outside world's attention to the next generation of models has also risen. Andrew directly raised the question that users are most concerned about: When will GPT-5 be released? Altman responded, "Maybe this summer, but I'm not sure about the exact time." He revealed that the company is facing a repeatedly discussed issue internally: Should the new version still be released in the form of "big fanfare" as in the past, or should it continue to iterate without changing the name, like GPT-4.
He further explained that today's model system structure is much more complex than in the past. It is no longer a linear process of "training once, launching once", but a dynamic system that supports continuous optimization. "We are now thinking about this question: if we continue to update GPT-5 after releasing it, should we call it GPT-5.1, 5.2, 5.3, or keep the name GPT-5?" Differences in user preferences also increase the complexity of decision-making: some users like snapshots, and some users hope to continue to improve, but the boundaries are difficult to unify.
Andrew pointed out that even people with a technical background can sometimes be confused about model selection, such as whether to use O3, O4 Mini, O4 Mini High, etc. The inconsistency of names exacerbates the difficulty of choosing.
In response, Altman gave a background explanation, saying that this is actually a "byproduct of paradigm shift." The current system is somewhat like running two sets of model architectures at the same time, but this chaotic state is coming to an end. He added that although he does not rule out the possibility of a new paradigm emerging in the future, which may cause the system to "split" again, "I am still looking forward to entering the GPT-5 and GPT-6 stage as soon as possible," when users will no longer be troubled by complex naming and model switching.
AI memory, personalization, and privacy controversy
Talking about the biggest experience change of ChatGPT recently, Sam Altman said frankly: "The memory function is probably my favorite new feature of ChatGPT recently." He recalled that when he first used GPT-3, the conversation with the computer was already amazing, but now the model can give accurate responses based on the user's background. This feeling of "knowing who you are" is an unprecedented leap. Altman believes that AI is opening a new stage. As long as the user is willing, it will have a deep understanding of the user's life and provide "extremely helpful answers" based on this.
However, functional evolution has also triggered more complex discussions at the social level. Andrew Mane mentioned that the New York Times recently filed a lawsuit against OpenAI, asking the court to force OpenAI to retain ChatGPT user data beyond the compliance period, which has attracted widespread attention. Altman said: "We will certainly oppose this request. I hope and believe that we will win." He criticized the other party for claiming to value privacy while making cross-border demands, and pointed out that this just exposes the current institutional gap regarding AI and privacy.
In Altman's view, although this lawsuit is regrettable, it also has the positive significance of "promoting society to seriously discuss AI and privacy." He emphasized that ChatGPT has become a "private conversation partner" in the daily lives of many users, which means that the platform must establish more serious institutional guarantees to ensure that sensitive information is not abused. He bluntly said: "Privacy must be a core principle for the use of AI."
The discussion further extended to data usage and advertising possibilities. Andrew questioned whether OpenAI could access user conversation data and whether the data would be used for training or commercial purposes. In response, Altman said that users can indeed choose to turn off the use of training data, and OpenAI has not yet launched any advertising products. He personally is not completely against advertising, "Some ads are good, for example, I have bought a lot of ads on Instagram." But he emphasized that in products like ChatGPT, "trust" is an extremely critical cornerstone.
Altman pointed out that social media and search platforms often make people feel "commoditized" and that content seems to exist for ad clicks. This structural problem is the source of widespread user concern. If the output of AI models is manipulated by advertising bids in the future, it will be a complete collapse of trust. "I would hate it myself."
On the contrary, he prefers to establish a "clear, transparent and consistent" business model: that is, users pay for high-quality services rather than being manipulated by hidden advertisements. Under controllable conditions, he does not rule out exploring models such as "platform commission after clicks" in the future, or displaying some practical advertisements in addition to output content, but the premise is that it will never affect the independence and reliability of the core output of the model.
Andrew expressed similar concerns and used Google as an example. He believes that the Gemini 1.5 model is excellent, but as an advertising-driven company, Google's underlying motivation makes it difficult to be completely assured. "I have no problem using their API, but when using chatbots, I always wonder: Is it really on my side?"
Altman expressed his understanding and admitted that he was also a loyal user of Google Search. "I really like Google Search." Although there are many advertisements, it was once "the best tool on the Internet." However, structural problems still exist. He praised the Apple model and believed that "paying for products in exchange for a clean experience" is a healthy logic. He also revealed that Apple had tried the advertising business iAd, but it was not successful. Perhaps it is not keen on this kind of business model in essence.
In their view, users also need to exercise judgment. "If we find that a product is suddenly "pushed very hard" one day, we have to ask one more question: What is the motivation behind it?" Andrew said. Altman added that no matter what business model OpenAI adopts in the future, it must always adhere to the principles of "extreme honesty, clarity, and transparency" to maintain the trust boundary of users in the platform.
Stargate, building a smart energy landscape
When the conversation turned to "The evolution of the relationship between AI and users", Altman first reviewed the structural errors of the social media era. He pointed out that "the most fatal problem of social platforms is the misaligned goals of recommendation algorithms - they just want you to stay longer, rather than really caring about what you need." The same risk may also appear in AI. He warned that if the model is optimized to "only cater to user preferences", it may seem friendly but may weaken the consistency and principles of the system, which will be harmful in the long run.
This deviation was already evident in DALL E 3. Andrew observed that early image generation had a distinct problem of single style, and although Altman did not confirm its training mechanism, he acknowledged the possibility. The two agreed that the new generation of image models had made significant improvements in quality and diversity.
The bigger challenge comes from the bottleneck of AI computing resources. Altman admitted that the biggest problem at present is that "we don't have enough computing power for everyone to use." For this reason, OpenAI launched Project Stargate. This is a global computing infrastructure financing and construction project, the goal of which is to integrate capital, technology and operational resources to create a computing platform of unprecedented scale.
"The core logic of Stargate is to lay a cost-controlled computing power base for intelligent services for all people." He explained that unlike any previous generation of technology, AI will have a huge infrastructure demand if it is to truly cover billions of users. Although OpenAI does not have a $500 billion budget in its account, Altman is confident in the implementation of the project and the performance of the partners, and revealed that its first construction site has started, accounting for about 10% of the total investment.
He was shocked by his on-site experience: "Although I know what a gigawatt-class data center is, when I actually saw thousands of people building a GPU computer room, I realized that the complexity of the system was beyond my imagination." He used the analogy of "no one can make a pencil alone" to emphasize the breadth of the industry mobilization behind Stargate, from mining, manufacturing, logistics to model calling, all of which are the ultimate embodiment of human engineering collaboration over thousands of years.
In the face of external doubts and interference, Altman responded for the first time to reports that Elon Musk tried to interfere with the Stargate project. He said, "I made a wrong judgment before. I thought Elon would not abuse the influence of the government to engage in unfair competition." He regretted this and emphasized that such behavior not only undermines industry trust, but is also not conducive to the overall development of the country. Fortunately, the government was not affected by it and stood firm on its legitimate position.
He is pleased with the current AI competition landscape. In the past, people generally had the anxiety of "winner takes all", but now more people realize that this is an ecological co-construction. "The birth of AI is very similar to the invention of transistors. Although it was only in the hands of a few people at the beginning, it will eventually form the foundation of the world's technology." He firmly believes that countless companies will create great applications and businesses based on this foundation, and AI is essentially a "positive-sum game."
When talking about the energy sources needed for computing power, Altman emphasized "all of them". Whether it is natural gas, solar energy, fission nuclear energy or future fusion technology, OpenAI must mobilize all means to meet the ultra-large-scale operation needs of AI systems. He pointed out that this is gradually breaking the geographical boundaries of traditional energy. Training centers can be deployed anywhere in the world with resources, and intelligent achievements can be disseminated at low cost through the Internet.
"Traditional energy cannot be dispatched globally, but intelligence can." In his view, this path of "converting energy into intelligence and then outputting it as value" is reshaping the entire human energy landscape.
This also extends to the field of scientific research. Andrew pointed out that the James Webb Space Telescope has accumulated a huge amount of data, but it is difficult to process due to the lack of scientists, resulting in a large number of "undeveloped scientific discoveries." In this regard, Altman imagines whether it is possible to have an AI that is smart enough in the future to deduce new scientific laws based on existing data without relying on new experiments or new equipment?
He mentioned that he once joked that OpenAI should build its own giant particle accelerator, but then he thought that perhaps AI could solve high-energy physics problems in a completely different way. "We have actually accumulated a lot of data, but the problem is that we don't yet understand the limits of intelligence itself."
In the field of drug discovery, such cases of "missing the known" are more frequent. Andrew mentioned that drugs such as Orlistat were discovered in the 1990s, but were shelved for decades due to limited perspectives and were not reused until today. Altman believes that "there may be a lot of such forgotten but valuable scientific materials, which can lead to huge breakthroughs with a little guidance."
Altman expressed great interest in the expectations for the next generation of models. He mentioned that Sora can understand classical physics, but whether it can advance deeper theoretical science remains to be verified. "The 'inference model' we are developing is expected to be the key to exploring this capability."
He further explained the difference between the inference model and the existing GPT series. "At the beginning, we found that as long as you tell the model to 'take it step by step', the quality of the answer will be greatly improved. This shows that the model has a potential reasoning path." The goal of the inference model is to enhance this ability in a systematic and structured way, so that the model can perform "internal monologue" like humans.
Andrew added the example of Anthropic using “thinking time” to evaluate model quality. Altman also expressed surprise: “I thought users hated waiting the most. But the fact is that as long as the answer is good enough, everyone is willing to wait.”
In his view, this is the watershed moment in the evolution of AI: no longer a mechanical response that pursues speed, but a move towards intelligent entities that truly understand, reason, and invent.
The next generation of hardware and the revolution of individual potential
Regarding OpenAI's hardware plans, Andrew mentioned the collaboration video between Sam Altman and Jony Ive, and asked directly whether the equipment has entered the trial stage.
Altman admitted that it was still early. He said that OpenAI has set a very high quality bar for this product, and it is not a goal that can be achieved in a short period of time. "The computers we use now, both hardware and software, are still essentially designed for a world without AI."
He pointed out that when AI can understand human context and make reasonable decisions on behalf of humans, the way people interact with machines will change completely. "You may want the device to be more sensitive, to be able to perceive the environment, and to understand the background of your life - you may also want it to be completely free of screens and keyboards." For this reason, they have been exploring new device forms and are very excited about some of the directions.
Altman described a new interaction paradigm - an AI that truly understands users and grasps context, can participate in meetings on behalf of users, understand content, manage information boundaries, contact relevant parties and promote decision execution. This will bring the relationship between people and devices into a new symbiotic state. "If you only say one sentence and it knows who to contact and how to act, the way you use computers will be completely different."
From the perspective of evolutionary logic, he believes that the current way we interact with ChatGPT is both "shaped by the device form" and "shaped the device form in turn." The two are in a continuous and dynamic co-evolution.
Andrew further pointed out that the popularity of mobile phones is largely due to their compatibility with "public use (screen viewing)" and "private use (voice calls)" scenarios. Therefore, the challenge for new devices is also: how to be "both private and universal" in a variety of scenarios. Altman agreed with this. He took listening to music as an example: using speakers at home and headphones on the street, this "public-private division" is natural. But he also emphasized that the new device form still needs to pursue stronger versatility to become a truly viable AI terminal.
When asked when we might see the product on the market, Altman didn't give a specific time, only saying that "it will take a while," but he believes that it will eventually "be worth the wait."
The conversation naturally transitioned to Altman's advice to young people. He said that the obvious strategic advice is: "Learn to use AI tools." In his opinion, "the world has quickly switched from 'you should learn to program' a few years ago to 'you should learn to use AI'." And this may still be just a phased transition, and he believes that new "key skills" will emerge in the future.
On a more macro level, he emphasized that many abilities that are traditionally considered "talents" or "character" can actually be trained and learned. Including resilience, adaptability, creativity, and even the intuition to recognize the real needs of others. "Although it is not as easy as practicing ChatGPT, these soft abilities can be trained through methods - and they will be extremely valuable in the future world."
When asked if he would give similar advice to a 45-year-old, Altman responded clearly: basically the same. Learning to use AI well in one's own professional scenario is a skill transfer challenge that must be addressed at any age.
Regarding the organizational changes after the arrival of AGI, Andrew raised a common question: "OpenAI is already so powerful, why is it still hiring?" He believes that some people mistakenly believe that AGI will directly replace everything. But Altman's answer is simple: "In the future, we will have more employees, but everyone's work efficiency will be far higher than before the AGI era."
He added that this is the essential goal of technological progress - not to replace humans, but to greatly enhance individual productivity. Technology is not an end, but a ladder to higher human potential.
