Written by: Techub News Compilation

OpenAI co-founder and CEO Sam Altman recently participated in an in-depth interview with Axios co-founder Mike Allen, expressing candid and urgent views on the arrival of superintelligence, the disruptive impact of AI on society, and the lagging nature of the current global governance system. In this conversation, Altman rarefully delineated specific time frames (such as "next year") to describe the real threats posed by AI and elaborated on the deeper considerations behind OpenAI's recently released "social resilience" blueprint.

Urgent Warning: The Speed of Social Change Exceeds Expectations

At the onset of the interview, Mike Allen pointed out that the agenda Altman presented was filled with "urgency." Altman agreed, emphasizing that although these are just thought starters and not final solutions, the time to start serious discussions is now. He believes AI has already begun to handle an increasing number of practical tasks and that the transformations speculated about in recent years "seem to be beginning to happen."

When asked why the public doesn't yet seem to fully perceive the coming upheaval, Altman highlighted that one of OpenAI's historical roles is as an "educator," to warn the public about the future based on what is seen in the research pipeline, in an appropriate and uncertain manner. He acknowledged that predictions can sometimes be incorrect but believes it is the company's responsibility to society. He admitted that many past predictions were initially considered a sort of "more polite version of madness," yet many have proven to be important foresights.

Altman emphasized that the current generation of models has already had a real impact on science and the economy, and that the next generation of models will represent a "very significant leap." He called for people to start thinking now about how the economy and how social operations and work forms will change as a result.

Next Generation AI: From "Efficiency Multiplier" to "Disruptive Force"

Altman specifically outlined the potential scenarios the next generation of AI models might bring. On a positive note:

• Scientific Discovery: The current models are already helping scientists achieve "small discoveries," while the next generation of models could assist researchers in making "the most important discoveries of a decade or even a career," achieving career-defining breakthroughs.
• Knowledge Work: Programmers have reported productivity increases of 2 to 3 times. In the future, it might be possible for "one person plus hundreds of GPUs to accomplish the work of an entire software team."

However, he stressed more concerning aspects. OpenAI's "Preparedness Framework" currently mainly tracks three key areas: cybersecurity, biosecurity, and "any issues related to timing."

Cybersecurity: Altman made it clear, "I suspect we will see significant cyber threats next year that must be mitigated." Current models are already quite powerful and will become stronger. He agreed with assessments from other AI companies that a world-shattering cyberattack is "entirely possible" this year. To prevent this, a "resilience" approach is needed, involving substantial efforts not only to secure individual AI models but also to involve cooperation from defenders, cybersecurity companies, major platforms, governments, and the entire open-source tech stack.

Biosecurity: AI models will become very adept at helping people conduct biological research at high levels, leading to "wonderful results" in curing many diseases. However, at the same time, there will be those who attempt to misuse this technology. Currently, cutting-edge models remain in the hands of relatively responsible companies, which can mitigate risks through model alignment, classifiers, and safety measures. But Altman warned, "We are not far from a world with extremely powerful, biology-savvy open-source models." At that point, society will need to build resilience to cope with terror organizations utilizing those models to create new pathogens. "This is no longer merely a theoretical possibility, or rather, it won't remain merely theoretical for long."

It is based on these imminent specific risks that Altman emphasized that safety in a world of powerful AI cannot be solely ensured by companies; it necessitates promoting concepts of "social resilience."

Nationalization vs. Deep Collaboration: Governance Challenges in the AI Era

Given the overwhelming power exhibited by AI, Mike Allen posed a sharp question: why not nationalize OpenAI and its competitors?

Altman provided an answer from a historical perspective. He believes that in another era, creating artificial general intelligence (AGI) should have been a government-led project like the Apollo program, the Eisenhower highway system, or the Manhattan project. However, under the current operational dynamics of the world, he thinks that government projects "are unlikely to succeed."

The strongest argument against nationalization is that the U.S. needs to successfully build superintelligence in a manner that is "in keeping with American democratic values," and do so ahead of others. As a government project, that is likely not feasible. Altman candidly stated, "I think this is a regrettable situation."

Nonetheless, he firmly believes that the companies developing this technology and the government "must work extremely closely together." This is not a traditional regulatory relationship, but rather requires a "very, very deep partnership." He admitted that this depth of cooperation has not been realized yet, but OpenAI is making efforts to closely collaborate with the government.

When asked about the most urgently needed gaps to bridge, Altman emphasized that cybersecurity and biosecurity are top priorities. Additionally, building infrastructure and some economic ideas in the blueprint also need to be expedited.

AGI Is Within Reach? The Pace of Change Exceeds Imagination

Regarding how far we are from superintelligence or AGI, Altman gave a significant response: "We are close enough to AGI that a precise definition becomes important." Some believe it has already been reached, while others think we have not. But the key point is that AI is already discovering "truly novel" scientific knowledge and is "truly, at scale, engaged in serious, valuable economic work." Regardless of how you mark this point of AGI on the evolutionary curve, "we have come a considerable way along this curve," entering a "new paradigm."

He emphasized that the pace of change is "faster than people expect, and faster than people are prepared for." Currently, the "main areas" of knowledge work have already changed. He used programmers as an example: "Being a programmer at the beginning of 2025 versus the beginning of 2026 will mean something entirely different."

Trust and Responsibility: Personal Feelings at the Historical Crossroads

Faced with the question of "why should people trust you to lead this change," Altman first clarified that AI will not alter the core of "what it means to be human" – interpersonal relationships and the pursuit of happiness and other "the same things that have driven us for millennia" will remain unchanged. What will truly change is how the economy and social operations function.

Regarding trust, he believes that those building AI must be "highly upright and trustworthy," and that the vast majority of people in the industry feel the "seriousness" of what they are doing and carry this weight daily. Simultaneously, he emphasized, "no single person should make decisions that will affect all of us alone," and OpenAI believes in the democratization of AI, believing tools should be handed to people to collectively determine how they are used and to consider how to integrate them into life for great benefit.

Discussing personal changes brought about by wielding immense power, Altman described it as a "very peculiar experience." He shared a conversation with a member of the team responsible for ChatGPT's "post-training" (determining its conversational style): never before in human history have nearly a billion people talked to the same "virtual brain." The subtle adjustments to the model's style have impacts that no individual could reach.

He admitted that in everyday trivial work, he rarely feels he is doing "significant, historically-weighty things," but occasionally pulling back to think, he realizes that this is an "incredible honor," both cool and terrifying. He recalled the most intense feelings during the months after completing the training of GPT-4, before its release. Only a few people in the world knew about this "incredible thing," knew the enormous impact it would have, and understood that how it was released could steer the world in a better or worse direction. "Experiencing that period, it is impossible not to be changed by it — both the good and the bad, but mainly in good ways, albeit in various forms."

As for the "bad aspects," he simply stated, "I carry a lot of pressure, anxiety, and weight... I no longer have a very normal life."

Blueprint in Political Realities: The "Feasible" and the "Moon Landing" from Energy to Taxation

When the discussion turned to the political feasibility of his policy blueprint in today's Washington, Altman acknowledged that these ideas are not "Trump-friendly," and have even been compared to "Bernie Sanders' radical dreams." However, he indicated that he has discussed some ideas with individuals from both the Democratic and Republican parties, aiming to open a dialogue.

He was surprised by the acceptance across the board of the fact that "we must try some different approaches, and the sooner we start discussing the better." He cited an example of a person he knows who can be described as "the most steadfast libertarian Republican" saying, "I hate giving any space to these types of ideas, but capitalism relies on some balance between labor and capital. If AI develops the way I (referring to this person) think it will, then the leverage will tilt too much towards capital rather than traditional labor (though of course, new jobs will emerge). We must think of new ideas to keep capitalism thriving." Altman believes that if even such a person is willing to contemplate how to include more people in the capitalist system in the AI era, and that the way of implementation may need to change accordingly, then society as a whole "is already quite ready to engage in this dialogue." He revealed that this "brother" is a Republican senator or senior policy figure.

Specifically regarding the content of the blueprint, Altman analyzed the feasibility of different ideas:

• Most Tactically Viable: Seeking faster construction of more energy to keep electricity prices low, which has garnered considerable political support.
• Less Controversial: Issues related to privacy and AI, he believes "something will happen."
• More Complex: Topics regarding how much auditing should be done on frontier systems, which require balancing between "not wanting to over-regulate" and "not wanting to take big risks."
• Still on the Edge of the "Overton Window": Thoughts on major adjustments to the tax system or the economy.

When questioned whether all of this requires a "functioning government," which the U.S. seems never to have truly possessed, Altman displayed an optimistic demeanor. He believes that part of democracy is about needing to "manage systemic chaos," and that things are not always as easy as one would wish, but he assumes "we can manage it."

AI as a Utility: Intelligence as Accessible as Water and Electricity

At the end of the interview, Altman responded to inquiries about his vision of "AI as a utility." He clarified that this is not about specific pricing, but rather an analogy for a usage model.

Just as after paying for electricity, individuals can freely use it to power computers, ovens, toys, and any other devices at home, in the future, people will have a "personal super assistant" running in the cloud. It will complete a large number of tasks, connecting to various devices and products, logging into different services, and thinking and processing matters for you in the background. If you use it heavily or at a higher intelligence level, your monthly bill will be higher; using it less will result in a lower bill. You will have this intelligent entity that truly understands you and can accompany you.

Regarding pricing, Altman predicts:

• Unit Intelligence Cost: If sufficient infrastructure can be built, the price of unit intelligence will continue to decline "at an astonishing rate." He worries that the world has huge demand for this, which could lead to "very strange supply constraints" that push prices up, thus substantial construction should be undertaken.
• Premium for Frontier Intelligence: People will always be willing to pay more for cutting-edge, more powerful intelligence. "Last year's intelligence" will become cheap, but the price of frontier intelligence will remain premium until its cost also decreases.

Altman concluded with this vision, depicting a future where intelligence is as ubiquitous, cheap, and accessible as water and electricity, which is exactly the scenario OpenAI is striving to bring to fruition.

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。