Sam Altman at TED 2025: Inside the most uncomfortable — and important — AI interview of the year

Openai Sam Altman CEO has revealed that his company has grown to 800 million users per week and faces “incredible” growth rates, during a tense interview at the TED 2025 conference in Vancouver last week.

“I have never seen a growth in any company, in which or not, like this,” said Altman to Ted President Chris Anderson during his conversation on the stage. “ChatGPT – it’s really fun. I feel deep proud. But it is crazy to live through it, and our difference is exhausting and tightening.”

The interview, which closed the last day of TED 2025: Humanity has been reinforced, not only offered the loud Openai’s success but also the increasing scrutiny the company faces as its technology turns society at the rate of even some of its supporters.

https://www.youtube.com/watch?

Our graphics processing units are melting: Openai is struggling to expand amid an unprecedented request

Altman drew a picture of a struggling company to keep pace with its success, noting that Openai graphics units are “melting” because of the popularity of new image generation features. He said: “Throughout the day, I invite people and beg them to give us their graphics processing units. We are very restricted.”

This si growth comes as Openai is considering launching its own social network to compete with Elon Musk’s X, according to CNBC. Altman did not confirm and did not deny these reports during a TED interview.

The company recently closed a $ 40 billion financing round, as it was evaluated at $ 300 billion – which is the largest special financing for technology in history – and this flow in capital is likely to help in facing some of these infrastructure challenges.

Without profitability to $ 300 billion: Altman responds to the accusations of “Ring of Power”

During the 47 -minute conversation, Anderson repeatedly pressed Openai from a non -profit research laboratory to a profit company with a value of $ 300 billion. Anderson expressed his concerns that critics are shared, including Elon Musk, who suggested that the Taman was “damaged through the link of power”, referring to the “master of the rings.”

Altman defended Openai’s path: “Our goal is to make AGI and distribute it, and make it safe to make a wide benefit of humanity. I think by all standards, we have done a lot in this direction. It is clear that our tactics had turned over time … We did not think we had to build a company about this.

When asked how to deal personally with the tremendous power he is now, Altman replied: “Shockingly, as before. I think you can get used to anything step by step … you are the same person. I’m sure I am not in all kinds of ways, but I don’t feel anything different.”

“Division of Revenue”: Openai plans to push artists whose methods are used by artificial intelligence

One of the most concrete policy ads from the interview is Altman that Openai is working on a system to compensate for artists whose patterns are simulated by artificial intelligence.

“I think there are incredible new business models, we and others are excited to explore it,” said Altman when he was pressed about the IP stealing in the images created from artificial intelligence. If I say, “I want to generate art in the style of these seven people, who all agreed to do so,” How do you divide the amount of money that you reach each? “

Currently, Openai’s photo generator rejects requests to imitate the pattern of living artists without approval, but it will generate art in the style of movements, species or studios. Altman suggested that the revenue sharing model may be imminent, although the details remain rare.

Independent artificial intelligence factors: faced “the most dependent safety challenge” Openai

The conversation became particularly tense when discussing “Auctic AI” – independent systems that can take online action on the user. The new “operator” tool from Openai of artificial intelligence allows tasks such as reservation restaurants, and increase concerns about safety and accountability.

Anderson challenged Altman: “One person can allow the agent there, and the agent can decide,” well, for implementation of this job, I got psychological copies everywhere. “Are there red lines that were clearly painted internally, where you know what are the moments of danger?”

Altman referred to Openai’s “alert frame”, but he provided some details on how to prevent the company from abuse independent agents.

“Amnesty International, which allows access to your systems, information, and the ability to click on your computer … when they make a mistake, they are much higher risks,” Altman admitted. “Our agents will not use if you do not trust that they will not empty your bank account or delete your data.”

“14 definitions of 10 researchers”: within the Openai struggle to select AGI

At a moment, Altman admitted that even inside Openai, there is no consensus on what constitutes the artificial general intelligence (AGI) – the company’s declared goal.

“It is like a joke, if you have 10 researchers in Openai in a room and ask Agi, you will get 14 definitions,” Altman said.

I suggest that instead of focusing on a specific moment when AGI arrives, we must realize that “models will become more intelligent, more capable, more intelligent and more capable of this long chy … We will have to compete and obtain great benefits from this amazing system.”

Reducing handrails: the new Openai approach in moderate content

Altman also revealed a major change in politics regarding moderate content, and revealed that Openai has reduced restrictions on photo generation models.

“Users have given us more freedom about what we were traditionally thinking about being hurt by speaking,” he explained. “I think part of the model alignment follows what the model user wants to do within the extensive limits of what society decides.”

This shift may indicate a broader step towards giving users more control of artificial intelligence outputs, and may be in line with Altman’s preference for allowing hundreds of millions of users – instead of “elite small peaks” – determining the right handrail.

Al -Tamman said: “One of the wonderful new things about artificial intelligence is that our AI can talk to everyone on Earth, and we can learn to prefer the collective value of what everyone wants, instead of having a group of people who enjoy society by sitting in a room and making these decisions.”

“My child will not be more intelligent than artificial intelligence”: Altman’s vision for an Amnesty International Future

The interview with the world concluded by the world inherited by his newborn son – one of them will exceed artificial intelligence.

He said: “My child will not be smarter than artificial intelligence. They will never grow up in a world where products and services are incredibly smart and incredibly able.” “It will be a world of amazing material … where the change rate is incredibly fast and amazing new things that happen.”

Anderson closed with a realistic note: “During the next few years, you will have some of the biggest opportunities, the largest ethical challenges, and the largest decisions that must be made from any human being in history.”

A budget law billion users: How Openai transports power, profit and purpose

The appearance of Altman’s Ted comes at an embarrassing turn for Openai and the broader artificial intelligence industry. The company faces the increasing legal challenges, including copyright lawsuits, authors and publishers, while simultaneously pays the limits of what artificial intelligence can do.

Recent developments such as Chatgpt viral images and Sora video generation feature showed a few months ago. At the same time, these tools have sparked discussions on copyright, originality and the future of creative work.

Taman’s willingness to interact with difficult questions about safety, ethics and societal influence of Amnesty International shows awareness of the dangers concerned. However, critics may notice that concrete answers to specific guarantees and policies have remained out of reach throughout the conversation.

The interview also revealed the competing tensions at the heart of the Openai mission: move quickly to progress in artificial intelligence technology while ensuring safety; Achieving a balance between profit motives with social benefit; Respecting creative rights while adding a democratic character to creative tools; And the movement between the experience of the elite and the general preference.

Anderson also noted in his final comment, the decisions made by Altman and his peers in the coming years may have unprecedented effects on the future of humanity. Whether Openai can upgrade to the declared level of its declared mission of ensuring “all the beneficiaries of mankind from artificial public intelligence” still must be seen.