Sam Altman and Chris Anderson on TED, AI Creativity, and the Questions That Follow

If you want to understand what Sam Altman and Chris Anderson actually discussed at TED, the conversation was not a sweeping prophecy about the future in the abstract. It was more immediate than that, and in some ways more revealing: the latest OpenAI models are already producing images, video, diagrams, and writing that feel surprisingly capable, and that raises practical questions about work, creativity, consent, and how society should respond.

Chris Anderson opened by welcoming Altman to TED and noting that OpenAI has been releasing “crazy insane new models” so frequently that it feels like something new arrives every other week. Anderson had been playing with a couple of them himself, and he used the stage to show what they could do.

## Sora and the surprising force of a prompt

The first example Anderson showed was Sora, the image and video generator. He had asked it: “What will it look like when you share some shocking revelations here at TED?” The result, shown to the audience, was funny and unexpectedly convincing. Anderson joked about the image’s details, including the fact that it had five fingers on all hands, and asked Altman how he would grade it.

Altman said it was “very close to what I’m wearing,” and Anderson replied that he had never seen him quite that animated. Altman answered, “No, I’m not that animated of a person.” The exchange made the demo feel less like a technical showcase and more like a live test of how far these systems have come in producing images that feel tailored to a moment.

Anderson gave it a B-plus, but the more striking example came next.

## A diagram about intelligence and consciousness

Anderson said he had asked the model to come up with a diagram showing the difference between intelligence and consciousness. He described the result as simple, but incredible. What impressed him was not just that it made an image, but that it seemed to understand the conceptual request at a deeper level.

He then asked what kind of process could allow that. His point was that this was clearly not just image generation in the narrow sense. It seemed to be linking into the core intelligences of the broader model.

Altman’s answer was direct: the new image generation model is part of GPT-4o, so it has all of the intelligence in there. He said that is one reason it has been able to do the things people really love. In other words, the image system is not operating as a separate toy layered on top of intelligence; it is connected to the model’s broader capabilities.

That distinction mattered throughout the rest of the conversation, because the examples Anderson raised were not just about pretty outputs. They were about systems that can reason, interpret prompts, and generate content in ways that affect how people think about their own work.

## What happens when AI starts doing your job?

Anderson raised the concern from the perspective of a management consultant playing with these tools and wondering, “uh oh, what does my future look like?” Altman said there are two ways to look at that question.

One view is fear: the system is doing everything you do, so what happens to you? The other view is the one that has accompanied every major technological revolution in history: now there is a new tool, so what can you do with it? Altman argued that while the expectation for someone in a particular job will increase, the capabilities will increase so dramatically that it will be easy to rise to that occasion.

The point was not that change will be painless. It was that the same technology that raises the bar can also expand what people are able to accomplish. Altman framed it as a familiar pattern: new tools change what is expected, but they also expand what is possible.

## Charlie Brown, AI, and the question of meaning

Anderson then moved to another example that had impressed him. He said he asked the model to imagine Charlie Brown as thinking of himself as an AI, and it came up with something that Anderson found “rather profound.”

Altman laughed and said the writing quality of some of the new models, not just in this case but in detail, is reaching a new level. But he added an important caution: there is really no way to know whether the model is actually “thinking that” or whether it simply saw that pattern many times in the training set.

He then made a broader point: if you can’t tell the difference, how much do you care? That was one of the more quietly provocative moments in the discussion. The issue was not just whether the output is good. It was whether the source of the output matters if the result is compelling enough.

## Does this look like IP theft?

Anderson pushed the issue further, saying that at first glance the Charlie Brown example looks like IP theft. He asked whether OpenAI had a deal with the “Peanuts” estate.

The audience responded with applause, laughter, and murmuring. Anderson said people could clap about that all they wanted, and that he would enjoy it. But he then shifted to the larger issue: the creative spirit of humanity is incredibly important, and the goal should be to build tools that lift that up, enabling new people to create better art, better content, and better novels that everyone enjoys.

Altman agreed with that basic principle. He said he believes very deeply that humans will be at the center of this. He also said that society probably needs to figure out some sort of new model around the economics of creative output.

His view was that people have been building on the creativity of others for a long time, and people take inspiration for a long time. But as access to creativity becomes incredibly democratized and people build off each other’s ideas all the time, there may be new business models that OpenAI and others are excited to explore. He said exactly what that will look like, he is not sure.

He drew a line between obvious violations and harder questions. There is clearly some cut-and-dry stuff, such as not copying someone else’s work. But beyond that, the boundaries are less obvious. If someone says, “I want to generate art in the style of these seven people,” and all seven have consented, how do you divide up how much money goes to each one? Altman described these as big questions.

His broader view was that every time throughout history we have put better and more powerful technology in the hands of creators, we collectively get better creative output, and people do more amazing things.

## Consent, style, and where the line should be

Anderson then raised an even bigger question: what happens when the people involved have not consented? He pointed to a previous example from the conference, where Carole Cadwalladr had shown ChatGPT being prompted to “give a talk in the style of Carole Cadwalladr.” The result was impressive, though not as good as her own talk, and she had said the key issue was that she had not consented to it.

Anderson asked how society is going to navigate that. Should it only be people who have consented? Or should there be a model where any named individual whose work is used gets something for it?

Altman answered that, at least right now, if you use OpenAI’s image generation tool and ask for something in the style of a living artist, it won’t do that. But if you ask for something in the style of a particular vibe, or a studio, or an art movement, it will.

He also said that if you want to output a song that is like a copy of a song, it won’t do that. The question, then, is where that line should be and how people decide when something crosses it. Altman said this is a question society has already sorted out in another context, through copyright law and ideas about fair use. He believes that in the world of AI, there will be a new model that people figure out.

## Creative fear, creative excitement