OpenAI CEO Sam Altman revealed that his firm has grown to 800 million weekly energetic customers and is experiencing “unbelievable” progress charges, throughout a typically tense interview on the TED 2025 convention in Vancouver final week.
“I’ve by no means seen progress in any firm, one which I’ve been concerned with or not, like this,” Altman informed TED head Chris Anderson throughout their on-stage dialog. “The expansion of ChatGPT — it’s actually enjoyable. I really feel deeply honored. However it’s loopy to dwell by means of, and our groups are exhausted and confused.”
The interview, which closed out the ultimate day of TED 2025: Humanity Reimagined, showcased not simply OpenAI’s skyrocketing success but in addition the rising scrutiny the corporate faces as its know-how transforms society at a tempo that alarms even a few of its supporters.
‘Our GPUs are melting’: OpenAI struggles to scale amid unprecedented demand
Altman painted an image of an organization struggling to maintain up with its personal success, noting that OpenAI’s GPUs are “melting” as a result of reputation of its new picture technology options. “All day lengthy, I name folks and beg them to offer us their GPUs. We’re so extremely constrained,” he mentioned.
This exponential progress comes as OpenAI is reportedly contemplating launching its personal social community to compete with Elon Musk’s X, in line with CNBC. Altman neither confirmed nor denied these reviews through the TED interview.
The corporate just lately closed a $40 billion funding spherical, valuing it at $300 billion — the most important non-public tech funding in historical past — and this inflow of capital will seemingly assist tackle a few of these infrastructure challenges.
From non-profit to $300 billion large: Altman responds to ‘Ring of Energy’ accusations
All through the 47-minute dialog, Anderson repeatedly pressed Altman on OpenAI’s transformation from a non-profit analysis lab to a for-profit firm with a $300 billion valuation. Anderson voiced issues shared by critics, together with Elon Musk, who has steered Altman has been “corrupted by the Ring of Energy,” referencing “The Lord of the Rings.”
Altman defended OpenAI’s path: “Our objective is to make AGI and distribute it, make it secure for the broad good thing about humanity. I feel by all accounts, we now have carried out lots in that path. Clearly, our ways have shifted over time… We didn’t assume we must construct an organization round this. We discovered lots about the way it goes and the realities of what these techniques had been going to take from capital.”
When requested how he personally handles the large energy he now wields, Altman responded: “Shockingly, the identical as earlier than. I feel you may get used to something step-by-step… You’re the identical particular person. I’m certain I’m not in all kinds of the way, however I don’t really feel any totally different.”
‘Divvying up income’: OpenAI plans to pay artists whose kinds are utilized by AI
One of the vital concrete coverage bulletins from the interview was Altman’s acknowledgment that OpenAI is engaged on a system to compensate artists whose kinds are emulated by AI.
“I feel there are unimaginable new enterprise fashions that we and others are excited to discover,” Altman mentioned when pressed about obvious IP theft in AI-generated photographs. “In case you say, ‘I wish to generate artwork within the fashion of those seven folks, all of whom have consented to that,’ how do you divvy up how a lot cash goes to every one?”
At the moment, OpenAI’s picture generator refuses requests to imitate the fashion of residing artists with out consent, however will generate artwork within the fashion of actions, genres, or studios. Altman steered a revenue-sharing mannequin may very well be forthcoming, although particulars stay scarce.
Autonomous AI brokers: The ‘most consequential security problem’ OpenAI has confronted
The dialog grew notably tense when discussing “agentic AI” — autonomous techniques that may take actions on the web on a consumer’s behalf. OpenAI’s new “Operator” software permits AI to carry out duties like reserving eating places, elevating issues about security and accountability.
Anderson challenged Altman: “A single particular person might let that agent on the market, and the agent might resolve, ‘Effectively, with a view to execute on that operate, I obtained to repeat myself all over the place.’ Are there crimson strains that you’ve got clearly drawn internally, the place you recognize what the hazard moments are?”
Altman referenced OpenAI’s “preparedness framework” however offered few specifics about how the corporate would stop misuse of autonomous brokers.
“AI that you just give entry to your techniques, your info, the power to click on round in your pc… after they make a mistake, it’s a lot larger stakes,” Altman acknowledged. “You’ll not use our brokers if you don’t belief that they’re not going to empty your checking account or delete your knowledge.”
’14 definitions from 10 researchers’: Inside OpenAI’s battle to outline AGI
In a revealing second, Altman admitted that even inside OpenAI, there’s no consensus on what constitutes synthetic normal intelligence (AGI) — the corporate’s said objective.
“It’s just like the joke, for those who’ve obtained 10 OpenAI researchers in a room and requested to outline AGI, you’d get 14 definitions,” Altman mentioned.
He steered that fairly than specializing in a selected second when AGI arrives, we must always acknowledge that “the fashions are simply going to get smarter and extra succesful and smarter and extra succesful on this lengthy exponential… We’re going to must contend and get fantastic advantages from this unimaginable system.”
Loosening the guardrails: OpenAI’s new strategy to content material moderation
Altman additionally disclosed a big coverage change concerning content material moderation, revealing that OpenAI has loosened restrictions on its picture technology fashions.
“We’ve given the customers far more freedom on what we might historically take into consideration as speech harms,” he defined. “I feel a part of mannequin alignment is following what the consumer of a mannequin desires it to do throughout the very broad bounds of what society decides.”
This shift might sign a broader transfer towards giving customers extra management over AI outputs, doubtlessly aligning with Altman’s expressed desire for letting the lots of of thousands and thousands of customers — fairly than “small elite summits” — decide acceptable guardrails.
“One of many cool new issues about AI is our AI can speak to everyone on Earth, and we are able to study the collective worth desire of what everyone desires, fairly than have a bunch of people who find themselves blessed by society to sit down in a room and make these selections,” Altman mentioned.
‘My child won’t ever be smarter than AI’: Altman’s imaginative and prescient of an AI-powered future
The interview concluded with Altman reflecting on the world his new child son will inherit — one the place AI will exceed human intelligence.
“My child won’t ever be smarter than AI. They may by no means develop up in a world the place services should not extremely sensible, extremely succesful,” he mentioned. “It’ll be a world of unimaginable materials abundance… the place the speed of change is extremely quick and superb new issues are taking place.”
Anderson closed with a sobering statement: “Over the subsequent few years, you’re going to have a number of the largest alternatives, the most important ethical challenges, the most important selections to make of maybe any human in historical past.”
The billion-user balancing act: How OpenAI navigates energy, revenue, and goal
Altman’s TED look comes at a crucial juncture for OpenAI and the broader AI business. The corporate faces mounting authorized challenges, together with copyright lawsuits from authors and publishers, whereas concurrently pushing the boundaries of what AI can do.
Latest developments like ChatGPT’s viral picture technology function and video technology software Sora have demonstrated capabilities that appeared unattainable simply months in the past. On the identical time, these instruments have sparked debates about copyright, authenticity, and the way forward for inventive work.
Altman’s willingness to interact with tough questions on security, ethics, and the societal impression of AI reveals an consciousness of the stakes concerned. Nonetheless, critics could observe that concrete solutions on particular safeguards and insurance policies remained elusive all through the dialog.
The interview additionally revealed the competing tensions on the coronary heart of OpenAI’s mission: shifting quick to advance AI know-how whereas guaranteeing security; balancing revenue motives with societal profit; respecting inventive rights whereas democratizing inventive instruments; and navigating between elite experience and public desire.
As Anderson famous in his closing remark, the selections Altman and his friends make within the coming years could have unprecedented impacts on humanity’s future. Whether or not OpenAI can dwell as much as its said mission of guaranteeing “all of humanity advantages from synthetic normal intelligence” stays to be seen.
