Sam Altman tries to kill the GPT-4 hype

OpenAI CEO Sam Altman has a reality check for everyone waiting on GPT-4: lower your expectations, because you are setting yourself up to be let down.

During an interview with StrictlyVC, Altman addressed the swirling hype around ChatGPT's anticipated successor, delivering a blunt warning to the tech world: people are begging to be disappointed, and they will be.

Tearing down the 100-trillion rumor mill

Speculation about GPT-4 has taken Silicon Valley by storm, turning into an internet meme of wild capability predictions. Chief among them was a viral chart claiming the new system would boast a staggering 100 trillion parameters, dwarfing GPT-3's 175 billion.

Altman dismissed the chart in no uncertain terms, calling it complete bullshit. He added that the entire rumor mill around the model has gotten ridiculous, noting that the hype reflects an expectation that OpenAI has already achieved artificial general intelligence, which it simply has not.

As for the release date, rumors had pointed to a launch in the first quarter or the first half of the year. Altman declined to give any fixed timeframe, emphasizing that the model will arrive only when the team is fully confident they can roll it out safely and responsibly.

On video generation, Google, and custom models

Beyond GPT-4, Altman touched on several broader bets OpenAI is making across the AI landscape.

Generative video is on the roadmap. With Google and Meta having already shown off prototype demos, Altman confirmed that OpenAI is actively researching video generation. It is a legitimate research project that will come eventually, though he declined to predict a timeline.

On the business front, Altman admitted OpenAI is currently making not much money, characterizing the effort as very early.

He also pushed back on claims that ChatGPT will swiftly kill Google. History shows that calling the end of an incumbent giant usually proves wrong, Altman argued, reminding onlookers that Google is smart, competent, and fully capable of mounting a countermove. Search will change, but not as dramatically in the short term as the hype suggests.

When asked about AI bias and censorship, Altman proposed a flexible future. Instead of a single moral code, he envisions a world where users could write out a few pages detailing their own values and behavioral rules, allowing their personal AI to read and adapt to those boundaries directly.

As for cheating fears in schools, Altman noted that detection tools are only temporary patches that determined students will bypass. Society simply needs to adapt to synthetic text the same way it once adapted to handheld calculators.

Why it matters

For everyday users and builders, Altman's comments are an essential grounding wire. GPT-4 will likely be an impressive technical iteration, but anyone holding out for an omniscient digital mind right out of the box will end up disappointed. The focus needs to be on utility, safety, and reliability rather than sheer parameter vanity.

For the broader industry, Altman is executing a masterclass in expectation management. When hype balloons beyond reality, even a monumental product launch can feel like a failure if it falls short of science fiction. Calling out the viral rumors directly helps OpenAI pull the conversation back toward safety and incremental progress.

We also think his respect for Google's counteroffensive is telling. Startup exuberance often underestimates the deep distribution and research moats of incumbents. That said, having sparked this level of global excitement, Altman may find that dampening public anticipation is far harder than igniting it.