AI Leaders Keep Moving the Goalposts on AGI

Ever notice how the people getting rich off AI keep telling us it might end the world — right before asking for more data centers? Yeah, us too.

That's the vibe behind a new running tally of the most ridiculous things AI bigwigs have said. The premise is simple: the leaders of the AI era love watching their valuations climb almost as much as they love warning us their software could destroy everything. AI is going to kill us all — now can we get more money for data centers, please?

Their other favorite hobby? Inventing trendy phrases that either explain how smart the next model will be or justify why last year's promised sentience never showed up. These buzzwords are hard to track and even harder to parse, so the log exists to remind us how often AI thought leaders contradict each other and move the goalposts.

The AGI predictions alone are a trip. Back in October 2023, Google's Blaise Agüera y Arcas and Stanford's Peter Norvig declared AGI was already here. By March 2024, Nvidia's Jensen Huang said we'd get there within five years, while Elon Musk figured it was one to two years away. A month later, Musk doubled down on 2025. OpenAI's Sam Altman agreed, saying AGI was coming in 2025.

Then the timelines started slipping. In January 2025, Google's Demis Hassabis guessed three to five years. By December 2025, Musk was saying xAI could hit AGI in the next few years, maybe 2026. August 2026 brought Altman saying OpenAI was "not quite yet" there, but that an internal system he'd call AGI would arrive by year's end. And in September 2026, Huang posted that GPT-6 Astra meant AGI had officially arrived.

When AGI stopped being convenient, the industry found a new promised land: recursive self-improvement. Anthropic's post "When AI builds itself" pushed the term into overdrive, describing a future where AI autonomously designs its own successor — while noting we're not there yet and it isn't inevitable. There's also "pace the frontier," a phrase that surfaced after 1,386 frontier AI employees signed a statement asking the U.S. government to help deliberately slow automated AI development.

So what's the takeaway for you? Mostly that the people selling AI keep rewriting the finish line, and the buzzwords change faster than the models do. Worth keeping an eye on — and maybe a grain of salt handy. What do you think: are these predictions bold vision or just good marketing?