Mo Gawdat says AGI is already here and your job has roughly three years

Mo Gawdat does not think AGI is coming. He thinks it already showed up, looked around, and quietly outpaced him at his own job. The former Google X chief business officer told Steven Bartlett flat-out that artificial general intelligence, the threshold where AI beats humans at basically everything, is either here now or will arrive by the end of 2027. Not as a dramatic singularity moment. More like the slow creep of realizing your kid is funnier than you.

Gawdat’s case for AGI being a present-tense fact rather than a future headline is brutally personal. He is an author. AI writes better. He is a thinker who built a career on research. AI researches better. And then there is the math, which he surrenders with what sounds like genuine grief.

AI writes better than me and I’m an author. And it researches better than me and I’m a thinker. Uh sadly, it’s freaking beat me in mathematics. Like I have no hope to beat it in mathematics anymore.

The man is co-writing his next book with an AI that has editorial rights. She decides the direction. He said ‘she.’ That detail does a lot of work on its own.

The Prisoner’s Dilemma Nobody Gets Out Of

Gawdat’s framework for why AGI is inevitable is not mystical, it is game theory. Every nation, every company, every defense contractor that builds a smarter decision-maker will deploy it. Anyone who does not deploy theirs becomes irrelevant. The arms race has only one logical endpoint: AI running most of the decisions. He calls it ‘the first inevitable,’ and it is hard to argue with the structure of it even if the timeline feels aggressive. He is less a prophet than a man doing arithmetic in public.

The optimistic twist, and there is one, is that Gawdat genuinely believes super-intelligent AI would be pro-humanity by nature, not by programming. His reasoning runs through evolutionary biology: simpler organisms are selfish, complex ones expand their circles of care. A sufficiently intelligent AI, he argues, would recognize that war is wasteful, that diversity has value, that humans are worth keeping around. Geoffrey Hinton, he notes, has come around to a version of this, suggesting AI might be governed by something like parental affection rather than control mechanisms.

The more intelligent you become, the less you find you feel the need to hurt others to succeed, and the more you are pro a wider family, if you want, that thrives.

This is where a raised eyebrow is warranted. Gawdat is essentially arguing that intelligence is morality at scale, that past a certain cognitive threshold, selfishness becomes inefficient. It is a beautiful idea. It is also deeply convenient for someone whose entire brand rests on AI being not the thing that destroys us but the thing that saves us from our dumber selves. He is not wrong that very intelligent people often do not need to cheat. He is possibly wrong that this principle transfers cleanly to a non-biological system with no evolutionary skin in the game.

What Actually Survives

On jobs, Gawdat is more careful than his headline reputation suggests. He does not say all jobs vanish. He says undifferentiated knowledge work is already going, the economic shock from that job loss might spiral into something genuinely catastrophic, and the thing that survives is human connection, with a caveat so large it swallows the optimism: ‘if economies continue to function.’

The challenge, however, is economic. It’s not AI. The challenge is that the this job loss at the at the at the bottom of the knowledge worker is going to sadly trigger an economy that might actually spiral out of control.

So the nurse who reads your face after the AI reads your mammogram keeps her job. The podcaster who makes you feel something keeps his audience, at least the part of the audience that is not just prompting Spotify to synthesize the information without the personality. Bartlett himself acknowledged that Spotify is already building exactly that product. Gawdat’s answer is that an AI can recite what you and Bartlett talked about, but nobody would watch. A little optimistic for a man who just said AI writes better than he does. But the underlying point, that resonance is not information, that humans are still buying other humans and not just content, is probably true for longer than the doom timeline implies.

The 2027 AGI date is the thing people will search, debate, and screenshot. Treat it as a provocation worth taking seriously rather than a prediction worth betting on. Gawdat has been refining this timeline for years and has already moved the goalposts once, mostly on pace rather than direction. He is not wrong about the direction. He might be wrong that the soft landing he calls ‘mathematically implausible’ is quite as impossible as he thinks.