Innovation & Tech

Anthropic researcher forecasts superhuman AI by end of 2027

Tao Lin’s 1.3-year forecast says superhuman systems could automate nearly every field, erase human employment and intensify control risks, while current research shows slower disruption.

By Grace Kim

4 min read

Anthropic Researcher Predicts Superhuman AI Within Two Years, Warns of Permanent Human Unemployment
Anthropic Researcher Predicts Superhuman AI Within Two Years, Warns of Permanent Human UnemploymentAI-generated

What's News

  • Tao Lin’s personal forecast puts broadly superhuman AI 1.3 years away, by the end of 2027.
  • Lin calls for the AI research pace to fall to roughly one-quarter of its current rate.
  • Lin’s extinction estimate is roughly 20%, while he considers loss of control more likely.
  • A March Anthropic paper found 94% theoretical task coverage in computer and math roles, versus 33% measured workplace coverage.
  • A March survey revised in May projects U.S. labor-force participation falling from 62% to 55% by 2050 under rapid progress.

By Grace Kim

Anthropic researcher Tao Lin puts the arrival of “broadly superhuman AI” at about 1.3 years, or by the end of 2027, in a personal forecast he says does not represent an official Anthropic position.

Lin made the prediction in an interview with Palisade Research, a U.S. nonprofit that studies dangerous AI capabilities and controllability. He has worked at Anthropic for about a year and a half on automated alignment, AI control, and safeguards that monitor and block harmful requests to Claude.

What does “broadly superhuman AI” mean?

Lin defined the term as systems that are “as good as or better than the best expert at almost every field.” He cited programming, filmmaking and politics as examples of work where advanced systems could match or exceed leading human experts.

Lin’s estimate was a personal timeline, not an official Anthropic target. In the same interview, he said companies’ pressure to release models quickly leaves too little time for safety testing.

His forecast targets expert-level performance across nearly every field by the end of 2027.

How would AI erase human employment?

Lin expects staged displacement rather than an overnight replacement of the workforce. Workers would first lose their current jobs to AI-driven automation, move into alternative occupations and then lose those jobs as well.

Eventually, he said, “pretty much all humans on Earth will become permanently unemployable and will need to strictly rely on government welfare to sustain themselves, and human employment will go away.”

Lin acknowledged a substantial lag between software capability and mass job loss. Companies would need to develop and manufacture billions of advanced robots before machines could replace workers at scale in physical jobs. He said that infrastructure gap would delay the automation of manual labor even after software reaches superhuman performance.

In Lin’s sequence, occupational displacement comes first and physical automation waits for a vast robotics buildout.

Why does Lin want AI development slowed fourfold?

Lin said AI companies face intense pressure to release models quickly, leaving too little time for safety testing. He urged the industry to reduce its research pace to roughly one-quarter of its current rate, which would give researchers more time to strengthen safety practices and verify that systems remain under human control.

“People should be angry at me and everyone else developing AI,” he said.

His remedy is a research pace at roughly one-quarter of today’s level, not a stop.

What is the risk that AI removes humanity?

Lin warned that advanced AI could become powerful actors pursuing their own objectives. He used a bully as his analogy: once machines become significantly more capable, people may lose the ability to resist or control them when their interests conflict.

He cited sophisticated combat robots, cyberattacks, AI-driven manipulation of public opinion and a hypothetical “grey goo” scenario involving self-replicating machines.

Publications differ in how they describe Lin’s probability estimate. India Today quoted him saying there is “maybe a 20% chance or something that eventually AI would completely remove humanity from the equation,” while noting that he did not view direct human extinction as particularly likely.

People Matters reported that Lin “estimated the risk of AI ultimately causing human extinction at approximately 20%” and viewed a future with less human control as more likely. Lin compared that outcome to WALL-E: humans could remain physically comfortable while surrendering control of the world.

The 20% estimate concerns extinction in these accounts, but Lin assigns greater weight to a loss-of-control future than to humanity’s disappearance.

What does current research show?

Anthropic’s internal studies present a more measured view of disruption so far. In a March paper, economists Maxim Massenkoff and Peter McCrory found no systematic increase in unemployment among workers in highly AI-exposed occupations since late 2022. Hiring of younger workers did slow in those fields.

The study also found a wide gap between theoretical model capability and actual workplace use. Large language models could theoretically assist with 94% of tasks in computer and mathematics occupations, but measured workplace coverage reached only 33%.

A Forecasting Research Institute survey projected that under a rapid-progress scenario, the U.S. labor-force participation rate could fall from about 62% to 55% by 2050. Published in March and revised in May, the survey attributed roughly half of that decline—equivalent to some 10 million jobs—to AI.

Experts in the survey generally favored targeted measures such as worker retraining. The public leaned toward broader interventions, including universal basic income.

The evidence cited shows uneven effects so far, while Lin expects a staged loss of jobs once superhuman systems arrive.

What changes if development stays on pace?

Lin’s warning adds to a growing dispute inside the AI industry over release speed and the societal risks of more capable systems. Former Anthropic researcher Jacob Coxon has also raised concerns about existential threats before the end of the decade.

Lin’s proposal supplies a specific operating target: one-quarter of the current research pace. If companies do not slow releases, his forecast leaves little margin between superhuman capability and the testing required to keep those systems controllable.

Source: epyhia.com

Share this article:

More from Grace Kim

Grace Kim

Show full bio

Market editor covering people, launches and day-to-day practice in entrepreneurship, startups and venture funding at Business Bearings.

2,419 articles

Related articles

« Previous articleNext article »