Anthropic CEO says AI could surpass human intelligence in a “small number of years”

Artificial intelligence may outrun human intelligence far sooner than many policymakers expect, according to Anthropic chief executive Dario Amodei. In recent public appearances he has argued that current systems are already on a steep trajectory and that models with greater cognitive abilities than people across most tasks could arrive within a small number of years.

His timeline is aggressive even within the tech industry, and it comes from someone who has built his career around both frontier AI research and safety work. The result is a prediction that blends optimism about medical breakthroughs and scientific discovery with stark warnings about misuse and civilisation-level risk.

From research leader to AI alarm bell

Dario Amodei, who leads Anthropic after senior roles at other AI labs, has become one of the most closely watched voices on advanced systems. His profile as CEO Dario Amodei matters because his company trains large-scale models that already sit near the top of benchmark charts and are deployed in products used by millions.

Case studies on Anthropic describe how CEO Dario Amodei has long argued that AI could drive unprecedented improvements in quality of life, from faster scientific discovery to new tools for education, while also creating serious risks that range from misinformation to existential threats. That dual focus shapes how he now talks about the next few years.

“Small number of years” to AI beyond most humans

At India’s AI Impact Summit, Anthropic CEO Dario Amodei told delegates in Feb that artificial intelligence is approaching a transformative moment. He said that systems are already “well advanced on that curve” and that there are only a small number of years before AI models surpass the cognitive capabilities of most humans for most things.

Coverage of the same remarks notes that the Anthropic CEO predicted artificial intelligence will surpass human cognitive abilities across most tasks within a “small number of years,” framing the shift as part of a broader trajectory where each generation of models gains more general problem-solving skills. Reports on the India Impact Summit highlight how he linked this to rapid progress in reasoning, coding and scientific assistance.

In a video summary of his talk, shared widely online, Amodei reiterated that AI systems are already matching or exceeding typical human performance on many benchmarks and that he expects that trend to continue steeply. The clip, circulated under the description that there are “only a small number of years” before AI models surpass human cognitive capabilities, has become a reference point for current AI timelines.

From AGI timelines to “AI geniuses”

Amodei has not kept his forecasts vague. In a detailed breakdown of his views, he has been quoted as saying that artificial general intelligence could be one to three years away, a claim summarized under the phrase Dario Amodei Says one to three Years Away, with His Full Breakdown focusing on scaling, data and algorithmic improvements.

On social media, a separate recap captured how ANTHROPIC CEO DARIO AMODEI SAYS AGI IS 1–3 YEARS AWAY, repeating his view that steadily larger models trained with better techniques are likely to reach what he considers general-purpose capability on par with top human experts.

Earlier commentary shared by AI specialist Syed Shahul Hameed on LinkedIn, under the heading Dario Amodei Predicts AI Elite Performance by 2026-2027, described Amodei’s expectation that AI could handle end-to-end software engineering and other elite tasks in that window, with the Video Player clip presenting him as confident that current trends will continue.

In a widely discussed interview summarized on Reddit, Dario said that many of his employees no longer write code themselves and instead rely on internal tools. When asked directly whether scaling alone would take the field to AGI, he said yes, and predicted that 2026 would see an explosion of AI capabilities which will accelerate everyone’s timelines. The same discussion thread relays his view that by 2027 AI will be able to handle complex reasoning tasks that currently trip up models, such as counting the letters in “strawberry.”

That belief in scaling as the main driver of progress underpins his timeline. He often points to something like a “Moore’s Law for intelligence” over the past decade, an argument that also appears in a Reuters video where the Anthropic CEO said that artificial intelligence will surpass human cognitive abilities across most tasks within a small number of years.

Public warnings on television and at global forums

Amodei has carried the same message into mainstream venues. In a clip shared on Bloomberg Television in Feb, Anthropic CEO Dario cautioned that AI models could surpass human cognitive capabilities within just a few years and stressed that society is not yet prepared for the consequences.

A separate video segment, promoted under the line Anthropic CEO: AI to surpass humans in “small number of years,” again presented the Anthropic CEO stating that Artificial intelligence will surpass human cognitive abilities in that compressed timeframe and urging governments to move faster on guardrails.

Earlier, during a panel at the World Economic Forum in Davos, Amodei described a “grand vision” in which AI powered advances in medicine could double human lifespans. He argued that if AI today can already help with drug discovery, then with five to ten more years of progress it might support treatments that radically extend healthy life, a claim summarized in coverage of how Amodei framed the opportunity.

In a separate feature, Anthropic CEO Dario Amodei said he thinks AI could help find cures for most cancers, prevent Alzheimer’s and even double the human lifespan, painting a picture of what healthcare might look like with advanced AI assistance. The segment described how he imagines AI systems working alongside doctors and researchers.

Civilisation-level risks and “AI geniuses”

Amodei’s optimism is matched by concern that humanity is not ready for what he calls “AI geniuses.” In a detailed essay summarized by SiliconANGLE, he warned that once society can be sure AI geniuses will not go rogue and overpower humanity, attention must shift to the misuse of AI by humans. He highlighted the risk that advanced models could significantly lower the barrier to developing biological weapons and other mass harm, a point captured in the line that Once those systems exist, their misuse could be catastrophic.

Another summary of his views on risk explains that Amodei organizes his concerns into five categories. Autonomy risks ask: Could AI systems develop goals misaligned with human values. Other categories cover misuse, economic disruption, loss of control and long-term existential threats, with the article noting that Amodei believes superhuman AI could arrive by 2027 with civilisation-level risks.

A policy-focused digest of his remarks similarly states that Anthropic chief executive Dario Amodei has issued a stark warning that superhuman AI could inflict civilisation-level damage unless governments and industry act far more quickly and seriously. That warning places his “small number of years” forecast into a broader argument for urgent regulation.

In a separate longform piece on AI’s potential, Amodei (Dario Amodei) presents an optimistic vision in which the technology helps solve disease, poverty and climate change. The same analysis stresses that he still sees deep downside risk if safety and governance lag behind capability.

Balancing hype, hope and hard timelines

For investors, regulators and the public, the question is how seriously to take Amodei’s compressed schedule. On one side, his argument that scaling existing techniques will be enough to reach AGI is supported by rapid gains in coding, language understanding and reasoning benchmarks, as well as internal anecdotes about Anthropic engineers who now rely on AI instead of writing code themselves.

On the other side, some researchers caution that current models still struggle with reliability, long-horizon planning and grounding in the physical world, and that these gaps might not close as quickly as raw compute curves suggest. Unverified based on available sources.

Like Fix It Homestead’s content? Be sure to follow us.

Here’s more from us:

• I made Joanna Gaines’s Friendsgiving casserole and here is what I would keep
• Pump Shotguns That Jam the Moment You Actually Need Them
• The First 5 Things Guests Notice About Your Living Room at Christmas
• What Caliber Works Best for Groundhogs, Armadillos, and Other Digging Pests?
• Rifles worth keeping by the back door on any rural property

*This article was developed with AI-powered tools and has been carefully reviewed by our editors.