AI Leaders Clash Over When Human-Level Intelligence Will Arrive

The Battle Lines Are Drawn

At this year's World Economic Forum in Davos, Switzerland, the world's most influential AI researchers engaged in a very public disagreement about one of the industry's most pressing questions: how close are we to achieving artificial general intelligence?

On one side, Anthropic CEO Dario Amodei delivered bold predictions that sent shockwaves through the tech community. He stated that AI would replace most, maybe all, of what software engineers do within six to twelve months. He went further, predicting AI would reach Nobel-level scientific research within two years and that fifty percent of white-collar jobs could disappear within five years. Amodei maintained his forecast that human-level AGI could emerge by twenty twenty-six or twenty twenty-seven.

The Sceptics Push Back

Demis Hassabis, the Nobel Prize-winning CEO of Google DeepMind, offered a starkly different view. He told audiences that current AI systems are nowhere near human-level AGI, estimating only a fifty percent chance of achieving such capabilities by twenty thirty. Hassabis identified critical gaps in current systems including the ability to learn from just a few examples, continuous learning, better long-term memory, and improved reasoning and planning.

Yann LeCun, the Turing Award recipient who recently left Meta to found AMI Labs, was even more direct in his criticism. He argued that large language models will never achieve human-like intelligence because they cannot build world models that connect cause and effect. We're never going to get to human-level intelligence by training LLMs or by training on text only, LeCun stated. We need the real world.

What's Really Happening Inside AI Companies

Amodei's predictions weren't just theoretical. He revealed that inside Anthropic, engineers have already shifted from writing code to reviewing and refining AI-generated output. One engineer who created Anthropic's coding tool said that one hundred percent of his contributions over the past month were written by AI itself.

The Economic Reality Check

Despite their disagreements on timelines, the leaders found common ground on economic impact. For business leaders attending the forum, the immediate challenge isn't about AGI timelines but implementation. Cognizant CEO Ravi Kumar told the forum that current AI technology could unlock approximately four point five trillion dollars in United States labour productivity, but only if companies restructure their businesses and reskill their workforces. Skilling is no longer a side thing, Kumar argued. It has to be part of the infrastructure story.

The debate reveals a fundamental split in the AI community about both the path to AGI and when it might arrive, with profound implications for workers, businesses, and society at large.