"It's the Decade of Agents, Not the Year": The Realists and Skeptics of AI

In the last post we heard the people who say the takeoff has already started. Here are six voices who are just as knowledgeable and who take a more cautious line. Some build the technology, some run big companies and one won a Nobel Prize. Read them together and you get a more balanced picture.

The six voices

Andrej Karpathy (ex-OpenAI, ex-Tesla), Dwarkesh Podcast, 17 October 2025: "In my mind, this is more accurately described as the decade of agents." In the same conversation he argues that the problems are tractable but still difficult, and that AGI is still about a decade away. The episode is titled "We're summoning ghosts, not building animals".

Yann LeCun (founder, AMI Labs), MIT Technology Review, 22 January 2026: "The breakthroughs are not going to come from scaling up LLMs." MIT Technology Review described his new Paris-based company as a contrarian bet against large language models.

Satya Nadella (CEO, Microsoft), February 2025: Claiming an AGI milestone is "just nonsensical benchmark hacking". The Register reported he is more interested in AI boosting GDP than in delivering superhuman intelligence, and says AI is yet to find its killer app.

Geoffrey Hinton (Nobel laureate), September 2025, reported by Fortune: AI is "going to create massive unemployment and a huge rise in profits", and the moment may be "amazingly good" or "amazingly bad".

Mustafa Suleyman (CEO, Microsoft AI), 19 August 2025, "Seemingly Conscious AI": "We must build AI for people; not to be a person." His essay worries about what happens in the run-up to superintelligence, when systems can seem conscious.

Jensen Huang (CEO, Nvidia), Milken Conference, 6 May 2025: "You're not going to lose your job to an AI, but you're going to lose your job to someone who uses AI."

What I take from the realists

1. Agents are real but unfinished. They work in narrow settings and break in open ones, which matches the shutdown and database stories earlier in this series.
2. Scaling alone may not be the whole path. Serious people disagree, and the disagreement is the honest state of the field.
3. The most actionable line is Huang's. It is also a CEO selling chips, so I read it as a useful frame, not a law.

An exercise for you

Choose one realist and one accelerationist from this series, read each original source in full, and write a paragraph on where they actually disagree. It is more instructive than any summary. You will usually find that they agree on the facts more than the headlines suggest, and disagree on timing, on risk and on what to do about it.

The second video is a long interview with Geoffrey Hinton, in which he explains his concerns in his own words. As with every source in this series, watch it critically and compare it with others. To see all the timelines side by side, read Ask Ten Leaders, Get Ten Timelines.

Keep reading

• "The Takeoff Has Started": What the Accelerationists Actually Said, in Their Own Words
• Human-Level AI: When? Ask Ten Leaders, Get Ten Timelines
• From AI That Talks to AI That Works: The Real Shift From Chatbots to Agents

See the complete AI Era series

---