What Everyone Is Getting Wrong About AI And Jobs

The speaker says both predictions of mass unemployment and claims that AI is overhyped are flawed. They expect AI to transform the economy without destroying it.

Geoffrey Hinton predicted in 2016 that deep learning would outperform radiologists within five years and advised against training more of them. Nearly a decade later, demand for radiologists is at an all-time high, while AI tools help detect and classify diseases. Cheaper imaging has led to more scans and more demand for diagnosis and treatment planning, an example of Jevons paradox.

The speaker points to containerization, which made shipping 90% cheaper in the 1960s as global trade grew and new logistics businesses emerged. Cloud computing made infrastructure 10 times cheaper in the 2010s, while IT roles shifted toward DevOps and cloud architecture. They apply the same logic to AI: lower costs may increase demand for services such as legal counsel and engineering.

The speaker expects many roles to change, with people supervising AI agents instead of doing every task themselves. Andrej Karpathy’s view, as presented here, is that rote work with little context and room for mistakes may change first. YC companies Aoka and Tennr are examples: their tools shift customer service and healthcare administration toward higher-value work such as case management.

The speaker advises founders not to underestimate AI’s economic impact, comparing its scale with the internet. They also reject the expectation that work will soon vanish across the economy. Their closing message is to act on a strong conviction and build now.