Karen Hao: Rebel against the Empire

In this episode, we sit down with Karen Hao—award-winning journalist, MIT-trained engineer, TIME 100 AI honoree, and author of the best-seller Empire of AI—to unpack the power structures shaping today’s AI, business, and competition.

She has a clear and punchy point of view. In our conversation we ask:

- Is the Silicon Valley–Wall Street model the only way forward, or others can chart a better path?

- Do paradigm shifts like AI inevitably create harmful externalities—hidden data workers, energy and water shortages—just as the Industrial Revolution created pollution and child labor?

- Are today’s billion-dollar AI investments solving big problems—or just fuelling hype?

- Are we consciously shaping AI—or sleepwalking in “a dream made in the Valley”?

Karen argues that today’s AI giants act as modern empires: claiming resources that aren’t theirs, exploiting vulnerable labour, monopolizing knowledge, and sustaining themselves with a narrative of existential competition. She shows how many tools we use were born from subjective and often ad-hoc decisions—like scraping even the murkiest corners of the internet—despite harmful downstream effects.

She is critical of current regulatory efforts, weakened by Big Tech lobbying and points to real alternatives. She warns of an inevitable financial bust—investments wildly outpacing business value. But her message is ultimately hopeful: beyond the Big Labs, smaller teams are building AI for healthcare and the environment, showing that another path is possible. 

Her call to action is clear: we must not sleepwalk into a future shaped by a self-serviing AI elite. We need to collectively imagine alternatives, and build a more balanced future. As Karen claims, something better is achievable: "The shape of a technology is never inevitable: there are many, many human choices."

Note: This podcasst was recorded as part of an INSEAD alumni webinar. Live questions were submitted by the audience. 

Empire of AI book link.

See omnystudio.com/listener for privacy information.