Quote · The Prof G Pod with Scott Galloway
The Crisis of Adulthood — with John Burn-Murdoch
Where this was said
AI Productivity: The Signal-to-Noise Problem
At 49:20 · chapter starts 47:42
Scott asks how a data journalist cuts through the wildly conflicting AI productivity headlines, given the financial incentives of the parties involved. John Burn-Murdoch offers a methodological answer: look for data that tracks the full value chain, not just proxies. The first wave of AI productivity studies measured software engineer output — lines of code written, pieces of software shipped — and found meaningful gains. But more recent work follows that pipeline all the way to product adoption, and the picture changes: very little additional real value has been captured at the end. Burn-Murdoch applies this framework to his own experience: using Claude and Codex, he generates far more intermediate material, but at week's end he still produced two articles — the same as before. The honest conclusion is that AI is very effective at accelerating low-level activity, but whether that translates into proportional high-value output for clients, readers, or customers remains genuinely elusive and will likely only be answerable in a few years' time.
Research suggests every major AI model nudges users toward more moderate, expert-aligned political views, reversing the polarising effect of social media, because their training data and business model reward accuracy over sensationalism.
Social media amplified fringe voices because anyone could post and attention was the currency. AI flips that model: trained on mainstream sources and rewarded for accuracy rather than clicks, every major model consistently nudges users toward moderate, expert-aligned views. It might be the first new media form since television to pull people toward the centre.
