
How machine learning and robots will displace work faster than institutions adapt.
Kaplan, who founded GO Corporation and Onsale and teaches at Stanford, argues that machine learning and robotics will displace labour on a scale that existing institutions cannot absorb, and that the resulting problem is distributional rather than technological.
His framing has two parts. Synthetic intellects are systems that learn statistical patterns from data without being programmed with rules, and forged laborers are machines that act physically in the world. He argues these together attack cognitive and manual work simultaneously, which is different from previous automation waves that displaced one at a time. He examines the effect on employment, on wealth concentration since capital rather than labour captures the gains, on legal liability when a learning system causes harm and no one wrote the rule it followed, and on financial markets where algorithmic agents already interact faster than any human oversight.
His proposals are unusual and are the most interesting part. He suggests job mortgages, which finance retraining against future earnings; and a public benefit index that would give tax advantages to corporations whose shareholding is broadly distributed, on the reasoning that if capital captures the returns then ownership of capital should be widened rather than income redistributed after the fact. He is sceptical of universal basic income on the grounds that it separates people from productive participation.
The book is from 2015 and predates the transformer architecture and large language models entirely, so its picture of what the technology can do is dated in specifics while its distributional argument is not. Kaplan writes for a general audience and the economics is presented without much engagement with the labour economics literature on automation, which has found more complex effects than straightforward displacement.
Investors and policy-minded operators who want an early framing of automation as a distributional problem, read with the knowledge that the technology section predates modern language models.
Chosen by shared topics, the entries that hold them, and the people who recommended both.
AI as a fall in the cost of prediction, and what that does to decisions and jobs.
Why automation of cognitive work may cause lasting unemployment.
Recommended by: Daniel Petre, Patrick O'Shaughnessy
How digital technology is changing work and wealth as the Industrial Revolution did.
“For additional discussion of freestyle chess, see Erik Brynjolfsson and Andrew McAfee, The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies (New York: W.W. Norton & Company, 2014)” — Michael Mauboussin
Source: Lessons from Freestyle Chess: Merging Fundamental and Quantitative Analysis, page 9
Recommended by: Michael Mauboussin, Mohnish Pabrai and 5 others
How Google and others turn behaviour into data to predict and steer what people do.
Recommended by: Vinod Khosla, Benedict Evans
Why technological progress is exponential, and what follows if the trend holds.
Recommended by: Brian Armstrong, Patrick Collison and 7 others
The theory that the neocortex works through thousands of models of the world.
Recommended by: Bill Gates