AI Investment & Labor Market Impact

Media exuberance about AI is not new; similar headlines appeared in 1958.

A 1958 New York Times article claimed the Navy-funded Perceptron would walk, talk, see, write, reproduce itself, and be conscious within a year for $100,000.

Why: Catchy terms like 'artificial intelligence' drive press excitement and investor imagination more than technical descriptors like 'symbolic processing.'

"The people at the New York Times were possibly a bit optimistic."

Technology replaces skills, not entire jobs, reshaping the nature of work.

US farm employment fell from 80% in 1870 to <2% today, yet new industries absorbed labor gradually as productivity gains created new wealth and roles.

Why: Increased productivity lowers costs and spurs demand, which in turn generates new types of employment that did not previously exist.

"Technology has increased worker productivity since it first appeared in the industrial revolution, and this increase in productivity meant that fewer people were needed to do the same amount of work, but the increased wealth that resulted from this new higher productivity ended up creating new jobs."

The speed of skill obsolescence, not automation itself, is the primary labor concern.

Oxford researchers estimated 47% of US jobs at high automation risk and 19% at medium risk; white-collar information processing roles are especially exposed.

Why: AI now targets less-routine cognitive tasks—writing, translation, graphic design—whereas past automation focused on repetitive physical tasks.

"The risk of AI bringing about a sudden shift as to which skills are valued in the labour market is what people worry the most about as the rate of technological progress speeds up."

AI simulates but does not possess intentionality or understanding.

John Searle argues computers manipulate symbols without meaning; generative AI can produce art yet includes distorted Getty Images logos because it lacks semantic comprehension.

Why: Without consciousness or intent, AI outputs are unmoored structures interpreted by humans, not genuine thought.

"Without understanding or intentionality, we can't describe what the machine is doing when it's running a program as thinking."

Broad diversification already provides AI exposure; chasing hot startups risks picking MySpace over Facebook.

59% of large companies claim an AI strategy and 70% say they understand how AI creates value, so most diversified portfolios are implicitly exposed.

Why: Fads like weed stocks, 3D printing, and NFTs show that trying to catch every wave often underperforms steady, broad-based investing.

"Odds are that you really don't need to change much about your investment portfolio to get exposure to artificial intelligence."

Human adaptability and preference for social interaction buffer pure automation.

People still attend live concerts, gyms with trainers, and cafés with baristas despite cheaper automated alternatives, valuing tailored advice and community.

Why: Many tasks remain unautomated because part of their value is relational, not transactional.

"A mobile phone is perfectly capable of playing you a recording, of a great piece of music, but people still pay to go to concerts, to see the piece performed live and to be part of the event."

AI boosts individual productivity but can temporarily reduce aggregate employment.

When some skills are automated, fewer total workers may be needed even if the job still exists; cashiers now handle exceptions rather than every item.

Why: Employers need tasks completed, not workers per se; partial automation raises output per worker, shrinking headcount until new demand emerges.

"Employers just need tasks completed rather than needing workers, and the new technology that makes one worker more productive is likely to, at least in the short term, put another worker out of a job."

Generative models recombine existing human-created data without true creativity.

Stable Diffusion was trained on scraped internet images, occasionally regurgitating distorted Getty logos—evidence it lacks semantic understanding of what it creates.

Why: Statistical pattern-matching across large datasets can mimic style but misses nuance and intent inherent in human artistry.

"Stable Diffusion relies on human created images for training data... it was noticed that Stable Diffusion sometimes includes a distorted Getty Images logo in the corner of the images that it creates."