The 7 Core Categories of Generational Investment
1. Human Capital
Raising the productive capacity of the future
Early childhood development
Education
Public health, nutrition, mental health
Why:
Largest long-run ROI
Compounds over a lifetime
Hardest to repair if neglected
While the paradox currently holds, it is possible that AI may eventually become highly reliable, legally accepted, and autonomous for common scans, then staffing pressure could become real.
For now, radiology looks like job redesign and productivity expansion than destruction.
The radiologist paradox is the idea that AI can automate parts of radiology, yet radiologists may become more valuable rather than less valuable.
The original fear was that AI can read scans, so radiologists will disappear but this has not happened.
Medicine also has high trust and liability barriers. Patients don't totally defer to AI interpretations so a human physician is needed to remain the accountable decision-maker.
Fifth, AI could increase herd behavior. If the Fed, banks, hedge funds, and regulators all use similar models trained on similar data, they may all react in the same direction at once. That could amplify instability.
The Fed should use AI heavily as an analytical tool, but not let AI set monetary policy. AI should advise but humans should still decide.
AI can help the Fed in several useful ways:
1. Parse huge amounts of data faster
2. Detect turning points earlier than lagged data
Fourth, accountability matters. If the Fed raises rates and unemployment rises, the public deserves accountable decision-makers, not “the model said so.” An independent central bank already has a legitimacy problem; outsourcing judgment to opaque AI would make that worse.
@realBigBrainAI RL is not a general intelligence substrate. It often needs hybrid approaches (model-based RL, imitation learning, human-in-the-loop, curricula).
Ongoing research targets sample efficiency, better exploration, and multi-agent setups.
@realBigBrainAI RL is a powerful but narrow tool—highly effective for polishing known strategies in simulated environments with dense feedback, but limited for open-ended discovery or data-scarce real-world settings.
@pmarca Lastly, the best strategic sectors are not necessarily the most labor-intensive sectors. Grid equipment, shipbuilding, and repair are more labor-intensive while semiconductors, chemicals, and battery materials are more capital-intensive.
@pmarca Second, if the U.S. cannot train enough workers, the bottleneck becomes labor scarcity. You get wage pressure and project delays, not unlimited job creation.
Third, tariffs and subsidies can create jobs in protected sectors while destroying jobs elsewhere through higher costs.
@chamath It would be healthy to have our tax revenue sources dynamically change as the economy changes over time
Imbalances will still accumulate but it will be slowly and maybe more obvious