The problem this solves
AI risk arrives in two shapes neither of which helps an adult make a decision: hype (everything is about to change, sign up before it's too late) and dismissal (it's just autocomplete, nothing to see here). Both are emotionally cheaper than the third position — taking the technology seriously, acknowledging genuine uncertainty, and making calm asymmetric bets.
This micro-course teaches you to read AI claims like an investor reads a pitch. Five categories of risk, a seven-question evaluation framework, time-horizon discipline, and cost-asymmetric responses. The point isn't to know the future. It's to make decisions that survive plausible versions of it.
A taste of the exercise
The preview lesson walks you through applying the seven-question evaluation framework to one confident AI claim from your last week of news consumption.
Key concepts
- Substitution risk
- Tasks (and parts of jobs) being done by AI more cheaply / faster. Already underway in writing, coding, design, paralegal, translation.
- Misuse risk
- Humans using AI for fraud, deepfakes, surveillance, disinformation, blackmail. Not new categories; AI cuts the cost.
- Systemic risk
- AI integrated into critical systems creating cascading failure modes or amplifying fragility.
- Concentration risk
- A small number of companies owning the substrate the economy runs on. Geopolitical, regulatory, commercial implications follow.
- Alignment risk
- The speculative category. Highly capable systems acting on misspecified objectives. Real research; contested timescales.
- Asymmetric bet
- An action whose cost is small if you're wrong and large benefit if you're right. Most good AI personal bets are asymmetric.
Common mistakes
- Treating ‘AI’ as one thing rather than a family of capabilities.
- Confusing a hype video with deployment reality.
- Forecasting from feelings (excitement, anxiety) rather than capability deltas.
- Reading only one tribe (optimists or doomers).
- Skipping the timescale question on every prediction.