What Will Get Better

What things will and won't get better in the next 50 years?

Will get better:

• Most kinds of technology.
  • Upsides of innovation outweigh downsides of introducing vulnerabilities for most technology.
  • E.g., phones, tablets continue to improve.
• Some kinds of forecasting where more data can help.
  • E.g., Weather forecasting (although it can never be perfect).
• Medicine: we'll get better at handling extreme conditions.
  • E.g. increasing life span of cancer patients, curing cancer more often.
  • The really hard part of medicine is making sure your intervention doesn't have serious side effects in the long term. But in extreme conditions almost any medical intervention can be justified, because the patient would die anyway. You care less about the side effects of the intervention.

Won't get better:

• Technological progress where risks are fat-tailed, i.e., risks are low probability but extremely costly.
  • E.g. nuclear tech (still on ancient computers), flight tech (still updated via floppy), bank transactions processing (still on mainframes).
  • The huge downside of introducing a bug outweighs upsides of innovating on such technology.
• Forecasting behaviour of complex adaptive systems.
  • E.g., stock market behaviour, wars between nation states, state of the economy.
  • The act of forecasting can change the behaviour of the system and thus invalidate the forecast.
• Politics. Tribalism, corruption, rent seeking continue.
• Medicine: not much progress on more modest conditions.
  • E.g. probably won't cure common cold.
  • Side effects matter a lot more here because patient would survive with or without medical intervention. It's hard to justify the value of the intervention without knowing long term side effects.
• Nutrition science. Progress will be hard beyond the basic facts already established (don't smoke, get enough vitamin C).
  • Effect sizes appear to be small for most nutritional interventions. E.g. Maybe turmeric does improve your long term health, but the effect is too small to be measureable unless the sample size is huge.