AI: It AI-n't What You Think!
summary
Argues that modern AI is accelerated inference not true intelligence, trained on human code so it often invents facts and produces arbitrarily inconsistent outputs (example: airline chatbot promising reimbursement that led to court rulings). Shows AI excels as a code investigator — it finds subtle bugs (order‑of‑evaluation C++ example) — but generated solutions vary wildly. Recommends use AI for ideas not solutions, keep human verification, and plan for data+specs to one day regenerate entire apps.
AI-written summary · about
chapters (14)
- 0:00Welcome
- 1:58AI revolutionary or evolutionary
- 5:07Clarke's laws and boundaries
- 8:21Historical tech parallels
- 13:23Generational shifts
- 19:01AI as alternative info
- 25:28Debugging vs code generation
- 31:55AI tool and limits
- 37:38AI for ideas not solutions
- 41:16AI as black swan
- 44:24AI compilers future
- 47:43Business agility focus
- 50:29Adapting to AI change
- 53:20Call to action
