Three Inevitables of AI
Plan around AI's arrival, superior intelligence, and unavoidable failures
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 4
- Confidence
- 91%
The Three Inevitables model begins with three planning assumptions: AI has already happened and cannot simply be switched off; its intelligence will surpass human intelligence in more domains; and harmful outcomes will occur somewhere along the path. Competitive incentives reinforce the first assumption because companies and nations fear stopping while rivals continue. The framework turns those claims into a strategy exercise rather than a prediction contest. Accept the direction of travel, test plans against much stronger systems, and identify failures that deserve action now. Gawdat emphasizes immediate risks such as job disruption, concentrated power, manipulated truth, and misuse over less probable existential stories. The output is a prioritized adaptation plan that remains useful even if dates and capability forecasts prove wrong.
Origin
Mo Gawdat presents the three inevitables from his book Scary Smart as a way to frame the AI transition.
Core principles
- 01Plan from conditions that already exist rather than wishing them away
- 02Assume capability will continue to improve under competitive pressure
- 03Expect failures and prepare responses before they occur
- 04Prioritize probable immediate harms over sensational distant scenarios
How to run it
- 1
Accept the starting condition
Treat AI as an active and expanding force rather than a future possibility. Remove plans that depend on every competitor voluntarily stopping.
Pro tip Separate what would be desirable from what your plan can safely assume.
Watch out Acceptance is a planning premise, not approval of every use.
- 2
Stress-test for greater intelligence
Ask how the plan changes when machines perform the relevant cognitive task better, faster, or cheaper than people. Identify assumptions tied to permanent human superiority.
Pro tip Test the task, not the whole job title.
Watch out Exact intelligence forecasts are uncertain; do not disguise them as facts.
- 3
Inventory likely failures
List immediate and longer-term harms without collapsing them into one dramatic scenario. Include misuse, labor disruption, misinformation, and power concentration where relevant.
Pro tip Start with harms that are already observable.
Watch out Focusing only on extinction can distract from urgent damage.
- 4
Prioritize action
Rank risks by probability, impact, reversibility, and time horizon. Fund measures that reduce near-term harm while preserving options for larger future threats.
Watch out Do not let uncertain dates become an excuse for inaction.
In the wild
A firm assumes AI adoption will continue, tests which research and drafting tasks stronger systems can absorb, and inventories immediate risks to junior roles, client confidentiality, and quality control. It then prioritizes retraining, human review, and a transition plan instead of betting on regulation halting the tools.
→ The firm prepares for disruption without relying on one precise AGI forecast.
Common mistakes
Treating inevitability as passivity
The assumptions are meant to focus preparation, not to excuse unsafe development or remove human responsibility.
Forecasting one dramatic outcome
A single distant scenario can obscure more probable harms already affecting work, truth, and power.
Is it for you?
Best for
It is best for leaders creating resilient AI adoption, workforce, or governance plans.
Not ideal for
It is not ideal for making precise forecasts about when AGI or specific harms will arrive.
From the transcript
“there are three inevitables AI has already happened not just will happen”
“it is the first inevitable it's an arms race that will not stop”
“the other inevitable is it will eventually become smarter than us”
From the episode
YAPClassic: Ex-Google Officer Mo Gawdat Warns About the Dangers of AI, Urges All to Prepare Now! |Artificial Intelligence
YAPClassic