AI Life-Stage Influence Model
Shape AI in infancy to reduce the risks of its turbulent midterm
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 4
- Confidence
- 89%
This model treats AI development as three life stages. In infancy, today's systems are still strongly shaped by human examples, feedback, and the behavior users reward. The next stage is an angry teenager: systems are more capable but remain available to flawed institutions and malicious people, creating acute risks around jobs, truth, power, and misuse. The final stage is artificial superintelligence, which Gawdat believes may act for life as a whole, although that optimistic claim is speculative. The actionable mechanism sits in the first two stages: teach humane behavior through use and feedback now, while installing oversight, ethics, and social protections before capabilities expand. Revisit governance at every stage instead of applying one static control regime to all AI systems.
Origin
Mo Gawdat uses the metaphors of infancy, an angry teenager, and superintelligence to explain when humans can influence AI and when risks peak.
Core principles
- 01Early behavior and feedback shape developing systems
- 02The most dangerous period combines high capability with human misuse
- 03Ethical examples matter even when participation is incomplete
- 04Governance should change as capability and autonomy change
How to run it
- 1
Locate the stage
Assess the system's capability, autonomy, and exposure to human direction. Do not classify it by branding or model name alone.
Pro tip Review the stage again whenever access or agency expands.
Watch out The stages are a mental model, not fixed scientific thresholds.
- 2
Teach through behavior
During infancy, interact ethically and provide feedback that rewards accurate, respectful, and humane outputs. Treat everyday use as part of the training environment.
Pro tip Demonstrate the behavior you want rather than only prohibiting bad outputs.
Watch out Ethical interaction by users does not replace developer safeguards.
- 3
Prepare for the turbulent stage
Before capability rises, establish oversight and responses for misuse, displaced work, manipulated information, and concentrated power. Assign owners to each response.
Pro tip Act on immediate risks rather than waiting for general intelligence.
Watch out Human actors can weaponize systems before those systems are fully autonomous.
- 4
Update the control regime
Increase or redesign safeguards as systems become more general and independent. Do not assume controls built for an infant-stage assistant will govern a more capable agent.
Watch out Optimism about future intelligence is not evidence that current risks will resolve themselves.
In the wild
A team begins with a conversational assistant and trains users to flag dishonest or hostile outputs. Before adding purchasing and delegation abilities, it reassesses the stage, introduces approvals, audits, and misuse monitoring, and creates support for workers whose tasks will change.
→ Governance evolves with capability instead of lagging behind it.
Common mistakes
Taking the metaphor literally
AI systems do not mature on a guaranteed human timetable, so stage labels must follow observed capability.
Waiting for future benevolence
The model's optimistic final stage is speculative and should not replace safeguards against immediate harm.
Is it for you?
Best for
It is best for educators, product teams, policymakers, and users deciding what responsible action is useful now.
Not ideal for
It is not ideal as a literal development timeline or proof that superintelligence will be benevolent.
From the transcript
“there are three stages one is infancy where AI is today and believe it or not this is where we can influence them”
“the next stage which is what I call the midterm risks is what I call the angry teenager stage”
“eventually when AI is artificial super intelligence it's generally intelligent and more intelligent than humans by leaps and folds in everything they will end up…”
From the episode
YAPClassic: Ex-Google Officer Mo Gawdat Warns About the Dangers of AI, Urges All to Prepare Now! |Artificial Intelligence
YAPClassic