AI as a Transformative Technology
Since the launch of ChatGPT by OpenAI in November 2022, AI has been regarded as a disruptive technology capable of transforming lives akin to the Industrial Revolution. However, researchers at Princeton University suggest that the transformative impact of AI will unfold over decades, not years.
AI as a 'Normal' Technology
- Transformation Timeline: AI's impact is expected to be gradual, comparable to other technologies like the Internet.
- Adoption Reality: The rapid adoption of AI is often exaggerated, and its productive use requires traversing a learning curve.
- Non-Linear Influence: Past technologies show similar patterns in development and societal impact, suggesting AI's influence won't be non-linear.
AI Takeover Scenarios
- Assumptions: AI won't gain power without proving reliability, contradicting how organizations adopt technology.
- Control Measures: Businesses will ensure human control over AI systems, as seen with companies like Waymo and Cruise.
Innovation-Diffusion Feedback Loop
AI development previously relied on large Internet datasets but will now require real-world interaction and tacit knowledge. This feedback loop implies gradual human adoption of AI, leading to incremental improvements.
Impact on Human Labor
- Past Technology Adoption: General-purpose technologies like electricity and the Internet took decades to show economic impact.
- Job Shifts: Automation will reduce costs and shift human jobs to areas unaffected by automation, maintaining human oversight.
Focus on Deployment Phase Risks
- Adoption Focus: Policy interventions should facilitate AI adoption by workforce training and setting standards.
- Addressing Risks: Ensuring AI reliability and deploying different defences depending on environments is critical.
Interventions and Resilience
- Temporary AI Development Pause: Calls for a pause in AI development to mitigate societal risks may not be effective.
- Resilience Approach: Building resilience by preventing the concentration of power and resources can help manage risks progressively, akin to developing an "immune system" against AI-related challenges.