Concrete AGI risk demos fall short as advocacy tool

The challenges of using concrete AI demonstrations to convince skeptics about the risks of Artificial General Intelligence (AGI) reveal a fundamental disconnect in how different experts interpret algorithmic behavior and its implications for future AI development.

The central challenge: Demonstrations of potentially concerning AI behavior often fail to persuade AGI optimists because they can explain away the behavior as predictable algorithmic outcomes.

• When presented with demonstrations of problematic AI behavior, technically knowledgeable optimists often dismiss concerns by pointing out that the algorithm simply performed as its code would suggest
• The predictability of algorithmic behavior, even if only obvious in hindsight, can make demonstrations feel less threatening to skeptics
• The lack of catastrophic real-world consequences in these demonstrations further reinforces optimists’ existing beliefs

Key counterarguments: The AGI risk community faces several distinct types of skepticism that require different approaches to address.

• Some skeptics maintain unrealistic optimism about algorithmic behavior
• Others place excessive faith in human ability to control AI systems
• Many have overconfident views about institutional safeguards
• A significant portion simply haven’t thoroughly analyzed the interconnected risks

Strategic implications: Effective advocacy for AGI risk awareness needs to target specific misconceptions held by different audiences.

• Demonstrations may still prove valuable for those who haven’t deeply considered algorithmic behavior
• Advocates should identify and directly quote claims that their demonstrations effectively counter
• Different approaches may be needed for technically sophisticated skeptics who focus on algorithmic predictability

Technical perspective: The focus on algorithmic predictability misses broader implications about AI risk.

• The mechanical nature of algorithms applies equally to human brains and any potential artificial intelligence
• Predictability at a code level doesn’t diminish the potential for emergent behaviors and risks
• Understanding how code works doesn’t necessarily translate to understanding its broader implications

Looking ahead: The challenge of convincing AGI optimists through demonstrations highlights the need for a more nuanced approach to AI safety advocacy that addresses specific forms of skepticism while acknowledging the limitations of any single persuasion strategy. Future advocacy efforts may need to combine technical demonstrations with broader evidence about institutional and human factors that contribute to AI risk.