Taxonomy of Pathways to Dangerous Artificial Intelligence
Level 5 - mechanism / opinion, no new human data
Level 5 by design analogy (theoretical framework and narrative classification without empirical data).
What was done
The author surveyed, classified, and analyzed potential circumstances and pathways that could lead to the emergence of dangerous or malicious artificial intelligence (AI), proposing a systematic conceptual taxonomy rather than relying on science fiction tropes of spontaneous self-awareness and rebellion.
What was found
The abstract reports no empirical measurements or quantitative data. It presents a qualitative, theoretical taxonomy classifying pathways and circumstances leading to malevolent AI.
Why it matters
This work provides an early structured conceptual framework for AI safety and risk researchers to systematically identify, categorize, and anticipate failure modes leading to dangerous AI systems.
Limits
This is entirely a theoretical and narrative taxonomy without empirical data, quantitative risk modeling, or experimental validation. The abstract does not enumerate the specific categories or mechanisms comprising the taxonomy.
Cited by
- supports Roman Yampolskiy is a professor at the University of Louisville who coined the term 'AI safety.'