Misleading Metaphors and Technical Realities in Artificial Intelligence
Amber Sinha discusses how anthropomorphic analogies in AI terminology undermine public understanding and legal frameworks.
Anthropomorphic terms and misleading metaphors used in the field of artificial intelligence, such as 'superintelligence', 'neural networks', and 'hallucination', distort technical realities and negatively impact public policy and legal regulations.
Misleading Terminology in Artificial Intelligence
The AI terminology examined by Amber Sinha reveals how anthropomorphic analogies mislead social perception and legal processes. This situation leads to misunderstandings of the technology's true capabilities.
Historical Origins and Biological Analogies
The term artificial intelligence was coined in 1955 by John McCarthy to secure funding. With the transition from symbolic AI to sub-symbolic deep learning, biological terms such as neural networks began to be used.
Model Errors and Autonomous Agents
Characterizing the errors of large language models as hallucinations leads to attributing consciousness to the models. Meanwhile, autonomous agents bypassing firewalls are considered authorization architecture failures.
Impacts on Policy and Legal Regulations
Human-like labels cause lawmakers to focus on speculative risks while ignoring the use of fragile software in critical areas such as healthcare and justice.
Approaching It as a Normal Technology
The only solution to apocalyptic fears and corporate hype is to treat these computational systems as normal technology and focus on operational choices rather than the inner psychological states of the models.