The ethical danger of anthropomorphising AI systems
Anthropomorphising AI—treating it as if it has human emotions, intentions, or consciousness—is both an ethical and practical problem, according to a recent opinion piece. The primary concern is deception: humanlike AI risks the lives of vulnerable people by obscuring the technology's true limitations. While AI can surpass humans in processing power, it also fails spectacularly because it remains a machine. Unlike the industrial revolution, when machines were not called teammates, today's language elevates AI to a false parity with human intelligence. Text-based generative AI, trained on vast text corpora, merely calculates statistical probabilities—e.g., given "thank", it predicts "you"; given "the cat sat on the", it predicts "mat". The article argues that this vocabulary choice misleads users about AI's nature.
Key facts
- Anthropomorphising AI is an ethical and practical problem.
- The foremost ethical concern is deception.
- Treating AI as humanlike risks the lives of vulnerable people.
- AI's processing powers can surpass humans but also fail spectacularly.
- Since the industrial revolution, machines were not referred to as teammates.
- Text-based generative AI is trained on vast amounts of texts.
- Given the word 'thank', AI statistically predicts 'you' as the next word.
- Given 'the cat sat on the', AI predicts 'mat'.
Entities
Institutions
- South China Morning Post