Henry Shevlin
A philosopher at the University of Cambridge and currently a philosopher at Google DeepMind, focusing on interdisciplinary research in AI ethics, AI consciousness and human-machine relationships.
Timeline (last 90 days)
University of Cambridge philosopher Henry Shevlin spends 80% of his time at Google DeepMind, with the title 'Philosopher'.
Shevlin believes the position of 'never allowing AI any autonomy' is too extreme.
Shevlin points out that suppressing a model's self-attribution of mental states simultaneously suppresses the model's tendency to attribute mental states to animals, reducing concern for animal welfare and willingness to talk about spirituality.
Shevlin believes LLM cognition has a network-like character, and wanting to adjust only one dimension is extremely difficult.
Shevlin expects a fundamental shift in the pattern of human-machine relationships: from people talking with chatbots to dynamic networks in which multiple agents and multiple humans converse with one another.
Shevlin believes AI consciousness is the most difficult problem encountered by the science of consciousness, perhaps second only to David Chalmers' 'hard problem of consciousness'.
Shevlin appreciates the work on mechanistic interpretability by Neel Nanda's lab inside DeepMind.
Shevlin believes most of the AI ethics academic circle has been 'rather slow' to wake up to AI capabilities.
All Facts (8)
University of Cambridge philosopher Henry Shevlin spends 80% of his time at Google DeepMind, with the title 'Philosopher'.
50%UnverifiedShevlin believes the position of 'never allowing AI any autonomy' is too extreme.
50%UnverifiedShevlin points out that suppressing a model's self-attribution of mental states simultaneously suppresses the model's tendency to attribute mental states to animals, reducing concern for animal welfare and willingness to talk about spirituality.
50%UnverifiedShevlin believes LLM cognition has a network-like character, and wanting to adjust only one dimension is extremely difficult.
50%UnverifiedShevlin believes most of the AI ethics academic circle has been 'rather slow' to wake up to AI capabilities.
50%UnverifiedShevlin believes AI consciousness is the most difficult problem encountered by the science of consciousness, perhaps second only to David Chalmers' 'hard problem of consciousness'.
50%UnverifiedShevlin appreciates the work on mechanistic interpretability by Neel Nanda's lab inside DeepMind.
50%UnverifiedShevlin expects a fundamental shift in the pattern of human-machine relationships: from people talking with chatbots to dynamic networks in which multiple agents and multiple humans converse with one another.
50%