Tamir compares the brain to large language models: both predict the next piece of information, and both can 'hallucinate.' No model is perfect, but models are either improved or left unimproved.
Where it appears 1
Connections
Related
Explored through
- Minds hallucinate like language models
Like an LLM, the brain predicts and can hallucinate, so we have to sit down deliberately and check that our predictions and interpretations aren't simply wrong.
There is no perfect model, but there are improved models, and then there are models that do not get improved.
- Imperfection is the human condition
If you're looking for growth, it isn't because you're broken but because being human means nothing is ever perfect, so there's always more to interpret and the model can always improve.