Max Tegmark mentions consciousness in his recent podcast with Lex Fridman. The idea that recurrent neural networks might be conscious (as opposed to the current architecture of LLMs—linear feedforward transformers) is intriguing. But whether it helps with AI Safety is another matter. The consciousness is highly likely to be very alien. And could be full of suffering as much as it is happy—and if the former, a massive threat to us. Also it seems highly unethical to experiment with creating digital consciousness, given the risk of creating enormous amounts of suffering.
Max Tegmark mentions consciousness in his recent podcast with Lex Fridman. The idea that recurrent neural networks might be conscious (as opposed to the current architecture of LLMs—linear feedforward transformers) is intriguing. But whether it helps with AI Safety is another matter. The consciousness is highly likely to be very alien. And could be full of suffering as much as it is happy—and if the former, a massive threat to us. Also it seems highly unethical to experiment with creating digital consciousness, given the risk of creating enormous amounts of suffering.