A succinct discussion of the current state of our understanding of consciousness. I love seeing things like this in mainstream media.
Interestingly, there’s also a reference to AI risk at the end:
As to conscious AI, Yoshua Bengio of the University of Montreal, a pioneer of the modern deep-learning approach to AI, told the meeting he believes it might be possible to achieve consciousness in a machine using the global-workspace approach. He explained the advantages this might bring, including being able to generalise results with fewer data than the present generation of enormous models require. His fear, though, is that someone will build a self-preservation instinct into a conscious AI, which could result in its running out of control.
This paragraph probably leaves readers with two misconceptions.
The wording implies that Bengio’s main worry is deliberate coding of a self-preservation instinct, whereas the more prevalent concern is instrumental convergence.
Readers may conclude that consciousness is required for AI to be dangerous, which is not the case.
It also would have been nice for the article to mention the ethical implications for how we treat nonhuman minds, but that’s usually too much to ask for.
Perhaps someone with better credentials than me could write them a letter.
Animal and AI Consciousness in The Economist
https://www.economist.com/science-and-technology/2023/06/28/thousands-of-species-of-animals-likely-have-consciousness
A succinct discussion of the current state of our understanding of consciousness. I love seeing things like this in mainstream media.
Interestingly, there’s also a reference to AI risk at the end:
This paragraph probably leaves readers with two misconceptions.
The wording implies that Bengio’s main worry is deliberate coding of a self-preservation instinct, whereas the more prevalent concern is instrumental convergence.
Readers may conclude that consciousness is required for AI to be dangerous, which is not the case.
It also would have been nice for the article to mention the ethical implications for how we treat nonhuman minds, but that’s usually too much to ask for.
Perhaps someone with better credentials than me could write them a letter.