Is your AI sentient? If Anthropic’s Claude or OpenAI’s Chat-GPT is conscious does it have a kind of soul? If it does have something like free will and even a soul, should we be granting AIs the same moral rights we give ourselves?
Last week the New York Times ran an amazing story about Anthropic bringing in religious scholars from across the globe (and across faiths) to discuss variations on the questions above. It is a move that seems like something out of science fiction. What it actually reveals is an unspoken philosophy that, I believe, has been baked into the field of artificial intelligence for a long time.
Understanding that this perspective is just that - a philosophy - and not “what science says” is another essential step in protecting our humanity as we enter this dangerous era of potent AI.
AI can do amazing things like make this image. But can it be sentient? and how can we see the assumptions hiding in these questions.
As a scientist - or just someone who reads a lot about science - there’s a basic perspective about what’s real that always hides below the surface. It goes by a lot of names but often it’s called “materialism”. The idea is that reality is just physical stuff. If you made a complete inventory of all that physical stuff then you’d exhaust the Universe’s possibilities. Just as important, the most elemental units of that stuff are all you need to explain everything you see (an approach called “reductionism”).
The problem with these views is not that they are wrong, it’s just that are literally views. They are a metaphysics - i.e. “beyond physics”. Not only are there other kinds of metaphysics (like Idealism which claims reality is a kind of All-Mind), but there are entirely other ways to frame the question. This includes seeing materialism (or idealism) as just limiting concepts (words) that can never exhaust what happens in human experience.
So how does this relate to AI and the question of souls?
Inherent, hiding and unspoken in the much of the AI field is the assumption that the human brain is just a kind of computer. In other words human experience must be reducible to computations which, in turn, must be reducible to matter and motion. More importantly, this implies that the particular matter “substrate” on which the computations occur is irrelevant. Humans use neurons… so what? The exact same kinds of computations and hence the same kind of sentience we have can be carried out with any kind of computer (i.e. silicon chips). So if humans have this capacity for consciousness experience - including moral dimensions of that experience (free will, suffering etc) - then it must be true that machines performing the same calculations must have those moral dimensions as well.
This is a really good book about why the brain is not a computer.
My point today is that kind of assumption has been thoroughly baked into much of the training AI scientists (and many cognitive and neuroscientists) get. Given that training, it’s is a pretty simple step to the idea of soon-to-be sentient machines, the need for AI moral consideration or even AIs having something like a soul.
We should be very, very skeptical of this kind of quick step hidden philosophical thinking. There is only one kind of “thing” in the entire Universe that is sentient, can have consciousness and display intelligence. That kind of thing is called Life. As I have discussed many times in this Newsletter, we are not even close to understanding organisms and the remarkable self-organization that makes their sentience, consciousness and intelligence possible. That means the burden of proof falls like a ton of bulldozers on anyone claiming that something so utterly unlike life could instantiate it’s most mysterious and extraordinary capacities.
To wrap this up there is, I’d argue, a kind of misplaced religiosity in a lot of Silicon Valley tech-bro culture. They’ve been waiting expectantly for their computer generated “Singularity” for decades: the coming of superintelligent god-like machines. They want it. They’re primed for it.
We don’t have to be.
========================================================
Please support Everyman’s Universe by sending this post to couple of friends and suggesting they subscribe (I mean if you liked it of course)
Thanks!
========================================================
PS: If you have specific questions or issues you want me to address leave a comment on the website or email me at [email protected]
PPS I could not get this proof-read so please excuse typos etc.
— Adam Frank 🚀
