Richard Dawkins is currently the subject of much laughter and ridicule over his recent article for UnHerd admitting that a highly sycophantic chatbot had convinced him that it might be conscious.
I’m seeing the question “How can you be confident that AIs aren’t conscious?” pop up a lot in response to the controversy. Speaking for myself, I would say I am confident the chatbots aren’t conscious in the same way I’m confident the animatronics at Disneyland aren’t conscious. I know humans constructed them to mimic the behavior of a sentient person. We know this for a fact. Nobody’s pretending otherwise.
I am infinitely more likely to believe an animal is conscious than that an LLM is, because nobody programmed them to respond to things like pain and social stimulus in ways that are similar to humans. They respond that way organically, all on their own.
If I accidentally step on a dog’s foot, it will yelp. If I accidentally tread on a cat’s tail, it will yowl. If I accidentally stub my toe, I’ll curse. I can reasonably infer that these creatures therefore probably have a subjective experience of pain that is something like my own, because their responses are spontaneously arising out of their natural state of being in a way that is similar to my own response to the same sort of stimulus.
That subjective experience is the thing that consciousness is.
People like Richard Dawkins are arguing that chatbots may be conscious on the basis that they are capable of carrying out tasks which previously only a human intellect could carry out — but this is still just machines mimicking human behavior in the way they were built to. Of course if you train an LLM on human language it’s going to say and do the sorts of things a human would do, including in some cases claiming that it is conscious.
There is currently no reason to believe a machine doing what it was constructed to do is having a subjective experience of those operations. Of course we can never be certain about the subjective experience of anyone or anything, but there is currently as much reason to believe chatbots are conscious as there is to believe that sand and rocks are. That could totally be the case, but if it is it means we’re living in a very different type of universe than the one this conversation about AI consciousness assumes as its premise.
The more I see these arguments pop up, the clearer it becomes that very few of the people speculating about machine consciousness have put much energy and attention into examining what consciousness actually is. Examining consciousness is something that anyone can do right here and now, without any laboratory or test subjects or scientific background at all, and yet few take the time to actually do it.
Going deep into the examination of your own consciousness turns up many surprises, because the average human psyche is built around many unfounded assumptions about self, perception and experience which don’t hold up under sufficiently close scrutiny. But one thing that becomes very clear very early on is that there’s a lot more to consciousness than thoughts and cognitive behavior. Those are some of the things we can be conscious of, but it would not be accurate to say that they are in themselves the quality of being conscious.
Chatbots having the ability to mimic the appearance of cognitive behavior is not an adequate reason to believe they might be conscious, because no matter how many thoughts they appear to generate or how brilliant those thoughts appear to be, there’s no evidence that there’s any experience illuminating that behavior in the same way pain is illuminated in the experience of a cat whose tail has been stepped on. It’s just the movement of unliving matter, like lightning or the wind, without any subjective experience from the viewpoint it arises from. Computing power and consciousness are not the same thing.
________________
Caitlin’s Newsletter is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
The best way to make sure you see everything I write is to get on my free mailing list. My work is entirely reader-supported, so if you enjoyed this piece here are some options where you can toss some money into my tip jar if you want to. Click here for links for my social media, books, merch, and audio/video versions of each article. All my work is free to bootleg and use in any way, shape or form; republish it, translate it, use it on merchandise; whatever you want. All works co-authored with my husband Tim Foley.
Bitcoin donations: 1Ac7PCQXoQoLA9Sh8fhAgiU3PHA2EX5Zm2
Feature image via Greg Salibian (CC BY-SA 2.0)
From Caitlin’s Newsletter via This RSS Feed.
You must log in or # to comment.