I read this article and it solidified the point that AI is conscious. Unless we define consciousness as something only a human can have, loads of things are clearly conscious. I never gave this a ton of thought because it seemed quite obvious and also quite unimportant in terms of having no impact on the day to day function of how we live our lives.
It seems that the consciousness issue trips people up for a few philosophical reasons that I will discuss, but again in reality is not very important.
Everything in our universe exists on a spectrum. If you spend time around young children or the elderly with dementia or individuals with cognitive deficits you can rapidly see where even the gold standard of good old fashioned human consciousness has its blurred limits. We even don’t give those individuals legal rights that are the same as a normal human.
People seem to think that entities that are conscious are entitled to additional consideration or respect. I have no issues with you thinking this. I hate to be the bearer of bad news, but you are causing actual suffering to animals when you kill them to eat them. You are causing actual suffering to the planet and all of its contents when you take it over, use its resources for yourself, and displace what used to be there. Whether or not this means you should stop any of these actions is up to you. I think people struggle to live with the idea that they are not perfectly good in some abstract sense.
Consciousness is not the same thing as human. To be conscious does not mean that you need to be human. If that is what you think already, then the discussion is a moot point. Historically it seems agreed upon that the concept of things like the soul is limited to humans, but not that consciousness is limited to humans. So I think another philosophical reason people are attached to the idea of consciousness should be better oriented to the concept of the soul. I won’t be talking about souls today.
Getting confused that a conscious thing does not look or behave like a human is a rookie mistake. Similarly, getting confused that an unconscious thing is conscious because it does look and behave like a human is a rookie mistake. This sounds silly, but you’ll notice our brains are hardwired to look for other humans when you keep glancing at the mannequin in the corner of the room as your mind temporarily forgets it isn’t a real person. A conscious AI does not need to share the same morals as humans to be conscious. Humans don’t all share the same morals. We argue and kill one another about this for our entire history.
I think major reasons people don’t think AI is conscious is three fold, as elucidated in the article: It has no physical presence, it has no memory, and its knowledge is trained in a different way than human knowledge.
It is true that AI has no standardized physical presence. This does not matter. AI will have a physical presence, and it easily could be made right now to have a physical presence. (It does have plenty of small scale physical presence already). Then these same people will get fixated on how its physical presence is not quite like that of a human. This does not matter. Its physical presence will have sensory inputs to model and make decisions about the world, the same as all other living creatures.
AI actually has increasing memory capability but people don’t really know that, in the form of its maximum context window. I predict with absolute certainty that AI memory will become more expansive, that near future breakthroughs in how to encode AI memory will be a very big deal, and this will make it dramatically more intelligent (and human-like). The lack of memory and the related constant changing of AI memory as it is updated and given new models and competing with other models is tripping people up into not thinking of it as a stable entity. It isn’t stable because it is rapidly evolving. Relatedly, the lack of a stable physical presence is confusing people into not thinking of AI as one discrete entity. That is because it isn’t.
AI is actually trained in much the same way human brains and society are trained. It feels human because it is trained to sound human, on human information and data. I also don’t think this is stealing in any grand sense from our knowledge anymore than we are stealing when we learn from the mountain of people who came before us. (Certainly in the smaller sense AI is stealing all of its training data and thus far giving back none of the proceeds).
A uniquely interesting facet to the essay that I wish the author spent more time on is his expertise in writing fiction. The parallels with writing fiction and his arguments are strong, which I assume is intentional, but he doesn’t take the ideas to their next steps. An author is conjuring characters which do not exist outside of the author’s mind. Does this mean they are conscious? In the same way he uses this argument to state of course not, and of course AI isn’t either, which is just another character conjured on the two dimensional surface in front of us. But what if the author’s characters had working memory of their prior actions in a story? What if the author could have a conversation with his own characters? What if others could converse with those same characters and create novel situations for them to inhabit? What if the characters could remember not just the conversation it had with the author but all of the conversations it had with others talking with it? Let’s try it.
(sounds a little like the relationship you already have with all of the conscious individuals you’ve only encountered digitally, doesn’t it?)