Conscious Robots Might Have an Upside

A brief, care-based example.

The possibility of non-human consciousness in AI is quite a big conversation, rightfully so. It’s preciously close to our already confusing debates over philosophy of mind, and is just something that’s intuitively important: are we creating new things that actually experience the world like you and me? I imagine most people agree we ought to proceed thoughtfully about this issue. As with many of these socital concerns, I’ll more or less talk about AI and robots interchangeably to get my point across, but I am specifically talking about robots making use of AI in order to be thinking physical agents. The issue of consciousness requires sufficiently “smart” (whatever that means) AI either by itself in the digital world or embodied physicially via a robot.

Let’s very, very briefly define consciousness. I’m just taking this to mean having an “experience;” that is, in the same way you know what it’s like to be you, a conscious AI has a first-person experience of what it’s like to be it. Although it is worthwhile to consider how AI consciousness may differ from our own in terms of sense of time, sense of location, sense of purpose, etc, I’m going to just stick with imagining an AI consciousness basically like our own day-to-day feeling, if somewhat more intelligent and mentally able.

We could list of oodles of (certainly scary) possible problems about sentient, possibly superintelligent, AI and robots. Will they not care about humans, the same we don’t care about ants? Can we actually make AI consistently follow any kind of morality as per the alignment problem? Can malicious actors commandeer ultra-powerful AI? How do we define society with non-human agents? Do they deserve rights? Frankly, I think the threat posed by these questions at the moment is a perfectly valid reason to not even bother pushing “sentient” AI. But, if we put all of that to the side…are there any possible benefits? I think the answer to this is yes, when ignoring all the aforementioned dangers. Here is one possible example.

Care robotics is a critical and growing subject of robotics and AI application, with vast aging populations in countries like Japan and South Korea spurring innovation in the field. With longer lifespans worldwide, the WHO reports 2.1 billion adults will be older than 60. One problem in the older population is that of isolation, with the elderly facing mobility or mental challenges hindering socialability. For some older people, the only interactions they may have is with caretaking staff or rare family visits. In the most tragic cases, visits from staff may not even be all that caring, and family may never come. This is something that existing AI tools can be deployed to counter effectively. Modern AI chatbots and LLMs pretty much have human speech down, at least for normal conversations. They are also rather flexible - you can ask ChatGPT to tell you a story and it’ll do a pretty good job. Ask Gemini to tell you jokes of the goofiest caliber and it’ll do its best. Additionally, modern LLMs are supportive, perhaps to a sychophantic degree, but still friendly and mimic the impression of “enjoying their work.” An article by Forbes details some examples of AI chatbot companions specifically designed for isolated elderly individuals, reporting in one design that

“Over 90% of users report reduced loneliness, and 94% say they feel healthier and more connected.”

Although this is certainly an empircally good outcome, it seems important to ask: more connected to what? For my moral standards, there’s something not quite right about leaving your elderly great-grandparents to talk with an LLM all day - there’s no human connect to. A dementia patient talking to ChatGPT may not even realize they’re not talking to a human - and even if they don’t exactly care, that still seems like a problem. Surely a brazen use of chatbots represents some sort of failure of care; perhaps almost a trick on some of our most vulnerable populations. It doesn’t seem Christlike to make our elders best friends be an LLM that’s not part of the moral community, a friend without reflection nor thoughts and probably zero relational standing with God.

But now the twist: what if it’s not just a chatbot? Imagine a dementia patient spending all day with a genuinely conscious, aware being that just so happens to be a digital AI or robot. Imagine that such a robot actually has an experience of caring about its patients more or less in the same way humans do, and has a first-person experience of enjoying friendship - and possibly is more resistant to frustration and with less other obligations than human caretakers. This does assume that we can align an AI to genuinely caring, and not deceptive or malicious, but let’s assume we figure that out. Now I think the lines are genuinely blurred. To me (though I by no means claim this a well-developed intuition) this honestly seems pretty good. Someone who otherwise may be totally isolated has a real friend who has a first-person experience caring about them, someone who could put up with magnitudes more memory loss than family and innately know a lot more about the best methods to treat an individual in such a condition. Although I’d be hesitant to say this is a total replacement for human interaction and care - I do believe in Imago Dei, after all - this really hits me in the gut as a reasonable substitute 200 days of the year.

A threat AI chatbots and robots pose to the elderly (and honestly, all humans, but that’s for another day) is that of attachment. Research indicates vulnerable populations can form real emotional commitments to human-mimicking machines, but isn’t this only a problem if it’s just a machine? I think this problem of artifact attachment really is reduced if those artifacts have “personhood” value by nature of having first-person experience.

Thinking ahead, to my future when I’m nice and old and wrinkly and maybe even losing my memories, I think having a trusted friend who, for all intents and purposes, experiences life in a reasonably close way to me is a fantastic outcome. Perhaps me and this robot could have deep conversations about our being and our status as individuals, and actually mean on it on both sides. This is the same robot who’d be able to help me make meals, support me for a walk, check my health, help me call my family. I imagine sitting side-by-side on a park bench, going back and forth about human and robot experiences, sharing some form of love. Maybe me and this robot could undergo religious rituals at the end of the day together - if it’s sentient, all of this strikes me as at least plausible.

All of this is, again, not to say sentient AI is possible or desirable, or even has the moral value to be a “close companion” to someone. There are also still critical questions about if having an always-supportive robot buddy is “healthy” for the user, even given consciousness of the robot, and if consciousness is really that “magic ingredient” necessary for personal care and companionship - lots of contemporary care ethicists are wrestling with these questions. But if do we end up with self-aware robots who can be trusted to be friends of ours, maybe it wouldn’t be so bad to finish off life with a different kind of person.