What Chappie Doesn't Know
Musings on embodiment, intelligence and dignity while visiting the Santa Fe Institute
I’m in New Mexico this week to visit the famed (and beautiful!) Santa Fe Institute and to participate in a meeting on AI and Human Multi-Scaled Intelligence. I’m excited to get to spend a few days with folks from Deep Mind and Open AI, as well as some of my favorite writers on AI, philosophy of science, and ethics.
The meeting has me brainstorming about the various ways to defend the idea that our embodiment gives our lives a special sort of dignity, one which AI will never possess. Within traditions like Judaism and Christianity there are a whole lot of theologically-grounded reasons for thinking our bodies are particularly special. But this is a meeting primarily of folks who work in and around frontier AI; there isn’t any common religious ground. So even though DELTA is rooted in Christian values, I think it is worthwhile for the DELTA Force to explore and debate ideas in the framework that don’t presuppose any specific religious commitments. We’ll play with these straight-up philosophical ideas from time to time in this newsletter.
To understand why embodiment matters, it probably helps to get clear on what we mean by “intelligence”. For today’s purposes, let’s define intelligence really simply as your ability to recall facts and skillfully manipulate information. For instance, your intelligence grows as you get better at remembering celebrity names and physics laws. It also grows as you get better at seeing solutions to math problems or fixing electrical problems in your house. This kind of intelligence can be measured, and now there are commercially available AIs which can absolutely smoke you on most of these skills.
In the DELTA framework, we don’t think that this type of intelligence is what makes human beings special or grounds our worth. But the fact that we live, think, and experience as embodied creatures – that is very important indeed. This raises two questions. (1) What kind of intelligence might we have in virtue of our bodies that AIs will never have? And (2) Is that kind of intelligence the right sort of thing to ground our moral worth and dignity – especially as we wrestle to understand that dignity now? Put another way: what good are our puny “wet brains” compared to these new silicon ones?
Here’s just one way of thinking about what’s special about our embodied way of knowing the world, and it is inspired by a classic thought experiment from Australian philosopher Frank Jackson. (If you are new to philosophy, we often use far-fetched examples to help you uncover and sharpen a philosophical belief you already have. This is one such wacky example that helps you think about an abstract question involving our minds.)
What Mary Didn’t Know:
Suppose that from birth, Mary was raised in a black and white room. She wore black and white clothes and black and white gloves and spent all of her days looking at a black and white screen. On that screen she studied every subject under the sun, reading articles and watching films, all in black and white. This includes studying every scientific fact about how light refracts to cause the color spectrum and how every species with eyes detect color. And Mary has a perfect IQ and inexhaustible attention span. She took in all of this information. She passed every physics exam inflicted on her.
Then suppose one day, the door to Mary’s cell is unlocked. She leaves the room and she sees with her own eyes a red apple on a table. Does Mary learn something new?
Many of us have the feeling that she does. Mary doesn’t learn a new fact about the nature of colors – by hypothesis she has studied all of this already. What she learns is what red is like.
My colleague Meg who writes here sometimes has a nickname for Chat GPT; she calls it Chappie. I find this hilarious.* Chappie knows a whole lot that we don’t. He’s aces at philosophy trivia questions. He can also do things that I personally do not know how to do. He’s damn good at coding up computer programs. He’s very serviceable at graphic design. I am miserable at both of those things. Chappie and other LLMs are similar to Mary, in the sense that they have had the opportunity to read almost everything conveyed in human language and to ingest a massive amount of spatial information, all translated into their language of data. They can pass the tests we throw at them.
What Chappie has never done is experience the world with his body. Because he has never had a body. I’d argue that Chappie does not know what red is like. He also doesn’t know what hunger is like. Or what it is like to love someone. Or what Hatch green chile tastes like. (FYI – delicious). Or what it is like to collapse with exhaustion after a transcontinental flight. The technical term for this is that Chappie lacks “qualia”. It’s an open question whether AI could have its own kind of qualia, but count me among those who hold that qualia are unique to embodied consciousness. It would take a really cool thought experiment to convince me otherwise.
So at a minimum, on this line of thought, you know some things that Chappie does not know and will never know. Around the Institute, when we work on a task without any AI help – which is often – we sometimes joke that we are deciding to use our “wet brains” (as opposed to the silicon brains). But the fact is, the good old wet brain has incredible capacities in virtue of the fact that it is a part of a body.
But wait, you ask, what if we gave Chappie better robotics? The ability to explore the world with wheels and propulsion and sensors? You can imagine us putting the AI into a small robotic interface and letting it out to bash about in the world, gathering data about whatever might come into its path. Still, I would argue, it would not know what red, hunger, love, and exhaustion are like – not the way that we do. Imagine Mary in her black and white room had a little remote control robot she could direct around the outside world doing experiments and gathering data for her black and white screen. She would know more facts about the world. But she still wouldn’t know what the world is like.
I would argue that the most morally important knowledge we gain about the world is the kind of knowledge that we only get with our bodies – the “what it’s like” knowledge. Aristotle says that the essence of being human is that we “take pleasure in our existence”, and indeed, one of the ways we come to love each other is by realizing that other people have these inner experiences. It is the reason our society is different from a herd of cows, who just eat side-by-side their whole lives. We share our inner lives with one another through attention and conversation. It is one of my favorite ideas in virtue ethics. (See Book IX of Aristotle’s Ethics).
This gives us maybe one reason to believe that advances in the ability of AI to systematize information should not feel particularly threatening to our dignity. Our experiential, “what it’s like” cognition is special to us, and more morally significant than our ability to recall information or optimize tasks. But this is a point that we will increasingly need to be reminded of, and that’s why virtue ethics is so important in AI ethics right now.
Many folks prize their intelligence in the sense of knowing facts and knowing how to do things. Indeed, we tend to value or devalue lives as this type of intelligence waxes and wanes. In 2014, UPenn bioethicist Ezekiel Emanuel published an article in The Atlantic: “Why I Hope to Die at 75”. He shares that he is winding down life-extending medical care because he acknowledges that aging beyond his 70s is likely to involve cognitive decline and increasing dependence on others. The article captures in a visceral way what Pope Leo calls the technocratic logic – the idea that we should only value our bodies insofar as they are vehicles for intelligence. I teach this article every year in my intro philosophy class and challenge my students to find counterexamples to this reasoning. For instance, how does Emmanuel think we should value people with cognitive disabilities? Does he think we have an obligation to try to soup up our intelligence? But most importantly, is there more to life than just working, being witty at the dinner table, being “quick on your feet”? Isn’t there more to having an inner life?
One of the beautiful things about “what it’s like” intelligence is that, for all we know, we had it when we were very small and we’ll have it until the end of our lives in these bodies.
There’s a lot more work to do trying to understand why qualia are so connected with moral worth and dignity. ButI have a conference to get back to – and Hatch chiles to eat.
Also don’t forget, tomorrow at 3pm EST I am doing a virtual panel with Archbishop Coakley (President of the US Catholic Bishops Conference), Bishop Paul Tighe (one of the Vatican’s leaders on AI policy) and others on Magnifica Humanitas. You can register here!
*I asked Claude to give me a funny nickname to use for it, and it suggested “Clod”, which I also find hilarious. I have a tendency to keep a bunch of AI tabs going on my laptop, and I am now going to call this ClodHopping.