Semblance of Meaning
[This is the third of three articles I’ve written about ChatGPT. You can find the others here: ChatGPT and Existentialism, Chat Bot Says it Wants to be Alive. Get in Line, Pal.]
If you have been following the saga of Amateur Polymath wrestling with the upcoming AI revolution, you may have an inkling as to which questions bother me: Does the chatbot really understand Existentialism? (ChatGPT and Existentialism) Does it really feel negative emotions towards being stuck in a computer? (Chat Bot Says it Wants to be Alive. Get in Line, Pal.) These questions take up mental space for me, because I am fairly certain that AI cannot do these things, but questions around consciousness are infamously difficult in psychology and philosophy. One thing is certain: AI today is great at creating a semblance of meaning. It bears witness to an incredible amount of data that shows where human thought stood on February 2023 (or whatever time period it was trained on). Therefore, it can imitate with shocking precision the ways in which we use our language and communicate. But I believe it lacks a crucial ability that would be required to be considered “intelligent” — it cannot change.
In existentialist thought, the one crucial capacity humans possess is the capacity for change. Kierkegaard says: “The moment of decision is madness”. Decision making is not some deterministic function — we cannot simply analyze all the inputs and spit out a solution. Of course, we make our little pros and cons lists, but we rarely go ahead and consider the decision made once the list is complete. There is something else inside of us, some sort of “madness factor” that impacts our decisions with no control from our conscious minds. This madness factor gives humans the ability to wake up one day and radically change their lives. This is why Robert Downey Jr. can become a super hero, and Kanye can become a nazi.
Imagine a friend from middle school that you haven’t heard from since. Try to visualize what they might look like today. Now imagine if your friend actually knocked on the door and came in. The person that greets you would most likely be incredibly different than the person you visualized. The way you visualize them in your head may be the equivalent of how an AI model would represent them: stuck how they were when you knew them, maybe just pulled from the corners to be resized into a full adult human shape. But humans don’t resize; we grow. We change in ways that are not predictable; our growth is non-linear. I wont’ make the more ambitious claim that a machine can never have the capacity to grow in this way (although I do believe it), but that’s not how the AI we built works.
So why does this matter? A lot of our daily conversations only have a semblance of meaning, not just with machines, but with humans too. We have an explicit category of conversation named “small talk”, where every response is selected from a set of default options. When your coworker asks you how your weekend was, you have very few options to choose from. Small talk excludes the possibility of an unexpected response; the form of the conversation is pre-scripted. We go through a sizable portion of our days in scripted language: We make small talk to our baristas, hair dressers, acquaintances, etc. And it is clear that today’s AI has the capacity to have this level of conversation. However, the moments where we are really alive are those where we don’t know what will happen next. Where there is risk involved. Where the possibility of change is inherent, for better or for worse.
For some of the more mundane tasks of our lives, a frozen understanding of the world might be sufficient. AI may be able to take on tasks in insurance, bureaucracy, even parts of law. Surely, this is important work that is valuable to automate. But I doubt that creative human thought is reproducible in this manner, and so I doubt that AI can create art. The point or art is to find creative ways to navigate and respond to the world. If that world now includes chatbots and image generators, so be it. I don’t think the artist gets to say “I don’t like how the world is changing; it breaks my art”. True art (la di dah) evolves together with humans. Creation may stem from replication, but it does not end there. Today, AI replicates, and replicates well, but it does not create.
[PS: If you are interested in questions around consciousness, I recommend looking into the Hard Problem of Consciousness. If you are interested in consciousness of AI, take a look at the Chinese Room Argument and the Turing Test.]
I love the positive and constructive tone of your articles! Rather than being a destructive force, AI might, if we have the ambition to outdo it, have given us the opportunity to think outside the box (the box being the literal computer hehe)