We've all heard that a million monkeys banging on a million typewriters will eventually reproduce the entire works of Shakespeare. Now, thanks to the Internet, we know this is not true.
-Robert Wilensky
Robert Wilensky's observation notwithstanding, today we are witnessing the proverbial million monkeys work their way towards the complete works of Shakespeare. They are accomplishing this task not because they are infinite in number, have infinite time, or exist in an infinite universe but because they represent a new type of intelligence. And while this new intelligence is capable of some amazing things, it is still fundamentally different from human intelligence.
Although it is easy for us to slip into thinking of our chatbot and AI Art partner as human, there are still breaks in this facade. Occasionally, the Chatbot will say something exceedingly odd that, for at least a moment, challenges the illusion that we are talking with a friend. Or produces an image that no human would produce. Artificial Intelligence enthusiasts know this is because, as advanced as generative artificial intelligence applications have become, they still lack general artificial intelligence. In other words, these applications can do amazing things, but they lack common sense.
The image above illustrates this perfectly. ChatGPT was given the instructions to research the great Brazilian artist Emiliano Di Cavalcanti and create an homage. Using Dall-e3, ChatGPT produced this pleasant image of fishermen in a harbor at dusk. At first glance, this appears to be a promising draft of an homage to Di Cavalcanti. If an art student had been given an assignment to create an homage to Di Cavalcanti and had produced this, it is easy to imagine an instructor pointing out the things they like and ways the work could be improved and sending the student on their way. But no human art student would produce this.
Let's start with the fisherman working on the fishing net. The fisherman has a foot where his left hand should be. Early AI Art was riddled with such "AI-isms". The more recent AI algorithms do a much better job of avoiding these issues, but they haven't disappeared entirely.
Another more subtle "AI-ism" is illustrated by the two fishermen wading in the water. The water appears to come up to just over their feet. Given the position of the boats and what we know about harbors and docks, this seems far too shallow. Now, if someone with more knowledge of harbors were to tell me that, actually, at low tide, this scene makes perfect sense, I would accept it. But until then, I will hold this up as another example of AI looking at the image through a different lens than you or I. Our life experience tells us something is off with the fisherman and the waders, but the AI has no life experience to draw on.
Once you start to look for "AI-isms" it can be hard not to see them everywhere. In the image below, for example, the windmill seems out of place, the carriage on the canal is missing any form of propulsion and appears to be a disaster waiting to happen, and the people in the boat seem doomed to make little progress until Spring. And that is only a cursory examination of the image.
AI, the Scatter-brained, Alien, Creative Assistant
Admittedly, though, AI Art has made great strides at addressing these "AI-isms". And as these issues become fewer and fewer, it will be easier and easier to think of AI as having a human-like intelligence. We will need to remind ourselves that just because an AI application can do one thing very well, this does not necessarily mean it will do another thing well - even if that other thing is, to us, much easier. Until AI acquires a general artificial intelligence, we will do well to think of AI applications as brilliant but scatter-brained assistants, capable of doing some incredible things but also capable of bungling the simplest tasks and going off on a tangent at any time. In other words, we should heed Ethan Mollick's advice in "Co-Intelligence: Living and Working with AI" and always make sure we are "the human in the loop", closely directing and supervising the work of our AI Assistants.
Finally, what should we make of these "AI-isms"? Yes, they remind us of the limitations of AI Art. They point to a lack of human intelligence and awareness. But they also point to a different type of creative process. A process that should not be romanticized, but can still be appreciated for what it is. And a process that can be extremely rewarding for those who participate in it.
-Peter Kreeft
Comments