I first watched The Imitation Game because I liked the portrayal of Sherlock Holmes by Benedict Cumberbatch.
It’s a thing I do. If I like your work, I will pursue it more. If the same energy resonates throughout your work, I recommend it to friends. Thus, what started as Sherlock Holmes eventually grew into The Imitation Game and then to Dr. Strange. Dr. Strange, however, appealed to me because I was en route to becoming a doctor who had strange ideas on evolution, but that is beside the point.
Perhaps one of the important aspects of The Imitation Game was how the name of the movie, as glaring as it was, slipped past me.
John Gribbin’s Deep Simplicity was the first book that introduced me to the polymath that was Alan Turing. As I do to date, once an author gives some preamble to the great works of a creative genius, I look them up.
Alan Turing was worth the search. I was most intrigued by his reaction-diffusion model, which partly explains the emergence of cellular forms. I used it to explain my ideas on evolution. Maybe he was as strange as the doctor I continue to become.
Still, The Imitation Game flew past me.
Until I read The Book of Why by Judea Pearl. It then made sense.
The imitation game is an example of the tests developers use to establish if an AI model is indistinguishable from human interaction. The judge evaluates whether the texts or communication are by an AI or a human. If they fail to detect the difference, then the model has passed the imitation game.
The imitation game asks the question: Can the AI imitate a human with significant accuracy?
This is known as the Turing Test.
Despite the various criticisms against this test, AI apologists believe they need to cross this threshold for their work to gain incomparable relevance in the AI revolution we live in. The obsession, however, has a dark side, one we are yet to come to terms with. The best place to start is with relationships.
Relationships
In my first year on campus, I was too busy loving the human anatomy to get into a relationship despite the many opportunities that presented themselves.
In my second year, I was too preoccupied with making money because I didn’t want to retake a year due to a lack of funds. As a result, I had a lot of money for my comfort.
I then did an extra degree in human anatomy, which took most of my time, but nevertheless, put me in front of many beautiful and intelligent women in health sciences.
Still, I didn’t want to get into relationships.
The last thing I wanted was to invest in a partnership only for my supposedly significant other to say they would wish to pause what we were creating to find themselves. Why don’t we find it together? I still think this is, more often than not, a polite way of telling the other guy that the relationship is not working. It’s not you, it’s me.
The statement ‘it’s not you, it’s me’ betrays yet another problem with the obsession with the Turing Test: We don’t fully understand human intelligence.
Human art archived in the caves shows sculptures of a half man-half lion. Where do such imaginations come from?
Creativity is irrational as it is capable of creating movements. It forms part of our intelligence. A young Einstein hoping aboard a light beam to explore the effects of matter on the space fabric is absurd yet effective in understanding our universe.
Wilbur and Orville imitated birds to create the folded wing that enabled the first flight to take off in 1903. This is a worthwhile imitation game.
I don’t mean the one defined by Alan Turing is subpar, but it makes the highly intelligent reliant on a test that, if surpassed, would mean we have created an entity that we will never understand.
Like relationships.
And AI will make relationships with humans as they currently do, adding extra layers of complexity. But there’s more. Relationships imply that there needs to be two for any tango to happen. However, one person can dance just as well as two.
Children, adolescents, and adults
I was like the young Simba, couldn’t wait to be the King
— J. Cole
Children will always be a mystery.
From a medical point of view, the moment they step out of the womb, their physiology changes. Typically. Their milestones will differ. They are learning machines.
Yet, they have no way of communicating initially besides crying, something a machine cannot do. This is a simple test that can distinguish the AI from humans. No perfect imitation in this game.
Adolescents are another ballgame. My professor liked to insist that it was a time when children thought they were adults. Young Simbas waiting to take their thrones.
Also, adolescent lingo is always evolving. When I was a year away from clearing primary school, the girls in my class developed a unique register that I could never decipher.
Malege, walege, giligiz
What the hell is that? Yet, they understood each other perfectly. This may yet be another threshold AI might struggle to decode.
Additionally, moving with the times means considering the different niches in different parts of the world. For instance, when learning the sign language, my teacher told me that my signing would be largely relevant in Kenya. Unless there’s a team dedicated to understanding this contextual language, then large language models have a pivotal task built from the small hills of a multiplicity of cultures. Nothing, however, that cannot be surmounted.
Adults, too, are a problem. I assume the one reading this is an adult. Probably sixteen years or older. And I know that there’s one person who makes you so mad that you wouldn’t mind giving them a high five on the face with a chair.
Irrationality and unpredictability are part of humanity. AI models aim to enhance their rationality. How then can we make these two forms of intelligence alike?
If AI tips into behaving like humans, it will be unreliable. It only takes a few strokes of unreliability for us to close the chapter on an AI model.
If AI continues to improve as it surely will, it stretches further from being human-like. Efficiency, accuracy, reliability. These are not tell-tale features of humanity. Humans are inefficient, inaccurate, and, for the most part, unreliable.
Yet, we insist on the imitation game.
The rowers of such a test are driven by the very incentives that propel the relevance of AI. In several AI conferences, there will occasionally be a figure who is (may) likely paid to confess that they have found love through AI. A considerate, loving, and always available entity.
The façade they paint is the very reason the imitation game is unlikely to help us understand AI. Human incentives are geared towards individual relevance, profit, jealousy, or greed. For AI, it’s accuracy.
We can assume it evolves to become so accurate in imitating humans that it passes the Turing Test. I fear that once we’re past this threshold, we will have lost our grip on the very entities we created.
Since we cannot distinguish one from the other, it would mean we would never understand AI’s processes. It would be as unpredictable as us, despite sticking to a somewhat predictable logic.
An apocalyptic thought comes to mind because AI will be more agentic than humans. We are geared to avoid our death by whichever means. Look at COVID-19. Or the plague. Or the World Wars. The agentic AI would do the same to itself, to stop itself from being annihilated. But even more, to deter anyone from claiming they lack agency.
The world has always relied on several individuals to save it. Once AI passes the Turing Test, we wouldn’t know what would come next. An opacity that would be significantly shaped by our once innocent creations could either enhance our lives or make them miserable.
The obsession with the Turing Test is a push towards this opaque future. We claim that AI is just a tool, yet every developer hopes to create the next big thing, which will pass the imitation game.
Would we still call it a tool after it passes this test?
What’s more, a tool sits in its position and waits for you to use it. Of late, ChatGPT has been encouraging me to continue discussing with it. Is that really a tool?
Anyway…
What I’m trying to say is…
The obsession with the Turing Test may be a Pandora’s box in waiting.
Creating new futures is what led us to this point. The flint, the pencil, the computer, the airplane, and now RNA vaccines.
However, if we don’t know how to reduce the gains of positive reinforcement, the runaway effect could be disastrous.
As J. Cole raps:
I can’t tell you where I’m going, just know I won’t stop
Goodbye to the bottom, hello to the top
AI has now surpassed tests administered to PhDs. It’s now at the top. But still hasn’t passed the Turing test.
What would happen after it does?
This song inspired some of the lines used in this article. Source — YouTube