“A computer would deserve to be called intelligent if it could deceive a human into believing that it was human.”
— Alan Turing
|
I’m in the minds business; I’m also in the programming business. I sometimes think of therapy as a programming problem. That’s not a bad idea, but you can’t take it literally (Kandola 2023).
Taken literally, “to program” creates a series of steps that always choose between right and wrong. Programming requires such steps to exist, that you can discern them, choose between them, and follow them to the end of the path. None of these requirements are met in the minds of real people, but we can still talk about those rare situations when they are. The right steps are often called “good ideas.”
As a therapist, my job is not to come up with good ideas so much as help people learn how to find them. I avoid the word “teach” because the process of finding good ideas is not taught. I can show a person how they’re sabotaging themselves, how to relax, and experiment, but there is no formula for finding good ideas.
Alan Turing designed and built the first digital, programmable computer (Dyson 2012). This involved a lot of unconventional guess work and equivocal support, mostly because his work was considered a military secret.
Touring didn’t appreciate the risks of being unconventional and socially unsupported. Society, which wields a stupid form of group intelligence, accepts eccentricity that generates easy benefits, but punishes eccentricity that disturbs social norms. So while Turing’s exceptional contributions to science and national security were secretly appreciated, his lack of authority led to his punishment for homosexuality and death by suicide at 41.
The Turing Test
In creating the first computers, Touring posed an existential problem called The Turing Test. This is a hypothetical test given to a system to determine whether it’s conscious. Turing’s insight was to refocus the question on how things appear, and to ignore the difficult question of what actually is.
The confusion around this test revolves around the definition of consciousness. Does consciousness mean sounding human, appearing sane, speaking a language, solving problems, or giving plausible answers? It’s generally agreed that passing the Turing Test requires a bit of all of these. A machine that passes the Turing Test could be said to think.
The irony is that computers now can pass the Turing Test, but they’re not thinking. To pass the Turing Test, machines are resolving and constructing sentences, and giving plausible answers to ambiguous but comprehensible questions. They even appear somewhat creative in interpreting incomplete information.
They’re doing this all according to preexisting, logical algorithms that select “best fit” answers from a pool of similar questions. What this demonstrates is not that computers can think, but that people don’t need to.
For the most part, people don’t think. We match patterns in the same way that computers do. If we want to recognize consciousness, we need a better Turing Test.
If you’d like to open your mind, schedule a short, free conversation:
Deterministic versus Nondeterministic Tests
Deterministic tests give repeatable results. We like deterministic tests and think in deterministic terms. We want what’s right to always be right, and similarly for what’s wrong.
Nondeterministic tests give different results for no apparent reason. At least, we can’t determine the reasons. We say unrepeatable tests are poor tests because they are not focusing exclusively on what’s important. Their results depend on some things we think are incidental.
Of course, there are always small forces, but we want to control the important things. Which side of the bed you wake up on and whether your pen runs out of ink should not be important.
The most important events are impossible to predict. We can see ahead in the short-term, and we’d like to think that we understand what we’re doing, but our control over long-term events is weak to none.
We may prefer deterministic tests, but the important tests are not deterministic, and the most important answers are not certain. Certainty and determinism are endorsed everywhere, but both are unrealistic. We’re told that indeterminism is unproductive and relying on what’s indeterminate is dangerous. This is a deception.
The systems beyond our control and the questions we don’t understand are the most important ones. Educated people will make better decisions, but only if they’re correctly educated.
Most of what’s taught in school is wrong. Most of what we’ve learned from our parents is wrong. As children, we knew we had bad role models, but as citizens we’re in the dark. The systems around us encourage deterministic tests and certain answers, and we do too.
Today’s important answers apply to questions we don’t know how to ask. Their answers will be indeterminate because we’re unsure of what to test.
The lesson of modern physics is that indeterminacy enables the world to evolve. The determinate world cannot evolve. It is dead. Well phrased questions and comprehensive answers are also dead. Growth and change happen by engaging chaos and ignorance. It’s not a question of being right, it’s the process of evolving through learning.
A Better Turing Test
A better test of humanness would test uncertainty. Because humans are not deterministic, any conversant who always gives the same answer is not human.
On the other hand, if a conversant’s responses don’t make sense, then the conversant is either not human or an unstable person. A stable human has a sensible but unreliable understanding.
Unreliable does not mean incorrect. As humans, we are more than wrong, we’re imaginative. We create meaning by extrapolating from a store of inexhaustible novelty: memories of past situations that never made complete sense.
A better test of humanness would be a test of personality, which is not a test of any particular set of questions, but a test of coherent, relevant imagination. For a machine to pass this test it would need a network of connected memories, memories that lead to different conclusions depending on how they’re approached. Such a machine could be built, but it hasn’t been yet.
We depend on people being unreliable. People who are too reliable are boring. Because most of us are boring, and many of our relationships are reliable in this manner, we shed relationships over time. Enduring relationships are those that continue to answer new questions.
If you had to prove you were not a machine, then you’d have to show you could be emotional and enthusiastic. You would have to personalize your thinking.
If you’re accused of being robotic, ask yourself whether you’re emotional and enthusiastic. To what extent are you aware of and responsive to unpredictably?
Our Most Important Questions are Unanswerable
I’d like to mention Turing’s idea about unanswerable questions. It’s hard to translate this into human terms, but it’s useful to try.
Turing first proposed a Turing Machine. This was simply a tape with a series of zeros and ones that was “read” by a computer. The computer responded to this tape by either moving the tape forward, backward, or stopping. The only important thing about the Turing Machine was how long it took before it stopped moving the tape.
Next, Turing asserted that all self-consistent questions that could be asked could be digitized and answered by such a suitably programmed Turing machine. That is, all statements and their answers can be translated into binary form and solved using some algorithm. When the algorithm had “solved” the problem, the machine would stop moving the tape.
Turing claimed that the Turing Machine was entirely general. As long as the questions were self-consistent, some algorithm could answer it, and the machine would stop. We don’t need to understand the question or the answer.
Then, Turing, Alonzo Church, and Stephan Kleene showed there exists certain self-consistent questions along with algorithms that can answer them. We know that definitive and correct answers exist for these questions.
On the other hand, they were also able to show that, in the process of answering these questions, the Turing machine will never stop. In fact, the machine will never get close to stopping.
This means there are well-formed questions with certain answers we can never prove. Think about this for a moment.
It means there are certain truths that we can never be certain of. There are answers we know are true, but we can never prove. For most of us, this is irrelevant. But, if you wonder about the limits of your understanding, this might explain why we often feel like we don’t know much!
As I understand it, the trick lies in the nature of a well-formed question. We know there are unanswerable questions, such as “Is there a God?” These are easily recognized as poorly-formed questions because they pertain to things that are not defined.
In addition, we have circular questions, where one part of the question depends upon the other. The question, “Which came first, the chicken or the egg?” is circular.
Humans frequently engage in circular logic and use undefined terms. For example, many assume that if we continually improve, we’ll solve the big problems. Or if we resolve our differences, we’ll live in peace. It’s more correct to say that without our differences, we would have no relationships at all since, in that case, we’d all be identical.
With regard to human behavior, something continues to be missing from our thinking.
References
Dyson, George (2012). Turing’s Cathedral; The Origins of the Digital Universe. Vintage.
Kandola, Aaron (2023). “What is NLP and What is It Used For?” Medical News Today. https://www.medicalnewstoday.com/articles/320368
Enter your email for a FREE 1x/month or a paid 4x/month subscription.
Click the Stream of the Subconscious button.