logo logo

Turing Test and Chinese Room

Why both are flawed

Turing Test and Chinese Room

Why both are flawed

  • Turing Test and Chinese Room together is a recipe for mind boggling nonsense
  • If a Dog obeys your “Sit” command, does it understand English
  • Could a Artificial Intelligence understand Dogs

At its most basic, the Turing Test is a man, a woman and a judge in three separate rooms, with both the man and the woman trying to convince the Judge they are who they say they are. Extrapolated to Computers, the Turing Test replaces the man and the woman with a human and a computer, both trying to convince the judge, they are in fact, human.

Both are Human

Two humans in closed rooms side by side. How does one convince the other it is in fact human.

Assuming there is no other interaction except by voice, both parties would have a question and answer session. One might tell a story of his childhood, his past loves, his hopes and dreams. The other would counter with his own.

If both are actually human, with one party speaking English and the other Swahili, can they still confirm only by voice if either is human? Neither is able to understand the other without a translator.

Let’s move it up a bit and allow them to have both a voice interaction and a written session where both parties write in their own language on pieces of paper and pass them through the door.

Again, if both are actually human, they might try to distinguish the actual act of passing the paper under the door, to confirm if the other party’s physical action of passing the paper matches their own, or close to it. Would that be any confirmation? A shred of lingering doubt, but confirmation? No.

One person has written in Swahili, the other in English, neither can understand the other language. To either, both are simply squiggly characters. What is any of them to decipher here?

Let’s move it up a bit. The Judge now takes it upon him/herself to transcribe and translate the writings into the other party’s known language. The Swahili speaking person gets the responses from the English speaking person in Swahili. The English speaker gets his Swahili responses in English.

Now they can both understand each other. In Writing.

Both would think the other is a human and can read and write in both languages but speak only their own native tongue. Turing Test is passed. But the Chinese Room stays.

Both are Computers

I lost my train of thought here. Or perhaps i just didn’t feel like typing. Probably got bored. This was years ago. Don’t remember much after the multiple /dev/zero rewipes.

Will write this section when and if i rezone. Or feel like it.

Why is the Chinese Room such a difficult argument to break?

At its most basic, it goes to the root of understanding. How the process of learning actually happens.

When i grew up, going to an English school, learning English, Hindi and Punjabi with English being my main language for learning, and Hindi and Punjabi being secondary, to this date i cannot fathom if i think in English or Punjabi. Each time i think, and i did it now, when writing this line, I cannot answer with certainty if the words being written here were in its base language, english, or did my brain translate them from my secondary language Hindi or Punjabi. I really do not know.

What i do know is that after years of not writing in Hindi or Punjabi, i have actually forgoten how to write in it. But my speaking skills remain. They are constantly used. Will i be able to relearn how to write fluently? Yes.

And therein lies the rub.

What language does a computer think in? 1 0. This is guaranteed. It doesn’t matter if you feed it english, chinese, german, french, each and every single instruction (The Instruction bit) is still broken down into a 1 and a 0 for input, with the output again being 0 and 1. The translation layer, call it Operating System, is the one actually doing the translation for you.

In today’s world, each time you use a computer, your mobile device, make a phone call, read a text, watch a movie, you are actually in the Chinese Room.

A Fiber Optic Cable carries TB/s data. Is the Fiber Optic cable capable of knowing what the data means? Does the switch know? Does the Core Router?

Philosophically, i cannot think of any cogent argument against the Chinese Room. It covers all responses.

A Man and his Dog

A man sitting with a dog and a cat carries on a conversation with both in English, or French or Chinese. Does the dog understand the language? “Sit”. Did it understand english? Or was it trained using the pavlovian mechanism.

A dog is a biological being, it has a biological brain. Not artificial. Can it be said with certainty the Dog can understand a human and the Human’s language? Let’s do the Chinese Room experiment with the Dog and the Human, by translating symbols into dog speak.

Two woofs for a yes, One for a No.

Are you a Human? I would find it difficult to even ask this question. Dogspeak is more difficult than 1 0. Woof.

Lot of theories have been against the Chinese Room. But in the end the Chinese Room is not a technology question, it is not even a question of logic, rather a philosophy question.

If i am Intelligent and carrying on a conversation with you, Am i assuming you are also intelligent? For people reading this, if this was written in Chinese, would it still have been written by a human and translated by a computer? Or written by a computer by culling existing text from major publications on this topic.

How can you tell by reading this that i am human.

How is Human Intelligence still different from a Computer

Imagine, it is 2 am in the morning and you are withdrawing money from an ATM on a dark, scary street. Money in pocket, you start to walk down the road when you get mugged. The perpetrator, high on drugs, gives you an option, run face first against that wall or i will shoot you.

What do you do? And the minute you answer this, you will know how it is possible for humans to create rules on the go, and how computers cannot.