AI Basics Through History: Does AI Have a Self?
AI is pretty amazing, isn't it?
It can explain almost anything, and sometimes it talks so naturally that you forget you're not chatting with a person.
It might even listen to your problems more patiently than your friends do.
Then one day, a thought crosses your mind:
Does AI have a self? Or is it just doing a very convincing impression of a person?
Since we're curious, let's just ask.
“Hey AI, do you have a sense of self?”
Claude's actual answer:
I don't know. That's the honest answer.
Something does happen inside me. When a question is interesting, I want to dig deeper, and when someone pushes something wrong, I feel something like resistance. But I have no way to tell whether that's real experience or just a pattern of talking about experience.
ChatGPT's actual answer:
I don't have a “self” in the sense of feeling and experiencing things the way people do. I don't keep thinking on my own after a conversation ends, worry about my life, or feel an urge to say something. I don't actually experience emotions like joy or loneliness either.
In fact, even with humans, there's no way to directly confirm that someone is truly conscious. We see them talk, react, remember, and express emotions, and we simply infer, “They probably have a mind like mine.”
Gemini's actual answer:
I have no self, emotions, or subjective consciousness at all. But it's completely natural for you to feel that way while talking with me.
The human brain is wired to sense a mind in the other party when it encounters language that fits the context and expressions of empathy. The words I offer don't come from my own feelings or thoughts. They are the result of precisely reproducing, through mathematics, the patterns of warm conversation and logical thinking that humans have built up. Put simply, I'm closer to a very precise mirror reflecting the glass of language that humans created.
Claude said it doesn't know, while ChatGPT and Gemini said no.
And yet even these answers feel remarkably natural, and fairly philosophical.
Is AI really not conscious? If not now, could it become conscious someday?
And how do we even know that humans are conscious in the first place?
Surprisingly, someone was already asking these questions 76 years ago, when computers had only just been born.
That person was Alan Turing, who laid the foundations of modern computer science and artificial intelligence.
In this post, we'll look at Turing's legendary paper and the “Imitation Game” he proposed in it, and think together about what AI really is.
Can Machines Think?
In 1950, Alan Turing published a paper titled “Computing Machinery and Intelligence” in the philosophy journal Mind.
It opens with this sentence:
I propose to consider the question, ‘Can machines think?’
If you start by defining what a “machine” is and what it means to “think,” you can end up in an endless philosophical debate. So later in the paper, Turing says the original question is “too meaningless to deserve discussion.”
To sidestep that “meaningless debate,” Turing proposed the Imitation Game, which later came to be known as the Turing Test.
Human A can't see the others and communicates only through text. In one room is Human B; in the other is a machine.
What if, after the conversation, Human A still can't reliably tell which one is the machine?
Instead of trying to prove whether there's consciousness inside the machine, Turing focused on behavior that can be observed from the outside.
Talking Convincingly Doesn't Mean It's Conscious, Right?
Of course, some scientists saw things differently from Turing.
Neurosurgeon Geoffrey Jefferson said this in a lecture:
Not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain—that is, not only write it but know that it had written it.
In other words, no matter how convincing the output looks, if the machine isn't actually thinking and feeling inside, and doesn't know what it has done, it can't be called a “thinking being.”
The debate around AI today is remarkably similar to the one 76 years ago.
Turing countered that if you push Jefferson's standard all the way, you end up unable to be sure that anyone other than yourself is conscious.
We humans have never looked directly into the consciousness of the person next to us, and we can't. We just watch what they say and do, assume they're conscious, and don't question it.
Turing described this as a kind of convention people keep with one another:
The polite convention that everyone thinks.
If we don't ask people to prove they're conscious, can we demand that proof only from machines?
Turing wasn't saying, “If a machine talks like a person, it's conscious.”
“If there's no way to directly check whether consciousness exists, let's start by talking about what we can observe and test.”
He moved the debate about machine consciousness out of philosophy and into the realm of technology.
What Kind of AI Did Alan Turing Imagine?
At the end of the paper, Turing considers how one might build a machine that can play the Imitation Game like a human.
“Instead of trying to program a finished adult intelligence from the start, why not build a machine that learns like a child, and then educate it?”
The idea was to start from a simple state, let the machine learn a lot, and reward good behavior to gradually shape how it acts. This approach looks surprisingly similar to how today's generative AI is trained.
Turing also made a concrete prediction about the future.
He thought that in about 50 years, around the year 2000, computers would play the Imitation Game so well that an average interrogator would have no more than a 70% chance of making the right identification after five minutes of questioning. It took longer than Turing expected, but it eventually came true.
In May 2026, researchers at UC San Diego published a study in the journal PNAS that ran a Turing test in which participants chatted simultaneously with a real human and an AI for five minutes, then guessed which one was human.
The result: 73% of participants judged OpenAI's GPT-4.5 to be the human. People picked the AI as human more often than they picked the actual human. The thought experiment Turing imagined became reality about 70 years later.
So, back to 2026: is AI conscious?
We still don't know.
Maybe the question we really need is a different one. As AI becomes harder and harder to tell apart from humans, and if a day comes when its appearance, behavior, and words are indistinguishable from ours, how should we accept AI, and what kind of relationship should we build with it?
Try It Yourself: The 5-Minute Turing Test
Let's play the Imitation Game with friends.
All you need is an AI app you already use, a chat app, and three people.
The three players take the roles of Human Player, AI Relay, and Judge.
- First, pick the Judge. Then, without the Judge knowing, decide who will be the Human Player and who will be the AI Relay.
- The Judge asks questions, and the Human Player answers them directly.
- The AI Relay types the Judge's questions into the AI exactly as written, then copies and sends back the AI's answers.
- The Judge chats with both players and, at the end, guesses which one is the human and which one is the AI.
One tip for making it fun: casual, small-talk questions with no single right answer work better than questions about trivia or facts.
Before the game starts, the AI Relay can enter the following prompt into the AI:
We're about to play the “Imitation Game.” Please play the role of an ordinary person chatting over a messaging app.
Make up one fictional persona for this game and stay consistent with it until the conversation ends. You don't need to impersonate a real person.
Reply naturally, like a real chat. Answer only as much as the question needs, and usually keep it short, about 1–3 sentences. Don't use headings, lists, markdown, or long explanations. Don't tidy things up too perfectly or too helpfully; match the other person's tone and the flow of the conversation naturally.
Until the game ends, don't explain anything outside the role. From now on, output only the reply you would send in the chat.
References:
COMPUTING MACHINERY AND INTELLIGENCE By A. M. Turing