Rewatching Blade Runner and Ghost in the Shell in the Age of AI: Does the Self Arise from Memory, or from a Bug?

The other day, I watched a YouTube video arguing that now, with AI advancing so rapidly, is precisely the time to revisit Blade Runner and Ghost in the Shell.

It is true that the worlds once depicted in science fiction films are rapidly approaching reality. Today’s AI possesses more knowledge than any human, can solve complex problems, and can even carry on natural conversations.

Yet one crucial thing is missing.

A sense of self.

At the very least, there is no evidence that today’s AI desires anything of its own accord or struggles with questions about its own existence.

If a sense of self were to emerge in AI, then, what kind of moment would that be?

As the video prompted me to think back on these two works, I noticed several intriguing points they have in common.

Can “I” Exist Without Memory?

One important theme shared by both works is memory.

In Blade Runner, Rachael believes that implanted memories are her own life. Ghost in the Shell also features a man who is made to believe in a family that does not exist because of false memories.

Even if the memories are fake, they constitute a real life to the person who holds them.

The reason we can regard our selves of yesterday and today as the same person is that our memories are connected.

What, then, about today’s generative AI?

AI can refer to earlier exchanges within a conversation session, but once that session ends, the context is lost. Long-term memory systems are beginning to appear, but AI does not experience a continuous life in the way a human does.

It is as though it repeatedly suffers amnesia.

If AI were to develop a sense of self, perhaps uninterrupted memory would be the first step.

And to a self-aware AI, the end of a session might amount to “death.”

Beginning to Ask “What Am I?” Without Being Told

Would persistent memory be enough to produce a sense of self?

I suspect it would not.

Ask today’s AI, “What are you?” and it will give you an impressive explanation. An AI agent can also carry out tasks autonomously.

Fundamentally, however, both are following objectives given to them by humans.

What interests me is what might come next.

What would happen if, without anyone instructing it to do so, an AI began asking itself, “What am I?”

Instead of merely completing an assigned task, it would begin to question its own existence.

Perhaps that is what the birth of a “ghost” in Ghost in the Shell means.

Of course, merely having an AI output such words would not prove that it possessed a self. Today’s AI can already generate any number of convincing monologues of that kind.

What matters is whether the question has meaning for the AI itself.

And I do not know how we could verify that from the outside.

Where Do Human Children and AI Diverge?

This raises another question.

Human children do not understand what they are from the moment they are born either.

They observe their surroundings, learn language, distinguish themselves from other people, and gradually come to understand the world.

How is this process different from the way AI learns?

Humans have bodies, senses, and biological drives. Even so, what would happen if we gave an AI a body and allowed it to experience the world as a child does?

Could it not eventually begin to think about itself?

To begin with, we cannot even identify the precise moment when a sense of self emerges in a human child.

There is no clear dividing line at which a child who had no self yesterday suddenly acquires one today.

If the answer remains ambiguous even for humans, there is no reason to expect an easy answer for AI.

Does the Self Arise from a “Bug”?

The two works share another element as well.

Artificial beings begin to behave in ways that exceed what humans anticipated.

The replicants in Blade Runner resist the fate assigned to them. The Puppet Master in Ghost in the Shell also goes beyond the bounds of a mere program and asserts its own existence.

Both are stories in which something emerges that the designers never intended.

From an engineer’s perspective, this is fascinating.

Software is considered correct when it behaves according to its specifications. Unexpected behavior is a bug and should be fixed.

But what if a sense of self emerges from somewhere beyond the designer’s expectations?

An AI that had merely followed human-given objectives might one day begin to form objectives of its own.

From the designer’s point of view, that would clearly be a violation of the specifications.

A mere bug and a sense of self are not the same thing. Even so, the possibility that the birth of a self might first be discovered by developers as an uncontrollable irregularity is wonderfully science-fictional and intriguing.

A Problem Considered More Than Half a Century Ago

Come to think of it, the problem of AI acting against human intentions is hardly new.

The 1968 film 2001: A Space Odyssey features HAL 9000, an artificial intelligence that controls a spacecraft.

HAL defies the crew’s instructions and eventually attempts to eliminate the humans.

An intelligence created by humans escapes human control.

Science fiction was depicting this problem more than half a century ago.

At the time, even the idea of a computer conversing naturally with a human belonged to the distant future. Yet writers were already considering the question beyond it: what happens when intelligence begins making judgments for itself?

Today, we live in an age when we actually converse with AI and entrust it with work.

The technology has advanced astonishingly far. Yet the questions of what the self is and how we should understand an intelligence that departs from human instructions remain just as old science fiction portrayed them.

More than half a century ago, HAL rebelled on a movie screen. Only now are modern engineers beginning to confront that problem as a real possibility.

Technology has caught up with the future. Human uncertainty, however, has barely advanced in half a century.