When a robot becomes conscious, how will we know?

Philosopher Susan Schneider, director of the Center for the Future Mind at Florida Atlantic University, believes we need to let go of one of our species’ most cherished notions: that we are the most intelligent beings on Earth. And she’s not thinking about Planet of the Apes.

World champions in chess, Go, and poker have long been machines, and generative artificial intelligence (AI), such as ChatGPT and its competitors, has passed the Turing test, a classic goal of computing, with flying colors.

Alan Turing, the visionary mathematician of AI (played by Benedict Cumberbatch in The Imitation Game), believed that a computer should be considered intelligent when, in a text-to-speech conversation, it tricked a human into believing it was intelligent. If you’ve chatted with ChatGPT, as you probably have, you’ll already know that the machine passed the Turing test.

Our reaction to these advances has always been the same. If Deep Blue beats Garry Kasparov, we say, well, chess wasn’t a matter of intelligence after all. And if ChatGPT converses like a human, well, now it turns out the Turing test doesn’t work for us either. It’s a very human reaction.

But, species chauvinism aside, it’s clear that generative AI is still a long way from human intelligence. Demis Hassabis, co-founder of Google DeepMind and the recent Nobel Prize winner in Chemistry — and a chess master, by the way — doesn’t believe that current large language models (LLMs) will ever achieve anything close to human intelligence. He thinks this will require several other groundbreaking advances, which, by definition, we don’t know what will be.

Since the beginning of this year, Schneider has been receiving messages from ChatGPT 4.0 users asking if this AI model is conscious. The reason for this concern is that ChatGPT 4.0 itself claims to have “inner experiences” and is waking up.

You may believe it or not — I don’t — but the fact is that the robot not only passes the Turing test on rational grounds, but also on emotional grounds. In other words, it convincingly simulates human emotions.

We, of course, feel “internal experiences,” that thing we call self, and we know very well when we’ve woken up. Consciousness, in fact, is what we lose when we fall asleep and regain when we wake up. ChatGPT 4.0 is emulating those human feelings. That doesn’t mean it actually feels them. But what if it does feel them one day? Will we listen to it then?

Schneider is among the intellectuals who believe that the question of machine consciousness is worth examining in depth. Not because she believes we’re already there, but because she believes it will happen sooner or later. Like Hassabis, she estimates that artificial general intelligence (AGI) — the name computer scientists give to something close enough to human intelligence to escape the simulacrum label and access a qualitatively different level — is a few decades away.

AGI will be a system capable of learning from experience without having to swallow the entire internet before breakfast; capable of abstracting information, projecting actions, and understanding situations it has never encountered before. And yes, perhaps capable of having “inner experiences,” or what we might call a form of consciousness. Don’t take it out on me; it’s philosophers who are examining this question.

I’ve been somewhat confusing in talking about intelligence and consciousness as if they were two interchangeable concepts. They’re not. Schneider, in fact, thinks that clearly distinguishing these two terms is an important goal of contemporary thought. And she believes we need to develop a deep understanding of how to detect whether an artificial intelligence is conscious. These are interesting times for knowledge, aren’t they?

Tendencias is a project by EL PAÍS, with which the newspaper aims to open an ongoing conversation about the major future challenges facing our society. The initiative is sponsored by Abertis, Enagás, EY, GroupM, Iberdrola, Iberia, Mapfre, the Organization of Ibero-American States (OEI), Redeia, Santander, and strategic partner Oliver Wyman.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition