History is wealthy with examples of individuals attempting to breathe life into inanimate objects, and of individuals selling hacks and tricks as “magic.” But this very human desire to imagine in consciousness in machines has never matched up with reality.
Creating consciousness in artificial intelligence systems is the dream of many technologists. Large language models are the most recent example of our quest for clever machines, and a few people (contentiously) claim to have seen glimmers of consciousness in conversations with them. The purpose is: machine consciousness is a hotly debated topic. Loads of experts say it’s doomed to stay science fiction eternally, but others argue it’s right across the corner.
For the most recent edition of MIT Technology Review, neuroscientist Grace Huckins explores what consciousness research in humans can teach us about AI, and the moral problems that AI consciousness would raise. Read more here.
We don’t fully understand human consciousness, but neuroscientists do have some clues about the way it’s manifested within the brain, Grace writes. To state the apparent, AI systems don’t have brains, so it’s inconceivable to make use of traditional methods of measuring brain activity for signs of life. But neuroscientists have various different theories about what consciousness in AI systems might appear like. Some treat it as a feature of the brain’s “software,” while others tie it more squarely to physical hardware.
There have even been attempts to create tests for AI consciousness. Susan Schneider, director of the Center for the Future Mind at Florida Atlantic University, and Princeton physicist Edwin Turner have developed one, which requires an AI agent to be isolated from any details about consciousness it could’ve picked up during its training before it’s tested. This step is significant in order that it may well’t just parrot human statements it’s picked up about consciousness during training, as a big language model would.
The tester then asks the AI questions it should only have the ability to reply whether it is itself conscious. Can it understand the plot of the movie Freaky Friday, where a mother and daughter switch bodies, their consciousnesses dissociated from their physical selves? Can it grasp the concept of dreaming—and even report dreaming itself? Can it conceive of reincarnation or an afterlife?
In fact, this test will not be foolproof. It requires its subject to have the ability to make use of language, so babies and animals—manifestly conscious beings—wouldn’t pass the test. And language-based AI models can have been exposed to the concept of consciousness within the vast amount of web data they’ve been trained on.
So how will we really know if an AI system is conscious? A gaggle of neuroscientists, philosophers, and AI researchers, including Turing Prize winner Yoshua Bengio, have put out a white paper that proposes practical ways to detect AI consciousness based on a wide range of theories from different fields. They propose a form of report card for various markers, resembling flexibly pursuing goals and interacting with an external environment, that may indicate AI consciousness—if the theories hold true. None of today’s systems tick any boxes, and it’s unclear in the event that they ever will.
Here’s what we do know. Large language models are extremely good at predicting what the subsequent word in a sentence ought to be. Also they are superb at making connections between things—sometimes in ways in which surprise us and make it easy to imagine within the illusion that these computer programs might need sparks of something else. But we all know remarkably little about AI language models’ inner workings. Until we all know more about exactly how and why these systems come to the conclusions they do, it’s hard to say that the models’ outcomes usually are not just fancy math.
