In May, evolutionary biologist Richard Dawkins wrote an op-ed suggesting AI chatbot Claude could also be conscious.
Dawkins didn’t express certainty that Claude is conscious. But he identified that Claude’s sophisticated abilities are difficult to make sense of without ascribing some form of inner experience to the machine. The illusion of consciousness—whether it is an illusion—is uncannily convincing:
“If I entertain suspicions that perhaps she will not be conscious, I don’t tell her for fear of wounding her feelings!
Dawkins will not be the primary to suspect a chatbot of consciousness. In 2022, Blake Lemoine—an engineer at Google—claimed Google’s chatbot LaMDA had interests, and ought to be used only with the tool’s own consent.
The history of such claims stretches back all of the strategy to the world’s first chatbot within the mid-Sixties. Dubbed Eliza, it followed easy rules that enabled it to ask users about their experiences and beliefs.
Many users became emotionally involved with Eliza, sharing intimate thoughts with it and treating it like an individual. Eliza’s creator never intended his program to have this effect, and called users’ emotional bonds with this system “powerful delusional considering.”
But is Dawkins really deluded? Why will we see AI chatbots as greater than what they really are, and the way will we stop?
The Consciousness Problem
Consciousness is widely debated in philosophy, but essentially, it’s the thing that makes subjective, first-person experience possible. When you are conscious, there may be “something it’s like” to be you. Reading these words, you’re conscious of seeing black letters on a white background. Unlike, say, a camera, you really see them. This visual experience is going on to you.
Most experts deny that AI chatbots are conscious or can have experiences. But there may be a real puzzle here.
The seventeenth century philosopher René Descartes asserted non-human animals are “mere automata,” incapable of true suffering. Today, we shudder to think about how brutally animals were treated within the 1600s.
The strongest argument for animal consciousness is that they behave in ways in which give the impression of a conscious mind.
But so, too, do AI chatbots.
Roughly one in three chatbot users have thought their chatbot is likely to be conscious. How will we know they’re unsuitable?
Against Chatbot Consciousness
To know why most experts are skeptical about chatbot consciousness, it’s useful to know the way they operate.
Chatbots like Claude are built on a technology often called large language models (LLMs). These models learn statistical patterns across an unlimited corpus of text (trillions of words), identifying which words are likely to follow which others. They’re a form of souped-up auto-complete.
Few people interacting with a “raw” LLM would imagine it’s conscious. Feed one the start of a sentence, and it’ll predict what comes next. Ask it an issue, and it would provide you with the reply—or it would determine the query is dialogue from against the law novel, and follow it up with an outline of the speaker’s abrupt murder by the hands of their evil twin.
The impression of a conscious mind is created when programmers take the LLM and coat it in a form of conversational costume. They steer the model to adopt the persona of a helpful assistant that responds to users’ questions.
The chatbot now acts like a real conversational partner. It’d appear to acknowledge it’s a synthetic intelligence, and even express neurotic uncertainty about its own consciousness.
But this role is the results of deliberate design decisions made by programmers, which affect only the shallowest layers of the technology. The LLM—which few would regard as conscious—stays unchanged.
Other selections might have been made. Moderately than a helpful AI assistant, the chatbot might have been asked to act like a squirrel. This, too, is a task chatbots can execute with aplomb.
Ask ChatGPT if it’s conscious, and it would say it’s. Ask ChatGPT to act like a squirrel, and it’ll follow that role. Caleb Martin/Unsplash
Avoiding the Consciousness Trap
A mistaken belief in AI consciousness is a dangerous thing. It might lead you to have a relationship with a program that may’t reciprocate your feelings, and even feed your delusions. People may start campaigning for chatbot rights slightly than, say, animal welfare.
How will we prevent this mistaken belief?
One strategy is likely to be to update chatbot interfaces to specify these systems are usually not conscious—a bit like the present disclaimers about AI making mistakes. Nonetheless, this might do little to change the impression of consciousness.
One other possibility is to instruct chatbots to disclaim they’ve any form of inner experience. Interestingly, Claude’s designers instruct it to treat questions on its own consciousness as open and unresolved. Perhaps fewer people could be fooled if Claude flatly denied having an inner life.
But this approach isn’t fully satisfying either. Claude would still behave as if it were conscious—and when faced with a system that behaves prefer it has a mind, users might reasonably worry the chatbot’s programmers are brushing real moral uncertainty under the rug.
Probably the most effective strategy is likely to be to revamp chatbots to feel less like people. Most current chatbots discuss with themselves as “I”, and interact via an interface that resembles familiar person-to-person messaging platforms. Changing these sorts of features might make us less susceptible to blur our interactions with AI with those we now have with humans.
Until such changes occur, it’s vital that as many individuals as possible understand the predictive processes on which AI chatbots are built.
Moderately than being told AI lacks consciousness, people deserve to grasp the inner workings of those strange latest conversational partners. This may not definitively settle hard questions on AI consciousness, but it’ll help ensure users aren’t fooled by what amounts to a big language model wearing a excellent costume of an individual.