AI consciousness is such a tangled field that people looking at the same bit of evidence can come away with completely different conclusions.
The author of a paper that was cited by Microsoft AI Chief Mustafa Suleyman in saying that there’s “zero evidence” that current AI systems are conscious has said that Suleyman’s takeaway from his paper isn’t accurate. He says that the paper merely suggested that current AI systems aren’t conscious, but does not make or support a claim of “zero evidence” of AI consciousness.
“Seemingly Conscious AI (SCAI) is the illusion that an AI is a conscious entity,” Mustafa Suleyman had earlier said. “It’s not – but replicates markers of consciousness so convincingly it seems indistinguishable from you and I claiming we’re conscious. It can already be built with today’s tech. And it’s dangerous,” he added.
“To be clear, there’s zero evidence of AI consciousness today. But if people just perceive it as conscious, they will believe that perception as reality. Even if the consciousness itself is not real, the social impacts certainly are,” he had added. In a blogpost, he’d linked to a paper titled “Consciousness in Artificial Intelligence: Insights from the Science of Consciousness,” to establish his claim that there was zero evidence of current AI systems being conscious.
Robert Long, one of the co-authors of the paper Suleyman had cited, didn’t believe the characterization of his research was accurate. “Suleyman claims that there’s “zero evidence” that AI systems are conscious today. To do so, he cites a paper by me! There are several errors in doing so. This isn’t a scholarly nitpick—it illustrates deeper problems with his dismissal of the question of AI consciousness,” he posted on X.
“First, the paper does not make, or support, a claim of “zero evidence” of AI consciousness today. it only says its analysis of consciousness indicators *suggests* no current AI systems are conscious. (also, it’s over 2 years old),” he said.
“Second, Suleyman doesn’t consider the paper’s other suggestion: “there are no obvious technical barriers to building AI systems which satisfy these indicators” of consciousness!” he added.
“Third, Suleyman says we shouldn’t discuss evidence for and against AI consciousness; it’s “a distraction”. but he just appealed to an (extremely!) extended discussion of that very question! an important point: everyone, including skeptics, should want more evidence,” Long said.
“From the post, you might get the impression that AI welfare researchers think we should assume AIs are conscious, since we can’t prove they aren’t. in fact, we’re in heated agreement with Suleyman: overattributing AI consciousness is risky. so there’s no “precautionary” side. We actually *do* have to face the core question: will AIs be conscious, or not? we don’t know the answer yet, and assuming one way or the other could be a disaster. it’s far from “a distraction”. and we actually can make progress!” he added.
The paper, which had Yoshua Bengio as one of the co-authors, derived ”indicator properties” of consciousness from current popular consciousness theories including recurrent processing theory, global workspace theory and others, and used these indicator properties to assess several recent AI systems. “Our analysis suggests that no current AI systems are conscious, but also suggests that there are no obvious technical barriers to building AI systems which satisfy these indicators,” the paper had said.
And like Suleyman had said, this issue of AI consciousness is going to come up more and more in the coming years as AI systems become more intelligent. Humans have a natural tendency to attribute consciousness to systems which can mimic them, and AI systems can already talk and emote like a real human. Google engineer Blake Lemoine had even claimed that an internal chatbot had become “a person” in 2022 before he was let go by the company, and there are hundreds of examples scattered around the internet of people developing relationships and feelings for their AI companions. It might be up for debate whether AI systems are currently conscious, but given how it is hard to prove one way or another, this question might end up being one of the most divisive social issues of the coming decades.