We have a framework for analyzing whether advanced AI is conscious or not
We (Jonathan Schooler, Marissa Ericson and myself) wrote a paper, published in the peer-reviewed journal Perspectives in Psychological Science, in early 2022, that provides a comprehensive framework for examining whether a given object is consciousness or not.
It’s called “Where’s My Consciousness-Ometer? How to Test for the Presence and Complexity of Consciousness.”
We look to various categories of “measurable correlates of consciousness,” including neural correlates, behavioral correlates, and creative correlates.
Here’s the abstract:
Tools and tests for measuring the presence and complexity of consciousness are becoming available, but there is no established theoretical approach for what these tools are measuring. This article examines several categories of tests for making reasonable inferences about the presence and complexity of consciousness (defined as the capacity for phenomenal/subjective experience) and also suggests ways in which different theories of consciousness may be empirically distinguished. We label the various ways to measure consciousness the measurable correlates of consciousness (MCC) and include three subcategories in our taxonomy: (a) neural correlates of consciousness, (b) behavioral correlates of consciousness, and © creative correlates of consciousness. Finally, we reflect on how broader philosophical views about the nature of consciousness, such as materialism and panpsychism, may also be informed by the scientific process.
Today’s advanced AIs, like OpenAI’s GPT4.0, are of course not based on biological neural structures, but they are based on “artificial neural networks.” Their behavior in some contexts is indistinguishable from human behavior, in terms of their verbal output. And their creative outputs, including words and the beautiful art that they are now producing through tools like DALL-E 2 and Midjourney, also suggest consciousness may be present.
My intuition remains that these current advanced AIs are very unlikely to be conscious in any way like our own, but I remain open to new facts and am in particular open to the idea that perhaps these large ANNs are physically based on EM field resonances that achieve various kind of standing waves that comprise collectively their own species of consciousness, in a way that is similar to but different to the collection of standing EM field waves that comprise human and other mammalian consciousness.
Current debates about AI safety don’t necessarily depend on whether any given AI is conscious or not. But it certainly could be a factor in terms of agency and desire to break through possible guardrails or rulesets designed to prevent harm to humans.
