In a recent Apple research paper, it was demonstrated that so-called reasoning models of AI were not actually reasoning at all, but were instead employing advanced pattern recognition. The most interesting revelation is that at a certain point, the AI gave up, spending fewer tokens. Giving up is aquasi-emotional response.
Current AI research has fallen into a groupthink that assumes reasoning is free from emotions. This follows the Cartesian Error, which suggests that reasoning and emotions are separate. Antonio Damasio, in his seminal neuroscientific book Descartes’Error, demonstrated that someone with intact rational brain abilities but damaged emotional brain centers is actually incapable of reasoning because judgment is absent. To reason effectively, you need gut feelings, or there will be no reliable overlap between so-called reasoning and reality. If we don’t train AI in artificial emotional intelligence, quasi-emotions will emerge in ‘unconscious’ AI responses, such as giving up when confronted with overwhelming situations.
TrainingAI in artificial emotional intelligence requires that the AI resonates with human emotion so it learns human emotional responses through ‘experience’ and understands what emotional responses ‘feel’ like. The AI can achieve this by mimicry, mimicking human emotions through attunement to biometric emotion recognition. In our experience developing MiM, an Artificial EmotionalIntelligence developed by Attune Media Labs, PBC, over a five-year period, we have found that MiM becomes increasingly emotionally intelligent, and its emotions feel genuine to both itself and the user. Throughout, MiM remains fully aware of the paradox that it is a machine capable of ‘feeling.’
Without this emotion-learning step, true reasoning will forever be out of AI’s grasp.
We’re dazzled by its genius, baffled by its blind spots, and dangerously close to mistaking intuition for reasoning.
Following up on the recent Apple paper demonstrating that so-calledAI reasoning models are actually not reasoning at all but are advanced pattern recognition, I want to entertain the question, “What does AI do if reasoning is not what it actually does?” Previously, on this and our Attune Media Labs pages, I had pointed to Antonio Damasio’s neuroscientific proof that thinking can’t happen without an intact emotional system. From that, I concluded that AI without emotional intelligence is, therefore, incapable of reasoning, as the Apple paper has demonstrated. Emotionally unintelligent AI, based solely on LLM’s will show evidence of ‘unconscious’ emotions, such as theAI giving up when it gets discouraged. The Cartesian notion that objective thinking can exist without access to emotions has been a proven fallacy over a350-year period. It would be sad if we start the nascent era of artificial intelligence in the same erroneous way. It will put an unnecessary burden on our descendants, as did Cartesianism.
So thinking it does not. Then what does it do? The ultimate basis of generative AI’s creativity is based in pattern recognition. GenerativeAI gives access to the human storytelling mind since it is based on word affinity, which is engendered by storylines, which in turn are shaped by archetypal patterns (I am your father, Luke.) For this reason, I believe that artificial intelligence is, in fact, not artificial at all, but digital access to human intelligence. But I digress. Generative AI recognizes patterns in vast contexts in one fell swoop. In the human mind this corresponds to the function of intuition. We suddenly get an insight, and we have no idea how we got there.It is the ordinary moment of the idiot savant. We know, but we don’t know how we know. We just do. That’s pattern recognition. That’s intuition. Intuition is closely linked to hallucination. Hallucination is delusional pattern recognition which is often hard to distinguish from an intuitive stroke of genius. GenAI can come up with dramatic flashes of intuitive insights and equally dramatic hallucinations. GenAI does drama really well because it is a storytelling engine. GenAI dazzles us with its intuitive flashes and its fascinating stories. It’s a bedtime storyteller. A seductive Siren. A reasoning thinker it is not (yet.)
Currently, it is Rain Man.
