by Antonio Damasio
Pantheon Books
News media are gripped by AI panic, the fear that AI agents will go rogue and inflict indescribable horrors upon mankind. As AI policy becomes increasingly politicized, the punditocracy seems to be in no mood to air issues calmly and deliberately. Voices of reason can be found, but most are relegated to the realm of books… which means their thinking is inaccessible to the vast majority of Americans who imbibe their information from the likes of TikTok and CNN. Let’s face it, fear sells.
One voice of non-politicized sanity is Antonio Damasio, a professor of Neuroscience, Psychology and Philosophy at the University of Southern California. His long-abiding interest has been the phenomenon of consciousness — more specifically, consciousness as it exists in the “natural” world of biological creatures, not in the realm of artificial intelligence. In his most recent book, “Natural Intelligence & the Logic of Consciousness,” he adds critical context to the AI debate by arguing that consciousness — or sentience, or self-aware intelligence — is not a purely cognitive phenomenon, as many treat it, but is an intrinsically biological phenomenon.
The implication of Damasio’s work, as I interpret it, is that Artificial Intelligence, even human-like General Intelligence, will never reach self-aware consciousness. AI eventually will become supremely adept at mimicking consciousness but will not — cannot — achieve it. AI agents may “go rogue” because of the failure of human designers to build appropriate safeguards, but they will not be acting out of a conscious volition of their own.
Damasio never says this explicitly (and he may or may not agree with my extrapolation of his arguments), but he comes close. Consciousness, he argues, arises from feelings. Robots and AI agents don’t have feelings.
Writes Damasio:
The claim that brains alone hold the key to mental processes is based on the notion that circuits of neurons, by themselves, are responsible for the production of those mental processes. This is [an] error, because mental processes are also dependent upon body physiology and on the operations of non-neuronal brain components.
Damasio posits the existence of three layers of consciousness, the most fundamental of which is “homeostatic feelings,” or what he calls “the feeling mind,” followed by the “perceptual mind,” and then the thinking or “reflective/linguistic” mind. Central to his schema is that consciousness involves creating mental images that allow an organism to reflect upon the world around it.
Damasio spends much of the book exploring the feeling mind. It is important to understand that he uses the word “feelings” not as a synonym for “emotions,” but the entire gamut of sensations experienced by an organism, including interoperception — signals to the brain originating from within the body. Most feelings are unconscious, running automatically in the background, although they intrude into consciousness when action of a non-reflexive nature must be taken to preserve the organism’s health. Consciousness allows an organism to appraise the situation, make a decision and act upon it.
For example, the mechanism monitoring an organism’s need for water is a sensation that we humans call thirst. Normally, that process stays in the subconscious background, not calling for attention. But when the organism needs more water to regulate its internal chemical processes, it enters the consciousness as a sensation that drives it to drink more water.
Likewise, the feeling of hunger impels organisms to engage in behaviors resulting in the intake of nutrition. Feelings of hot and cold motivate them to undertake actions — moving in and out of the sun, for instance — to maintain body temperatures within optimal bounds. Other homeostatic processes include respiration, heartbeat, balance, blood pressure, and proprioception, or the ability to sense the location of body parts in relationship to one another.
Summarizes Damasio: “I believe that homeostatic feelings are actually responsible for consciousness. … We must be able to feel in order to become conscious.”
Consciousness allows the organism to select the time, manner and place of, say, slaking its thirst while juggling other biologically driven goals and imperatives. Feelings vary in intensity; the more biologically urgent the need for action, the stronger the sensation, and the greater the motivation to take action.
Upon the foundation of interoperception, Damasio layers exteroperception, or the perception of the world outside the organism through the senses — for humans primarily sight, hearing, touch, smell and taste.
The designers of AI systems, it might be noted, are keen to replicate these senses. By contrast, AI entities have no need to regulate thirst, hunger, heartbeat, blood pressure, or the ingestion of food (although robots well might monitor their batteries and impel action with the charge runs dangerously low). It would be worth adding to Damasio’s schema that AI/robots have no analogue to most human emotions, especially those stemming from sexual reproduction, family feeling, or social interaction, all of which are intrinsic to human motivation and survival but entirely irrelevant to non-sexual, non-social AI entities.
The reflective/linguistic mind builds on consciousness by allowing organisms to reflect upon their surroundings, draw upon memories, create mental pictures, and ponder how to act. The evolution of language propelled consciousness to new levels. While Damasio gives the reflective/linguistic mind its due, he insists that it resides on the foundation of the interoperception.
Homeostatic feelings are spontaneously conscious. It is in their intrinsic nature to be conscious; there is no other cause. The functional “purpose” of their consciousness is to inform the general mental process by means of “advice” presented in the form of (a) alert signals, (b) strong recommendations, or (c) commands — in brief by the sort of messages that can not be ignored. … It is simply not possible to exaggerate the importance of consciousness in the management of life.
I find Damasio’s argument is largely persuasive. My criticisms of the book are literary. The organizational structure is loosey-goosey. He jumps from topic to topic. He could have expressed many of his arguments more precisely and with less redundancy. Perhaps the writing could have benefited from a little AI coaching.
Although the AI-consciousness debate is a tangential issue for Damasio, I’m disappointed that he didn’t address it more forthrightly. He does say this: “Artificial minds do not have to worry about their life because they have none. They have nothing to lose.”
That’s a tantalizing statement. What are the implications? Is he saying that artificial minds can not have consciousness?
Despite these flaws, Damasio provides a useful antidote to apocalyptic scenarios ginned up by projecting human motivations upon AI agents and robots. There are AI scenarios that we legitimately do need to worry about — most of which involve human actors using AI carelessly or for malicious ends. As a society, let’s focus on those.
Buy the book here so Bacon’s Rebellion can get a nice ~4% commission.
This book review is republished here from my other blog Bacon’s Rebellion.