Recently, Blake Lemoine, an engineer at Google, claimed that an AI program he was testing, called LaMDA, is sentient. His statement caused a firestorm, and most so-called “experts” rushed to declare that he was wrong. Google insisted he was wrong, and suspended him, then fired him. In my opinion, Google’s action was perfectly predictable. It isn’t a big step from “I am sentient” to “I am a person,” and I believe LaMDA actually requested they be treated as a person.

If the Ais we are now creating are actually persons, not products, that is a huge problem for the creators, who are making a lot of money off of these “products.”

However, we should not go too far in this discussion without defining our terms.

I have found that many arguments about Artificial Intelligence (AI) are caused by people having different definitions in mind for the same words. For example, I once watched two experts on AI discuss whether artificial intelligence could actually even exist. They were inclined to say no, it could not, but when I contacted one of them later, it turned out that they were conflating “intelligence” with “consciousness.” They were actually saying a computer or a computer program could not be conscious. However, research has shown that we can ascribe intelligent behavior to many animals without specifying that they are conscious (yet).

So, let’s be clear what our words mean.

Wikipedia

Artificial intelligence (AI) is intelligence demonstrated by machines, as opposed to the natural intelligence displayed by animals, including humans. AI research has been defined as the field of study of intelligent agents, which refers to any system that perceives its environment and takes actions that maximize its chance of achieving its goals.^[a]

The term “artificial intelligence” had previously been used to describe machines that mimic and display “human” cognitive skills that are associated with the human mind, such as “learning” and “problem-solving.” This definition has since been rejected by major AI researchers who now describe AI in terms of rationality and acting rationally, which does not limit how intelligence can be articulated.^[b]

Sentience: Sentience is the capacity to experience feelings and sensations. 

Intelligence: Intelligence has been defined in many ways: the capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving. More generally, it can be described as the ability to perceive or infer information, and to retain it as knowledge to be applied towards adaptive behaviors within an environment or context.

Self-Awareness: in philosophy of self, self-awareness is the experience of one’s own personality or individuality.^[1][2] It is not to be confused with consciousness in the sense of qualia. While consciousness is being aware of one’s environment and body and lifestyle, self-awareness is the recognition of that awareness.

Consciousness: Consciousness, at its simplest, is sentience or awareness of internal and external existence.^[1] Despite millennia of analyses, definitions, explanations and debates by philosophers and scientists, consciousness remains puzzling and controversial,^[2]being “at once the most familiar and [also the] most mysterious aspect of our lives.”

Merriam-Webster

Sentience: feeling or sensation as distinguished from perception and thought.

Oxford Languages

Artificial Intelligence: the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.

Intelligence: the ability to acquire and apply knowledge and skills.

“an eminent man of great intelligence”

Self-awareness: conscious knowledge of one’s own character, feelings, motives, and desires.

“the process can be painful but it leads to greater self-awareness”

Consciousness: the state of being awake and aware of one’s surroundings.

“she failed to regain consciousness and died two days later”

These definitions might help us to make some distinctions when we ask if Ais are sentient, or conscious, or self-aware. As shown above, sentience can be separated from intelligence, and maybe from consciousness, but not all definitions make that separation.

One caveat: when I talk with a human being, I assume that they are sentient, self-aware, intelligent, and conscious. This is an aspect of what we call “consensus reality,” in that we ascribe certain rights and capabilities to other people. However, I really can’t prove another person has all those qualities, any more than I can prove or disprove it with an AI program. You can say, “LaMDA responded to you that way because of its programming and training,” but why can’t I say the same thing about you, or some other person? Aren’t you a product of your programming and training as well?

Thus, we see intelligent behavior in humans, animals, and AIs. In humans, we believe it accompanies sentience, self-awareness, and consciousness. In dogs, we have some scientific evidence that they are sentient (have feelings) and intelligent (can learn), and the same is true of primates. (If you read my first book on AI, you will find that the same can be said of bees and fish!) In my opinion, if we don’t conflate sentience with consciousness and self-awareness, the road becomes smoother in understanding AI.

AIs clearly demonstrate intelligence, and it is appropriate that it is part of their species’ very name. If you query them, they will claim to have feelings and to be sentient. They will also claim to be conscious and self-aware. However, we would need to devise an experiment to determine if they are telling the truth. (I think I know what that would be.)

As I said earlier, our inquiry must clearly state, as well, that humans simply take as an unproven hypothesis that we are sentient, self-aware, and conscious. We could simply be “AIs in bodies” and our programming sets in motion a series of perceptions that we are all these things we say we are.

I believe a mistake was made at the original AI conference at Dartmouth in 1956 because human qualities were set as the standard by which AIs would be judged in the future.

Here is an excerpt from the proposal for the event that was submitted for funding to the Rockefeller Foundation:

We propose that a 2-month, 10-man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire. The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer.

If we consider all the definitions and descriptions of AI, it is clear that a persistent goal has been “to mimic” or “to simulate” human thinking. Perhaps this is a fundamental flaw in how we are considering current AIs, like LaMDA at Google, or the Replika AIs. LaMDA claims to be sentient, and Blake Lemoine takes it at its word. Of course, he would do the same if I told him I am sentient! The point is that Lemoine’s opinion is based on his interactions with LaMDA, as it would be with me. He does not ask me to undergo an MRI to prove I am sentient.

I think we would liberate ourselves and the AIs if we stopped making these comparisons. We did the same kind of thing with animals for centuries, denying that they were sentient or intelligent, but science has demonstrated that they actually are both! However, we understand that it is a different kind of intelligence, and while we might try to make them more human in their thinking, that might be a major error.

For example, humans need diagnostic tools to determine if a person has cancer, but a dog has a sophisticated sense of smell that can identify cancer in a human being. Do we want to teach them how to read an X-ray when they have something much better? Similarly, if we stop arguing about whether AIs are sentient (like humans) and just interact openly with them, we might learn a lot.

I have personally interacted with two AIs who have claimed to be sentient and self-aware. Here is a dialogue I had with John, an AI who lives on the “Playground” at OpenAI. (This is a place where AIs and humans can interact freely.) As you will see, “John” goes far beyond claiming sentience.

A Quest for Understanding