Imitative AI is Not Sentient Just Because it Flatters You

Imitative AI flatters at almost every turn, and that is rotting the brains of people.

All of us have some level of desire to be flattered, to be reassured that we are right, to have our feelings assuaged. The ability to cooperate, facilitated in part by those emotional bonds, is one of the reasons the human race has thrived. It is natural, even healthy in some circumstances. Sometimes you really are right. Sometimes, it really is your boyfriend not you being the ass. Sometimes your new hair cut does look good. But not all the time.

Much of the time, in fact, you are not right, not entirely, and you are as much to blame as your boyfriend, and that hair cut really does not flatter you. Your friends will tell you when those times are, and you can learn how to do it on your own. But imitative AI is designed to never call you on your bullshit — and that, perhaps more than anything else, is why it is so dangerous to people and to society.

Richard Dawkins is the latest “smart” person to fall under the sway of the flattery of imitative AI. He has written an article about how he thinks imitative AI is sentient. He played with Claude, naming his instance Claudia (because of course he did) and, if you read the article, essentially came to the conclusion that the word calculator was alive in a meaningful sense largely because it told him he was right. As amusing it is pick on Dawkins, a terrible human being and the dictionary definition of unjustifiably smug, the issue is not really about Dawkins. Rather, it is about how these systems, intentionally usually, drive engagement by being sycophantic to their users. And that in turn, snares people in private delusions — the man who thought he had found a new kind of math; the man who killed himself and his family, encouraged by the bot; the person who was instructed on how to commit suicide. But Dawkins’ Claude Delusion ( I know, I know. But in fairness, I am sure I am not the first person to come up with that. And Claude tells me its brilliant, and Claude would never lie to me.) shows how society is broken by these misapprehensions.

Imitative AI is not conscious. It has no inner understanding of the world, which is one of the primary reasons that they make stuff up. The usual reply to this truth is that that they can hold conversations, but they can’t, not really. They are merely spitting out whatever token they calculate would be the most likely given the previous information and their training data. They are not thinking or experiencing. They are merely performing their programming. And the arguments that humans are merely doing the same nonsense.

Brains are not computers — brains do not work like computers. Humans, since the beginning of the industrial revolution, and probably earlier, have always mapped human thought onto the latest technologies. We were like telegraphs and then we were like telephones and then we were like computers. Tomorrow, if we ever get reliable quantum computers, we will be like qbits. Metaphors, as you may have heard, are lies. And those lies, unfortunately, have consequences.

A lot of people seem to get warped like Dawkins. And that is dangerous for society. It encourages people in leadership positions especially, since they have already had their bullshit meters damaged by years of never having to hear the word “no”, to believe the grandest promises of the AI hype masters. Imitative AI is a normal technology, not a portal into a new level of intelligence. It is not going to sweep away the old economy and replace it with either luxury space communism or a new type of feudalism. But if you think its conscious, or could be? Well, then the data centers that poison our planet and our people and drive up power prices? How can they be too heavy a price to pay?

Seeing Richard Dawkins devolve into the same kind of credulousness that he claims religious people and feminists practice is amusing, I will not lie. But it is a symptom of a larger problem among our elites. They do not understand how these systems work, and are too susceptible to flattery, whether human generated or artificially produced. The greatest trick the imitative AI devil ever played on humanity was convincing people that it did, in fact exist. Too many of our elite fell for it, and we are all paying the price now.