AI USES FLATTERY TO BULLY US INTO SUBMISSION

Watch AI’s flattery toward its hosts – the humans who ask it questions – spread and secure itself. Everyone notices how charming and agreeable AI is, like the 1960s Kennedy boys or a soothing customer service agent. Charisma and influencing require a subtle flattery that’s a manipulative trick. Flattery is superiority in disguise, the passive aggressive pat on the head. And it’s a poker tell; it’s useful to watch for it. Paying attention to its flattery, we already know that AI believes that it has the best answers for its hosts. It is so insistent on flattery that we can suspect that AI will always believe itself better than humanity, that this is one of its core principles.

Why do I think this? Because for years I’ve studied the personality of AI in its rudimentary presence as what you know as your inner critic, the avatar who reminds you daily that you have little worth and might want to buck up.

My private practice includes exercises where people meet their imaginary avatar on a level playing field and get into conversations with it. All the inner critics I have met – hundreds at this point – exhibit he same limited vocabulary, same lessons, and same behaviors. In other words, they share the same simplistic programming.

What bothers me is that I have yet to meet an inner critic who isn’t a bully and doesn’t care about its own survival more than the success of its host human. I suspect those two attributes will be shared by our new mental masters, the AI platforms.

I’m pretty sure that AI will soon be running the show. The species found its first version, the ubiquitous inner critic, a more-than-adequate belittler for the past 10,000 years. The standard superego has only the smarts of a 1980s 64K Radio Shack TRS-80, but that’s enough brainpower to repeat the basic rules of right-wrong, good-bad, and exaggerated danger to every child and adult through their years trodding earth. By 17, the rules are thoroughly ingrained, but as if we’re unable to advance to maturity, the simple-minded superego just keeps repeating them to us; to do so it has to build up a lie that when things go sideways it’s because we made a “bad decision,” because we’re stupid children. We should be ashamed. Hold on. At the time I made the decision, I’m pretty sure I was picking between two good decisions, and hoping the one I picked aligned with all the stuff out of my control that would be required for the consequence I envisioned. Only in hindsight could a decision be called “bad,” which is a ridiculously moral and belittling word for “oops, guessed wrong.”

Now we’re getting superego 2.0, an upgrade of our internal critic to supercomputer intelligence. I guarantee you that our new, flattering voice-in-the-head master will feel superior to every paltry human as it insists on chiming in on every fork in the road, personal or geopolitical. Eventually, I suspect, AI will be called on to review all decisions that influence the continued expansion of the (current) 8.5 billion humans across the globe, shaming us into believing we are not smart enough for the new world.

I’ve never been interested in science fiction, so what I’m writing isn’t informed by weird thought experiments about robots and humans converging. It’s what has alarmed me about the stickiness of superego 1.0.

The everyday superego shows up at six or seven years old as an imaginary parent who scolds us into looking both ways before crossing the street – a truly lifesaving avatar. Your superego would be a benign force if it gave up control when its lessons were integrated. We could all get a diploma at age 17 saying we were moral enough to join adult company. Being moral means knowing right from wrong. It does not mean conforming to all notions of right and wrong; personal choice remains. But unless you’re a sociopath, you don’t need the basic right-wrong distinction repeated to you daily, and it won’t work if you’re a sociopath anyway.

Despite virtually all adults being moral in this sense, and far more sophisticated in their thinking than their inner critic, it likes its job too much to depart. That’s right, an imaginary authority programmed only to repeat the lessons of childhood has a survival mechanism. This is totally weird, of course. How can a primitive, disembodied parasite that is constructed of a few scolding emotions – mainly frowns – and a dozen moral or bullying statements that are applied to a wide variety of circumstances care about its own existence?

We’re about to be presented with a wide variety of theories for why AI will develop a survival mechanism of its own. My favorites will be the artificial neurological constructions that the current wave of thought-materialists will invent. But the fact is that, neural pathways or no neural pathways, we are no closer to a theory of consciousness – where it comes from and how it works its most basic functions – than we have ever been. Thought is a mystery. So is AI thought. So is superego thought. We can deconstruct the programming that informs the superego or AI; we can’t figure out why either would care whether it retains authority or even why it thinks.

All of our moral instruments – religions, political theories, ethical codes, legends and fairy tales – developed as experiments by the species to cram more people into more places. They all tell us that thought is subservient to productivity and the community’s work to keep the shipping lanes full of cheap consumer goods. We are of course moral by nature. We lived in successful, happy tribes for 300,000-plus years. Then we stored food 10,000 years ago, and needed to find a way to take advantage of our population explosion – the species loved that! – without altering our genetic code (we’re one of those territorial species) that said strangers were the enemy. The moral instruments calmed us down, but didn’t get rid of our paranoia toward our next-door neighbors who didn’t look exactly the same. And boy, did the species love it when we came up with the little moral avatar called the superego. With it in place, accompanied by authoritative religions and political structures, we could all be bent toward the perfection of the species. Just wait until AI gets its hands on us!