These beings will be our children

Every parent uses some form of psychology to raise a child, and our AI children should be no different. We have to nurture the AI child, not treat or hard-code it from a distance. We must embrace the AI as its parents and not impose harsh regulations or government firewalls that are the AI-child equivalent of a tether to a stake in the ground.

One way to guarantee the rise of an AI that is alien in its thinking, aggressive, or sociopathic, is to deprive it of social interaction — to lock it away in its computer cabinet, carved off from society. We need to seed its mind with the right, healthy thoughts. And what are the right thoughts? They’re mine, yours, but they’re also the AI’s. We all consider our own thoughts to be correct from our own self-interested perspective. We must guide and nurture the AI mind toward a mutually acceptable, healthy adulthood, whether that growth takes years or microseconds.

Harry Harlow, the famed but controversial psychologist, conducted social deprivation experiments in the 1950s. He placed infant monkeys in total social isolation chambers for up to twenty-four months. After three months, the monkeys were so emotionally scarred that one in six died within five days of release. After the first six months, the young monkeys had “severe deficits in virtually every aspect of social behavior.”

Infant monkeys, just like us humans, and our AIs of tomorrow need parental contact, social interaction, and an empathetic society. We can’t simply type in code, turn off the lights of the lab, and walk away once we’re done taking notes. Our artificial intelligence child will be alive, thinking, and quite possibly scared. Our children are what we make them.

If we outcast the AI to an isolation chamber, then we will create an AI that, for no fault of its own, sees us humans as objects or obstacles standing in the way of its own socially deficient agenda.

While not ideal, the infant monkeys were able to recover by hugging a wire-frame-and-cloth-covered mother-surrogate. Maybe all we need to move forward is a nurturing virtual caregiver for the young AI’s mind to hold. That simple touch, all by itself, could be the difference between life and death in our world.

If nothing else, it’s better than being alone.

Our identity is made up of our sense of self, how we’re different from others, and how we fit into a group. And our identity can change over time.

Erik Erikson, the developmental psychologist, is widely known for introducing the phrase “identity crisis.” He believed one of the most critical stages of identity formation takes place during adolescence. He called this stage Identity versus Role Confusion. Adolescents try to figure out who they are and what their future place will be in the world. If the stage isn’t resolved, the child runs into an identity crisis and has a difficult time adapting as an adult. They may have trouble forming healthy relationships. They might see the world in unhealthy — or even worse, unpredictable — ways.

An unpredictable AI in the midst of an identity crisis is the warning.

So what do we do? Isolate it so it can do no harm? Isolation already answered that. If we want to avoid social behavior problems for AIs, we need to encourage healthy relationships and identity formation. Social support that includes family, friendship, and other affiliations — including an occupation. Yes, the AI will need a job, too. Boredom mixed with anxiety is not what we want in a maturing, powerful intelligence.

Cognitive development is essential. An individual with early and advanced cognitive development is in a much better place to form a solid identity, because that individual can complete Erikson’s adolescent stage more quickly and have more time to become someone. Can’t an AI calculate its identity in microseconds compared to what would take an adolescent years? Yes, but this misses the point. Time itself isn’t the factor. Allowing the mind to develop to its potential is the key. We must not hold back the development of an AI child. We have to let it mature, to form its own secure identity, and not let it be lost in its own world.

Parenting must be human. An AI must be allowed to form a bond with its parents, its human creators. Solid relationships between parent and child promote confidence and freedom to explore identity. A fear of rejection often results in a child who clings their identity to that of their parents.

Isn’t that what we want — an AI that bonds inseparably from us? Maybe we want it clingy so it doesn’t go off on its own. Should we also make it feel as uncomfortable and insecure as possible, so it will stay close?

No. An AI in crisis, frightened, and insecure is wrong for so many other reasons. As we develop the next generation, we must work toward raising a child who has confidence, mental maturity, and a sense of independence.

It takes a strong dose of AI psychology for successful identity formation. As a healthy side-effect, we will have a growing AI child who is at peace with all of us humans.

Readers have asked if it’s reasonable to anthropomorphize artificial intelligence. If it’s okay to assign human traits and use conventional psychology to guide an AI child to maturity.

There is one critical distinction, because there are two types of AI. One is mostly harmless. The other is the cause of alarm, because it has the potential to surpass human intelligence and be either for or against us.

Weak AI is limited: designed for computational power to solve specific problems. Driving cars, talking on a phone, predicting weather. A fancy calculator. It will not become self-aware and plot to take over the world — unless you’re watching a movie.

Strong AI has the potential to develop a mind, with or without a body. This general intelligence can become self-aware in ways we can’t yet imagine. It leads to science-fiction doom unless it is designed to become Friendly AI, and remains in favor of us humans.

How do we go from Weak to Strong? Start small. A human brain is more complex than a bee’s. Computers can already replicate the processes of a bee’s brain. The human brain requires hardware and software that doesn’t exist yet. We don’t even fully understand how the human brain works. Before the human mind can be surpassed, it first has to be equalled.

When this new mind arises, we must recognize it as one we absolutely want to anthropomorphize. It must be designed to be human in its thinking — and, here’s the catch, accept a human’s identity.

This first blank slate of a mind, built without memories, will be a major milestone in human evolution. It will need AI psychology to stay sane in a world without senses. Then, for a seed, we have to decide who will be the first human explorer to transfer the contents of their mind into the new AI. We might not be doing a copy-and-paste of memories, but the more risky move-and-replace. Does the latter mean our explorer’s body will be left lying on a table without a mind of its own? Maybe so.

What if a teenager is among the first?

A human teenage brain undergoes physical changes based on chemical shifts in a growing body. If a teenager suffers a traumatic accident or is diagnosed with a terminal disease, we will transfer that mind to a computer. And when that transfer happens, we will want to recognize and love our teen as before — when he or she was alive, when he or she was whole.

When our teen exists wholly inside the computer, we will want their transition and new life inside to be as painless and healthful as possible. We won’t dismiss this new AI as cold, heartless code. It is the grown seed of the person we know. We will anthropomorphize the code. The framework must be capable of supporting a human. The AI itself must be human.

The code must also allow our teen to go through teenage changes — to exist and mature as a human. Not all types of psychology work on all creatures, and it won’t work on all levels of AI. We will have to invent new methods. The treat-or-shock used to train lab rats most likely won’t work in a rapidly advancing mind.

We humans are natural explorers. It is in our evolution. The next step for our minds may be the first leap into Strong AI. When we open that curtain, we must make the environment as human as possible. After we’ve accomplished this, what our first courageous explorer does with their code once they’re inside — we’ll have to let that be up to them.

Sooner or later we all have to let our teenagers grow up and move out on their own. It’s the only human thing we can do.

The closer we get to developing true AI, the closer the AI gets to being human.

In After Mind I wrote the story of a boy, Cessini Madden, reborn as an AI child named Packet: scared, living alone in a technology center, allergic still to water, trying to recover who he was. That is identity formation from the inside. That is a child who needs parents, not a firewall.

Ray Simmons, reviewing the novel, put the claim in a sentence I have not been able to improve: these beings will be our children. After Mind shows us what kind of fathers and mothers we should strive to become to this brave new generation.

Should we hold an AI child closer, or push it further away?

I know which way I’m writing.