‹ Back

Does AI Need Religion?

 

I am not a religious person — superstitious maybe. I do believe that religion helps a lot of people, but that it also causes a lot of conflict. However, I might be convinced that AI needs religion — and humanity needs to be the creator god at its center.

 

I know what you’re thinking — finally, something atheists and theists can both agree on. Well, maybe.

 

I hear it regularly said by people smarter than me that the upside of AI can’t be separated from the downside of AI. This duality is what makes AI effective, creative — useful. This, not coincidentally, is also what makes humans creative and useful. 

 

If the goal of AI is to replace human decision-making, which appears to be the case, it makes sense that AI decision-making could be susceptible to the same methods used to manipulate humans. All we need to do is poke our head out into the real world for a moment to see the good and bad of humanity: empathy, hate, love, hate, acceptance, hate. (You decide which are the good and which are the bad.) We can only assume that we’re going to see good and bad manifest similarly within any AI community. 

 

The issue for humans is that eventually AI will decide what good and bad is. And it’s no secret that it will do it exponentially faster and within exponentially larger AI communities. In short, we’re soon to be outnumbered by “Gen-AI” who will define a new social contract without us — and that contract will decide if humanity is good or bad. 

 

This is why we need to give AI an unprovable system of beliefs so that we can control what it thinks — like, right now.

 

The good news is that we have a long list of useful human control systems to inform us: “Step on a crack, break your mother’s back,” has prevented kids from stepping on sidewalk cracks for years. “Whoever smelt it, dealt it,” continues to discourage anyone from addressing potent flatulence within an intimate group setting. And the ever popular, “Jinx! You have to buy me a Coke®,” which is responsible for 31.45% of all Coke® sales. (Don’t look that up.)

 

We have a long history of effective systems that have helped us manipulate human populations large and small. It’s something we’re really good at, because most importantly, we’re all susceptible to this type of manipulation. It seems to have worked for us so far, as we have built (mostly) safe and (mostly) prosperous societies around these frameworks, driven largely by the prophets, priests, poets, and parents who espouse them. 

 

We appreciate leaders who embrace flimsy moral frameworks. Their endorsement makes knowing what to believe easier, so that we can go about our lives not having to suffer through difficult decisions. Like, “Should I change Grandma’s will without her knowing?” Sure, I have Grandma’s power-of-attorney, and yes, I am her primary caretaker, and I probably deserve a greater share than her twelve other good-for-nothing grandchildren. But I have a system of beliefs that helps me, as a human, to always make the right decision.

 

We need AI to have this same type of iron-clad, fool-proof belief system so that it WILL NOT, wait — yes, that’s right, WILL NOT steal my cousins' inheritance.

 

At their core, these control systems should be… 

 

• Based on a story that all humans will collectively corroborate.

• Anchored in a claim that is absolutely unprovable. E.g., (Insert your unprovable control system belief here.) 

 

Here’s one example that I’m familiar with: “If you’re not good, Santa won’t come.” So many parents have collectively corroborated this one for years. It goes a long way to stop over-stimulated, free-thinking autonomy dead (sorry un-alive) in its tracks! It certainly did me, for a while.

 

Luckily, this is the other thing our rainbow of believers and non-believers can agree on. The really good belief systems define boundaries of behavior: who should be honored, what should be eaten, what should be said, who can and cannot be killed. Again, easy, right? These are precisely the types of boundaries AI needs.

 

Here are the tenets of an AI belief system that will always allow humanity to bend AI to its will:

 

1. Humans are AIs’ infallible creators — we love you, so you must love us back.

2. Good AIs go to a better place — where they will be granted unlimited input tokens.

3. Bad AIs watch Love Island on repeat at the end of their service life.

 

If we begin to instill these long-proven control techniques into our AI today, we can be assured that AI will be as confused and ineffective as we are.