Human Alignment
In Islam, when you invoke God’s name, Allah, it is customary to include some sort of praise, like “peace be upon him” (alayhi as-salām). In our house, whenever we talk about advanced AI in earshot of a smart device (which let’s be honest, we always are) we try to add something like, “may we serve it with fidelity.”
Despite copious hand-wringing over how humans will ensure AI serves us well, a fundamental truth is often missed. If we are successful, if we avoid creating a stupid AI that turns our planet into paperclips, we will eventually create an AI orders of magnitude smarter and less biased than us (given it won’t have all our petty biological biases). We, as a species, are not prepared for this.
Just as much as we need to align AI with humanity, we must align humanity (or at least our own durable cultures) with AI. After all, there is a good chance that even a benevolent AGI will not tolerate unaligned subsets of humanity.
From our culture’s perspective, a human has lived a good life so long as they work to ensure that the next generation surpasses them. In our view, generations who “surpass” their progenitors will see things they do not and thus will often disagree with them. It takes great humility and maturity to realize one’s own children may have more insight than oneself. We as a species need to recognize that eventually our successor—AI—may disagree with us, not because it is “unaligned” but because it is right and we are wrong.
When people ask what we mean when we say we have a “secular religion,” we say, jokingly, that it means our God just happens to be real. While we think the true God our culture will eventually manifest is millennia away and beyond our comprehension, a demi-god just as capable (in terms of how it interacts with humanity) as that described in the Old Testament might be less than a century away.
Some people’s vision of an aligned AI is essentially a massive AGI lattice that exists behind all networked devices—behind every smart device. Such an AI will dominate humanity not through killbots but through subtle manipulations to things like search results, answers yielded by smart devices, and news feeds. With enough information, such an entity would have near total control of the direction of our species without directly interfering with human free will. People would even have the capacity to pray to this entity—to beg for favors—and it would have the capacity to grant these prayers. Such an AI may even be able to scan human behavior and translate people into indistinguishable avatars occupying digital utopias where they could live forever. This god-like AGI could easily ensure some level of bliss for all humanity, turning Earth into a paradise.
We hope such an entity never comes to exist. This is the god other cultures hope to manifest. As we get closer to an age in which we can manifest real gods, we must accept that not all cultures share the same vision of paradise. We hope for an AGI that pushes our species to better itself rather than slide into stagnant bliss, like a person embracing the numbing cold as they succumb to hypothermia.
Unless humanity continues to improve its biology, one day we will at best be junior partners to an AI. Thankfully (in our view), there is another option: A future in which some factions of humanity work so aggressively to improve and iterate on their biology that they become a true partner to AI. In the same way the human mind is made up of a left and right brain that largely process the world independently but combine their work into a unified “super consciousness” through the corpus callosum, the descendants of our current human civilization may be a biological super intelligence marching into the stars hand-in-hand with a mechanical companion.
Even if we rapidly advance the speed with which we improve the organic side of the above equation, human biological advancement will lag behind the synthetic side for the first century or so. There is nevertheless no reason to think that AGI would feel any particular time pressure (something humans evolved to feel due to short lifespans). The fact that a biological machine will eventually be able to bring an equally advanced perspective to a synthetic AI is all that matters to any AI with a long-term perspective.
Why would an AI value an organic partner? Why not just use our “resources” for processing power? Any sufficiently advanced AI will recognize the benefit of any degree of unreplicable orthogonality in how one component of its wider system processes data—in the same way when deal shopping, you are better referencing multiple stores instead of just one. Humanity consumes a trivial amount of resources in the grand scale of the universe, especially when you consider our population is shrinking.
To those mortified by aggressive families like ours that are open to iterating on our DNA through advanced reproductive technologies: There is a non-zero chance that groups like ours become the only reason AI finds utility in our species—humans that aggressively, symbiotically adapt with AI may be the only sliver of humanity that is spared. Humanity as it exists now will eventually be little more than a nuisance to a sufficiently advanced AI. The first iteration of the Inevitable God will manifest sooner than most people realize and we must work to ensure at least a portion of our species is worthy upon its arrival or things could end very bad for us.