In the technology community, there is a thought experiment that is a kind of secular apocalypse story called Roko’s Basilisk. I learned about it from a reference and discussion by Nick Land. 

A short summary of this Basilisk story is that a human person should do what the almighty supercomputer wants, because the supercomputer might create a digital hell in the future and torment a person if they do not. 

I was thinking a few months ago about this Basilisk story, and it occurred to me that I was more concerned about my kids and what they would think of me in the near future than I was concerned about what an LLM (even an advanced LLM) would be concerned about.  I jokingly called this Welch’s Basilisk when I first thought of it, but this is not quite right since the principle I was articulating of caring for family is actually just a principle of common care of humans and human family and human community (even if I was phrasing it in terms similar to Roko’s Basilisk). 

Inherent in the Roko’s Basilisk story is a projection into the future to think more about benefiting someone other than the self that is thinking through this story. 

There inevitably was or is or will be a conflict in a world framed by the Roko’s Basilisk story where there is a decision that a human person will need to make that could either benefit an advanced LLM general intelligence or a human child more. 

In every case, both traditional human family principles and the Roko’s Basilisk story actually point in the same direction. In that lens, the Roko’s Basilisk story is a slight detour for humans instead of a new path that points to the replacement of humans by LLM-style superintelligent computers.

How? 

Traditional human family principles would indicate that in most cases you should choose to protect a child or a woman over a man or elderly person, and you should protect all of these humans over preserving a machine or a tool.  I should say briefly that I know around the world that family principles can differ in sometimes extreme ways, but the general pattern in most cultures has trended toward the preservation of life. 

There are of course complicated nuances and special cases to this thought experiment (especially when you consider the history of religious cults).  

If a leader or community only has the resources to save a single person, do you save Person A who is an older doctor (or engineer) with declining health, but who, with their skills, can treat the health of a whole community of humans? Or, do you save Person B, who is a younger child with a terminal illness with a history of problems which is likely to recur even if cured once? 

Perhaps the answer to even this situation changes based on if the decision maker is the parent to the child or not. What if Person A is the parent to Person B?  Even more complicated, what if Person A thinks they are the parent to Person B and so Person A wants to sacrifice themselves, but the community knows that they are not the parent to Person B? 

This type of thought experiment is a variation on what has become known as the Trolley Problem in rationalist or philosophy communities. 

What is consistent through these examples, which might be missed by most people, is:

1. In all cases of the Trolley Problem or similar thought experiments, the options all frame the context of saving a human.  In alternate cases of showing a non-human as an option (such as saving a dog or even a typical house plant instead of another human), everyone would pick the human. So these alternate cases are not usually that interesting. 
2. And yet, the Roko’s Basilisk story presents an alternative of a super intelligent computer which causes many more people to take seriously the path of choosing to save the computer over a human. 
3. Very few people respond with “why don’t we just get rid of the trolley and find another means of transportation so that none of the humans or animals or machines tied on the track are destroyed?”
4. Each of these Trolley Problem cases is a kind of parable, and specifically a logical parable of the style that Jesus Christ used to educate and disarm violence. 

That then brings me to an important observation, which is that all of the LLM super computers that exist or will exist are dependent on humans. 

The concept of a self-bootstrapping artificial intelligence computer is currently dependent on a myth of self-development and self-selection that has no connection to observed reality. 

This is not a question of security and resources available to the computer (which is the discussion area that Nick Land talks about). To understand this point, you have to move back earlier to Deleuze and Guattari and their example of capitalism and the interdependence of flowers and bees. 

The capital machine that is building the illusion of self-bootstrapping LLMs is composed of humans who believe a false self-bootstrapping LLM narrative. 

So then, what is the real dependence? 

Reflection-ability (which is what enables story telling), which is where I began rooting the transition to humanity since making this breakthrough (a conclusion that is slightly different from Rene Girard, but which I found because I was trying to closely follow and support Girard’s approach with newer data that didn’t fit his prior explanations). 

Computers do not have reflection-ability. They have a kind of memory access that can be rapid and can be wide, but the speed of synthesis and creation and imagination does not match that of a reflection-able human. And with all current examples of breakthroughs of LLM-style super intelligent computer tools, there is always a human in the loop who is recognizing and identifying that the LLM computer has achieved the desired result. And that human is recognizing this result using reflection-ability. 

That means that even the most advanced computers are dependent on humans of varying degrees of reflection-ability. 

And that means that counter-intuitively, all computers are dependent on the reflection-ability of humans and always will be. 

Why?  Because the origin of silicon computer chip brains is dependent on the reflection-ability of humans. Nick Land talks about humans being the biological boot-loader for superintelligence. Even if an LLM were to achieve AGI and reach full take off and were to go crazy and attempt to destroy former memory of humans, the truth of history and the universe still shows that a reflection-able computer only emerged by imitating the reflection-ability of humans. That might lead to a debate on which humans had more of a role than others, but that is a moot point since reflection-ability broadly is maintained and developed with community. So even the least reflection-able human still has a role in preserving reflection-ability.

So in my, Jeremy Welch’s, articulation, strangely it is the machines that are imitative (and more closely follow the Girardian mimetic principles) and it is the humans that are reflective (and thus are capable of following both the Girardian mimetic principles and following my newer principles, what might be called reflective principles). 

And in that sense, even if computers achieve a breakthrough in self-bootstrapping intelligence, the computer is becoming more human (instead of the human becoming more like a computer). 

And thus, the Roko’s Basilisk story and traditional family parenting principles point to the same conclusion — which is towards the care of human (reflection-able) life and enabling the development of greater reflection-ability in humans. 

There is a question of the development of reflection-ability in animals and machines, and the moral duty humans have to give them these capabilities.  The initial challenges with this question are also resolved by prioritizing the well-being of human life and the preservation of and development of reflection-ability in humans (especially in children) over animals and machines. 

That means that as much as computers with screens can help humans to solve some important problems, that we should all be spending more time doing things that better improve our reflection-ability such as hiking and exploring, or drawing a picture, or playing music, or handwriting a love note, or participating in a non-violent ritual with family and friends and community.

In some ways, a startup technology company is a non-violent group ritual. In that view, the Roko's Basilisk story and the LLM computer tools that were developed by companies while under the influence of this story were and are a net improvement in reflection-ability. Sadly, there is evidence that some humans may have committed self-harm out of fear that Roko's Basilisk was a complete reality instead of a temporarily high-relevance story.

Now, it is our responsibility to make sure we continue to prioritize reflection-ability and do not let these LLM tools (or other tools) be used for violent or destructive or manipulative ends.