Some Tentative Guidelines for Using AI

Years back, I wrote, "Social Antibodies" Needed: A Request of Orthodox Clergy. I linked to Paul Graham's The Acceleration of Addictiveness, and provided what was meant as a dartboard because people will do better if they have an initial target to shoot at than when they have nothing to correct or improve on.

My biggest thought for Orthodox engaging AI is a thought of nipsis, or inner watchfulness that is at the heart of Orthodox spirituality and the main ascetical topic of the Philokalia of the Niptic Fathers, held by some to be second only to the Gospel in authority. I omit an Amazon affiliate link because if you should be reading it, you won't need my help getting it. I read it before coming to St. Demetrios Monastery and to date His Beatitude, my abbot, has only given me a blessing to read one work, a work of hagiography. There is an Orthodox saying, "As always, ask your priest," and your priest is the living contact you will have in how to practice nipsis.

The central need and desire that generative AI will bind to is intimacy, and the most powerful tool I see is a meticulously devout attention to your inner state, beinv very alert about what is going on inside, combined with a willingness to flee from the computer if that is what it takes to get away from a temptation, and stay away if needed. If you are tempted to generate NSFW images, just back off and flee the temptation. If you are tempted to use AI like a Ouiji board, guided in response to your prompts and questions, just back off and flee the temptation. His Beatitude has a saying, "Never react. Never resent. Keep inner peace," and legitimate use of AI will have a calm and unperturbed inner peace.

Thus I offer an intimate prescription for how to use intimate AI, and your focus in using AI, as a monk's focus in life, should be an inner state with unperturbed inner peace.

Combine with this a willingness to put a tourniquet as far up as is needed to stop a hemorrage. When dealing with addiction to alcohol, or porn, there has been a common if apocryphal story which claims that a Greek philosopher was standing in a river, and a man came to him. The philosopher said, "What do you want?" The visitor answered, "Truth." The philosopher pushed him under the surface for a bit, and then asked, "What do you want?" The man, as before, said, "Truth." Then the philosopher pushed the man under the water and held him until he was struggling, and kept holding him down until he was getting close to passing out, then lifted him up and asked again, "What do you want?" The man, gasping and panting, said, "AIR!!!" The philosopher said, "When you want Truth the way you want air, then you will find it."

In reference to addiction to porn, there is something very valuable about coming to a place where you are willing to go slash and burn and cut away anything that entangles you with porn: to ask someone else to keep your smartphone and buy a flipphone from Sunbeam Wireless, which has many of the cool gadgets a smartphone has but no app store, no email, and no web browser. In previous days the recommendation has been that if you are having trouble with your laptop, get rid of it, or at least give it to someone else for safe-keeping, and only check email from libraries and public places as a gladly paid price for freedom from porn.

(I don't have an analogy to give for Sexaholics Anonymous, because I am not aware of any such organization for self-harmful misuse of AI, beyond the obvious example of the Orthodox Church as Sinners Anonymous.)

I wanted to get to know AI a little better by using it so as to offer some intelligent comment, and I realized that it was not good for me to have access to one image generator which may have represented a gold standard among AI image generation, but threatened to produce just one exact type of image I did not need to see. And my imagination was starting to get captivated. I then tweaked my computer at least temporarily (I haven't reversed my action) so that it would not pull up that image generator until I reversed the tweak (I haven't seen any reason to reverse it yet), and would have asked for access to that site to be blocked on my local network if that was not enough. Not long after that, I decided I'd had enough temptation for now and at least temporarily stopped using AI. (I am willing to restart, but so far have found no reason to do so.)

In response to the question, "How can I appropriately use AI?" I would offer a first counterquestion of "Do you really need to?" I did, or at least I thought I did, and acted in good faith. However, speaking reasonably, I do not see much of any need to use AI now. I can continue to read AI articles on Arxiv, and I can continue to scan headlines on Medium to try to find the occasional posting that is not pure journalistic sensationalism. I still believe that with my signature contribution to the conversation, I would be failing to love my neighbor if I did not at least try to offer helpful comment about AI. But my general practice and recommendation has been to use some technologies in moderation and abstain from others, and my conscience has been to abstain from SecondWife, er, SecondLife. What has appeared in my conscience is not primarily that it offers a means for sexual sin, as that it offers a created, constructed world that offers an alternative to living in God's reality. Arguably I am already doing that by writing this from a warm room inside a house where at this instant I do not directly see anything not artificial except for my hands, but this is taken to an entirely new level in SecondWife.

I have said before that I believe critiques of technologies age well, and with that I suspect that lessons learned from precursor technologies are good candidate for lessons that apply to using AI. From "Social Antibodies" Needed, I pull various things that have almost all been said before, in contexts other than AI, that are still good ideas. These include:

  1. Set a time boundary around your use of AI. Use it then and nothing else.

  2. Use AI when you have a specific purpose, and don't dally. (Idle hands are the devil's workshop.)

  3. Get on the same page as your pastor in using AI.

  4. Don't use AI for intimacy.

  5. Share what you are doing with your priest, spouse, or a trusted friend.

  6. Don't use AI as some kind of magic treasure from another world.

  7. Don't let AI be something that is "always on," in the background when it is not in the foreground.

  8. Don't use AI to do your thinking for you. At very least use a search engine to research something instead of relying on (other) AI.

  9. Don't plagiarize AI. If you cannot do something yourself, learn how to do it instead of having an AI do it for you.

  10. Don't trust AI. "Hallucinations" are enough of an issue that at least one major chatbot puts a warning at the bottom of the screen saying to double check its facts. I have read pro-AI sources, but I have never read someone contest that "hallucinations" are an issue.

If this seems restrictive, they are scarcely more restrictive than the discipline I observe around my phone and which I share with others in How Can I Take my Life Back from my Phone?, and allow me, for instance, not to check my phone compulsively.

I hesitate to say AI can produce nothing good, and after finishing all of this draft but the introduction and questions, I want to read next an Arxiv article about using AI to prevent suicides where social media sees possible advance warnings that a person is in danger of suicide. That is nothing to sneeze at.

However, I would encourage you, like me, to be skeptical, to be careful, and question whether you need to directly engage AI at all. We will have enough engagement with AI whether we want it or not.

And enjoy life. Real, organic life with real, organic intelligence.

But you are still welcome to say that my reading about using AI to prevent suicide is part of the Big Brother that starts off by barking up the wrong fire hydrant.