The Art of Spiritual Prompting

I'm blogging early this week because I have a lot of ideas swimming in my head. All these ideas deserve their day in my blog. In fact, I keep a list of future blog titles that I plan to tackle whenever I'm inspired to do so. I'm writing this from Coffee Bean at what used to be known as the Da Men Mall. This mall is undergoing a makeover, and it has changed its name to EasyHome. For now, the place is quite deserted. I'm happy that at least Coffee Bean is still here. I've spent many productive hours here in the past before working on my projects. Let's see if I can also make this session, fueled by a hot Americano, a productive one too.

Today, I would like to discuss the concept of 'prompting', which has become familiar to us in the era of LLM-based generative AI tools like ChatGPT.  There are so many YouTube videos and articles out there that teach you how to coax LLMs to produce better outputs using the correct prompts. Apparently, it's mostly the quality of the prompt that determines the quality and accuracy of the generated response from the LLM. It's basically baked into the way LLMs work, for the LLM's response is composed of the predicted words or tokens that probabilistically match the context of the question or preamble text provided by the user. 

The surprising thing that LLMs have taught us is that what we would normally consider as 'intelligence' has a lot to do with how we associate words and string them together coherently to produce thoughts and ideas. LLMs weave words together like threads in a tapestry, predicting the next strand until meaning emerges. 

With a correctly engineered prompt, we could induce the LLM to output a chunk of code, even an entire application, that works right out of the box. Nowadays, you see many content providers trying to show off their coolest "one-shot" prompt-engineered application on YouTube. 

Let me now try to tie the concept of prompt engineering to spirituality. Intriguingly, all religious traditions--scriptures, their iconography, rituals and liturgy--serve as a kind of hyper-prompt to induce some kind of spiritual insight. For example, the student of the Bible would claim to be divinely inspired (or perhaps prompted?) by the text to experience certain epiphanies. Without the required contextual input of scriptural text coupled with his life experiences, he certainly would not be able to gain the same insight. 

Religious awakenings often come unexpectedly in a flash. It is simply the culminating effect of the right multimodal prompt. It could be an evocative turn of phrase from a religious verse, or a chance conjunction of events--golden sunlight through the clouds, the swell waves in the ocean, or the reverberating call for prayer from the mosque--that evokes a sense of awe and mystery, which immediately points to glory and the existence of God.

All spiritual traditions are collections of curated prompts that serve to produce in the believer states of divine rapture. The LLM gains its intelligence by seeing the interconnections between words through its laboriously trained neural network. Religion acts as a similar framework for believers to tap into its potential network of embedded wisdom. 

Just like how LLMs under certain conditions could hallucinate, religions could also induce their followers to go astray with false beliefs. The line between wisdom and violence can often be a fine one. We should learn from how we use LLMs and apply the same principles to religion. Only correct prompting would determine if the output contains erudite insights or hallucinatory dogma. Whether in code or in scripture, the art of prompting reminds us that wisdom depends not just on the system, but on the care with which we ask our questions.