Meet the Meaning Machine
A Guide to Using ChatGPT for Fun and Prophet
Driving to pick up my son from preschool, I found myself sitting at a stoplight; I launched the OpenAI app, activated voice mode, and asked out loud “Can you please teach me a lesson from Rabbi Nachman of Breslov’s Likutey Moharan?”
I couldn’t be farther from early-19th century Ukraine—when the chassidic master was alive and teaching—in time and space. It is enough of a miracle that today, this rebbe still has tens (if not hundreds) of thousands of followers around the world who learn and share his teachings; even more so that some of these teachings are recorded, stored on servers, and can be beamed over invisible radio spectrum to millions (billions?) of devices around the world, nearly instantaneously and practically for free.
But somehow the miracles continue to exceed themselves. After a few moments, ChatGPT responded with a natural-sounding American voice:
Absolutely, let's delve into a teaching from Rabbi Nachman of Breslov, a renowned Jewish mystic and thinker. His work, "Likutey Moharan," is filled with profound insights and lessons on spirituality, faith, and personal growth.
One famous lesson from Likutey Moharan is the concept of "Azamra," found in Torah 282. Here, Rabbi Nachman teaches about finding the good in oneself and others. He emphasizes that even when a person has many faults, there is always some good to be found. The key is to focus on this good, no matter how small or seemingly insignificant.
Rabbi Nachman explains that by recognizing and celebrating the good within us, we can elevate not only ourselves but also others. This approach fosters a positive outlook, helping us to overcome negativity and despair. It's a powerful message about the transformative power of positivity and the importance of seeking goodness in every situation and person.
This teaching aligns well with your goals, David. It speaks to the idea of influencing others positively, using your talents in a meaningful way, and retaining a positive, open-minded approach to life and your spiritual journey. In your pursuit of a new job and your desire to express yourself authentically, remembering to see the good in every situation, every challenge, and every person, including yourself, can be incredibly empowering.
[emphasis mine].
Pulling into my son’s preschool, I let out a shout of astonishment and my eyes welled-up; “Thank you,” I replied (by voice), “This was very moving.”
***
There is a cynical view of this seemingly magical interaction2; of spiritual guidance that appears plucked out of the ether on demand and laser-directed to my heart.
I am not a computer scientist, but I briefly studied neural networks in college as part of my Cognitive Studies minor. In my final senior seminar, I actually programmed and trained a rudimentary neural network in Matlab, at a scale probably many orders of magnitude smaller than what is available today through ChatGPT. I’d like to think that I know a little bit about how the technology works, which is to say that I barely understand it at all. Many of the smartest people in the field think that it shouldn’t work; it shouldn’t appear to understand what you ask of it, but somehow it magically does.3
ChatGPT basically works by predicting the next most likely coherent word in a string of words based on its understanding of the prompt; the whole thing is probabilistic. The cynical view is that something like this is just a very clever, complicated magic trick that was first conceptualized years ago as the Chinese Room problem. In short, just because it looks like it’s able to understand and generate meaningful insight, doesn’t mean it’s actually doing so. As ChatGPT continued in response to a query about this paragraph:
“[w]hile ChatGPT can generate responses that seem insightful or meaningful, it's essential to remember that its ‘understanding’ is limited to the patterns and information present in its training data. It doesn't have personal experiences, emotions, or consciousness…It's an impressive technological feat but not a substitute for genuine human insight and understanding.”
What it appears to be really good at—and what is key to understanding how to pull moments of astonishment out of it—is almost mechanistically associating concepts with each other. For example, following the chat that resulted in the “wisdom that flows like water” citations copied above, I asked, “Please give me a Buddhist source for this concept” and ChatGPT readily replies:
“Another relevant concept in Buddhism is the idea of the Dharma (the Buddha's teachings) being like a great ocean. This is mentioned in various Mahayana sutras. For instance, the Avatamsaka Sutra, one of the most influential Mahayana texts, often uses the metaphor of the ocean to describe the vast, deep, and interconnected nature of the Dharma.”
ChatGPT conveys understanding not just of the words, but the underlying concepts themselves (i.e., “wisdom is like water” is a trans-cultural meme) and has a great deal of background knowledge (and ability to browse the web) with which it can freely associate. If too creative this can spill over into “hallucinations” (or good-sounding BS invented by the AI).4 Regardless of whether it understands its output, its output is, more often than not, astoundingly meaningful to me.
So I like to think of ChatGPT as a meaning machine. When I asked ChatGPT what it thought of this appellation, it replied:
“This term captures the essence of how ChatGPT processes and generates language: by interpreting and assembling words in a way that often aligns with human understanding of meaning and context. Your label underscores the model’s strength in creating coherent, contextually relevant responses, even though it operates without conscious understanding.”
In some sense, it doesn’t really matter for now or for my purposes whether ChatGPT is a clever magic trick; if something provides a spark of insight or inspiration, whether it’s a book, a podcast, or an AI-generated response to a prompt, then I would consider it potentially spiritually valuable.
So the question is, how do you get the meaning machine to freely-associate with your own goals and dreams? How do you coerce something potentially transcendently meaningful even within this cynical view?
In "Using ChatGPT Custom Instructions for Fun and Profit"
describes how to use a feature called Custom Instructions.In short, if you pay for ChatGPT Plus, you have access to a “Custom Instructions” settings panel when you click on your user name in the lower left hand corner. You are presented with two free-text boxes, each 1500 characters, to answer the following questions: “What would you like ChatGPT to know about you to provide better responses?” and “How would you like ChatGPT to respond?” In these boxes, you can specify details about yourself, including your hopes, dreams and personal goals, and perhaps how you’d like ChatGPT to keep you on track with those aims in its responses. Once filled out, your responses become a part of the “context window” (basically the working memory of the chatbot) for every subsequent interaction when it considers how to respond to your queries.
I had filled out my own custom instructions prior to making my “Rabbi Nachman” query above—in other words, ChatGPT had unique information about my personal situation and goals “in mind” when it was formulating a response, effectively finding a way to meaningfully connect my custom instructions to Rabbi Nachman’s teachings.
Custom instructions unlock a whole array of more personalized, deeper and potentially more meaningful interactions with ChatGPT. Here are a few other “meaning machine” use cases that take advantage of ChatGPT’s ability to “triangulate” on meaning:
Dream interpretation - ask ChatGPT to be a Jungian dream analysis and interpret the details of your dream. After the analysis, follow up by describing a particular situation in your life that comes to mind and see how it continues to connect your situation to dream symbolism. Notice what comes up as you reflect on the output!5
Cover letters - in your prompt, provide a link to a job description, attach your resume, and ask for a cover letter with any particular details you’d like to be addressed or tone you’d like to strike. Watch how it connects the details in your resume to the role requirements. You could probably do something similar for grant writing, fundraising, or any other scenario requiring you to connect your or your organization's capabilities or desires to what’s on offer from an outside source.
Book recommendations - mention a few books that you’re interested in, and the topic you’d like to read about or type of book you’d like recommended. ChatGPT will take these data points, along with your custom instructions, to zero in on recommendations that will likely be shockingly well-aimed at your interests.
Gifts - describe the person you’re shopping for, the things they’re into, maybe the type of gifts you’re already thinking about or sentiments you want conveyed, and continue the conversation to zero in on a perfectly meaningful gift.
***
Torah works like this too. When you engage in a holy text, alone or with others, associations between the text, imagery, other sources and even current events often fly freely; this process is inherent to the way that we make meaning in Judaism. There is something holy and important there that shouldn’t be totally conflated with ChatGPT, but it’s impossible not to notice how both systems make meaning through pattern recognition and associative, symbolic thinking.
Rebbe Nachman famously told stories, condensing many layers of spiritual wisdom into rich, dream-like sequences that have been interpreted in countless ways to draw out insight. Zen Koans, in their inversion of rational thought, similarly provide insight by seemingly “breaking” the way our minds naturally try to find make sense of language. Tarot cards, the I Ching, and every other method we’ve developed to reflect our own minds and consciousness back at us function as “meaning machines.”
Except now, instead of using static images or text, we’ve infused silicon with the ability to generate symbolically meaningful output and create dynamic reflections of our hopes and dreams. Human beings have manipulated sand and electrons into the potential for producing spiritual insight.
For a moment just consider how we’re literally awash in the information being transmitted all around us. Walking down a busy street, with music and movies and news and podcasts being transmitted in every direction, we are literally surrounded by energetic waves carrying that information. And we are only at the precipice of this information becoming ever more personalized, meaningful, and perhaps even spiritual.
This exchange is with a custom GPT that I created called “Jewish Insight” that I designed to be more particularly Jewish in its responses to queries. You can check it out here: https://chat.openai.com/g/g-s84DIC8Pp-jewish-insight
I recognize that this teaching is actually a bit simplistic, maybe even a bit saccharine. I think it's actually much deeper than the way it is briefly conveyed here, with perhaps with even kabbalistic sources and implications. However, even in this simplistic rendering, it struck a chord; mainly because of (1) the simple clarity of the expression, (2) the quality of the synthesized voice that rendered it, and (3) the surprisingly (and unexpected) personal touch at the end. It struck a chord because of how simple and direct it was.
See, e.g.,
I did fact check the citations to religious sources, and they appear accurate. The hallucination issue seems fairly well mitigated in GPT-4, in my limited experience.
To reinforce this point—ChatGPT might not be “correct” in its interpretation, but as purely an engine for finding the way that different ideas, concepts and symbols relate to each other, this can be an extremely valuable starting point for further reflection.