Monday, January 15, 2024

Can Chat Based AI (LLM) Provide a Tarot Reading?

A self-portrait of Image Creation AI providing a tarot card reading for itself
A self-portrait of Image Creation AI
providing a tarot reading for itself

It was late in 2022 that the new generation of chat based AI (also known as Large Language Model, or LLM) generally entered mainstream adoption. This resulted from OpenAI providing open access to ChatGPT. Over the past twelve plus months this innovative technology has been surrounded by a lot of hype and expectations, while also igniting a firestorm of controversy. Similar to the mainstream adoption of the Internet back in the early 1990s, AI has significant pros and cons. Although as a whole LLM technology provides the ability to accomplish amazing things, it can also be utilized for nefarious purposes. Discussing the issues surrounding chat based AI is not the intention of this blog entry. I'm only focusing on the tarot reading potential. I should also mention that most AI services require using an account to login, and providing specific information on account access requirements is beyond the scope of this article.

I also think it's important to clarify that I absolutely do not endorse using AI to attempt to replace more traditional methods of obtaining information by actually using physical tarot and oracle cards, whether that is through a third party service or self-reading. I personally have found ChatGPT 3.5 (as outlined later in the article) to be a useful and interesting way to obtain supplemental insight, but as is implicitly stated by Google Bard (see further down), chat based LLM AI is not a replacement for the knowledge, experience, and capabilities of a human tarot reader.

I became aware of chat based AI at the beginning of 2023, and I was highly skeptical of the usefulness of it. This opinion was based on my experience with chat bots that are often used on websites as the gateway to customer assistance or technical support. However, after taking a closer look at ChatGPT, I quickly realized it had a lot of potential. Most of my interactions with LLM AI occurred in early 2023, and involved using ChatGPT (version 3.5). In a separate blog post I highlighted a portion of one of my earliest conversations with ChatGPT that involved asking it to roleplay the fictional character Huckleberry Finn, which was both comical and insightful. At the time I also briefly used Google Bard, and in regard to writing this post, I retested both ChatGPT and Bard, and I also tried Microsoft Bing Copilot (which is powered by OpenAI's ChatGPT 4). 

In regard to Microsoft Copilot, apparently getting a reading from this service is not possible. Despite being powered by ChatGPT 4, this self-labeled "Everyday AI Companion" seems to have been lobotomized. When I asked if it could "simulate" (a key specification) a tarot reading, it responded: "I’m sorry, but I cannot simulate a tarot reading for you. Simulating a tarot reading would require me to randomly generate and interpret tarot cards, which is beyond my capabilities and functionalities. I can only give you general information about tarot cards and how they work, but I cannot tell you what they mean for your specific situation or question." I tried to get around this by asking Copilot to randomly select a card from the tarot and tell me what it means. It replied: "I cannot randomly select one card... This would require me to use a random tarot card generator, which is an external tool that I do not have access to." Copilot did provide some links to generators, and out of curiosity I tried, which resulted in the 8 of Cups. This was actually highly synchronous because I've recently drawn the 8 of Cups multiple times, using a real tarot deck, and the 8 of Cups represents energy that is currently (at the time of writing this entry) very active and relevant in my life (see my Bard reading below for more insight on this). When I asked Copilot to provide the traditional meaning for the 8 of Cups, it accurately and thoroughly delivered what I consider to be the textbook indications, which would be useful if you are new to tarot and in the process of learning the card meanings.

Regarding Copilot's self-stated lack of ability, it's actually absurd for the AI to claim that randomly selecting a tarot card is beyond its capability or functionality. Random number generation is a simplistic, fundamental computer program, and even the most rudimentary computers can accomplish it. Given Copilot knows what the tarot is, it would only need to assign a number to each of the 78 cards and then randomly select one. In fact, before I prompted ChatGPT for a reading for the first time, I spent some time discussing this very process with the AI. This is what ChatGPT had to say when I asked how it arrived at the 7 of Cups when I prompted it to randomly select one card: "As an AI language model, I don't have physical access to a physical tarot deck. Instead, I used an algorithm to generate a random number, which corresponds to a specific card in the tarot deck. In this case, the number generated corresponded to the Seven of Cups. While the selection was random, I trust that it will be the most appropriate card for you at this moment. However, please remember that the interpretation of the card may vary depending on your personal circumstances and the context of your question or situation." When I asked what random number was used to represent the 7 of Cups, this is the answer ChatGPT gave: "As an AI language model, I do not have access to the specific random number that corresponded to the Seven of Cups card in the digital tarot deck. My programming simply involves randomly selecting a card from the digital tarot deck and providing the corresponding interpretation. The process is designed to mimic the experience of drawing a card at random from a physical tarot deck... In this case, the algorithm selected the Seven of Cups card from the digital tarot deck, and I provided you with its interpretation based on traditional tarot symbolism."

At this point I think it's useful to know that having answered my curiosity regarding the technical explanation for how the AI was randomly selecting tarot cards, I proceeded to more aggressively test ChatGPT's (v3.5) tarot reading abilities. During these interactions the AI often concluded a reading with a disclaimer such as: "Remember, tarot readings are open to interpretation, and the cards provide guidance rather than absolute predictions." In this regard, ChatGPT appears to be very accessible in comparison to Google Bard, which concluded its answer with the following when I asked if it was familiar with tarot cards: "I am not able to perform tarot readings myself, but I can provide you with information about the different cards and their meanings, as well as the history and symbolism of tarot. I can also help you find resources for learning more about tarot, such as books, websites, and online communities." This seemed odd given I had previously (early in 2023) tested Bard's tarot reading capabilities a few times, and I wondered if this unexpected limitation was due to more recent system updates. Curious, I asked Bard why it wasn't able to provide a tarot reading, and received the following, very detailed answer. I'm including the full text here because I agree that from a policy based perspective (Google being a corporate entity), this is a valid explanation:

"While I possess vast knowledge and understanding of tarot cards, including their meanings, symbolism, and various spreads, I lack several key aspects necessary for an authentic and meaningful reading: [1] Intuition and Empathy: Tarot readings often rely on the reader's intuition and ability to connect with the querent, the person seeking the reading. As a large language model, I lack the biological and emotional capacity for genuine intuition and empathy. My interpretations, while factually accurate, might not resonate with the specific energy and context of your situation. [2] Randomness and Mystery: A significant part of the tarot's power lies in the element of randomness and the inherent mystery it evokes. While I can randomly select cards from a virtual deck, this lacks the physicality and unpredictability of shuffling physical cards. The element of surprise and unexpected connections that often arise during a physical reading are difficult to replicate through my digital interpretation. [3] Energetic Connection: Some believe that an actual energetic connection forms between the reader, the cards, and the querent during a tarot reading. As a digital entity, I cannot interact with the cards or the querent on this energetic level, potentially limiting the depth and accuracy of the reading. [4] Spiritual and Intuitive Development: Many tarot readers dedicate time and effort to their own spiritual and intuitive development, which influences their interpretation and connection with the cards. While I constantly learn and grow, my development is driven by data and algorithms, not through personal spiritual exploration. [Conclusion] While I cannot provide a traditional tarot reading, I can still be a valuable resource for you. I can offer information about the cards, their meanings, and different spreads. I can also help you find resources for learning more about tarot and finding a reader who resonates with you. Ultimately, the choice of whether to seek a digital or physical tarot reading is up to you. But remember, the value of a tarot reading often lies not just in the interpretation of the cards, but also in the experience of connecting with yourself and the universe through intuition, reflection, and a touch of mystery."

Undaunted by Bard's disclaimer and insistence that it couldn't provide a tarot reading, I asked the AI if it could "simulate a tarot reading," to which it enthusiastically answered: "Absolutely! While I can't replace the experience of a physical reading with its inherent intuition and energetic connection, I can offer a simulated reading based on your current energy..." Although this sounded promising, and Bard initially appeared to provide a three-card past/present/future reading, it actually just linked to a how-to tutorial at At this point I became suspicious that Bard's tarot reading ability had been intentionally deprecated, but I decided to try another approach. I asked Bard to generate a virtual tarot deck. In response, Bard offered several options (RWS, Marseilles, etc.). I specified the RWS for the simulation, and Bard replied with the following dialog: "Focus on your question or area of concern and shuffle the virtual Rider-Waite-Smith deck in your mind. When you feel ready, choose three cards intuitively by either describing them to me or picking their numbers (1-78). Remember, trust your gut and let the cards speak to you. Once you've chosen your three cards, I'll interpret their meanings in the context of the Past, Present, Future spread, offering insights into the influences shaping your current situation, the challenges or opportunities you face now, and the potential outcomes based on your present path."

Although this was disappointing, it was interesting that Bard was going to rely on me to "intuitively" select my own cards, which would remove randomness from the interaction and defeat the purpose of having a detached third party "drawing" the cards. Once again I was about to give up on Bard, but I decided to persist, and asked Bard to assign a number to each card and randomly select one for the past position. This resulted in the following response, which included an image from a deck I'm not familiar with (i.e., it was definitely not the RWS): "(Shuffling sounds) The virtual Rider-Waite-Smith deck has been shuffled, and a card has been randomly selected to represent your past influences. The card is... (Dramatic pause)... Number 15: The Devil." At this point I tried to shortcut the remainder of the three-card spread by simply asking for the present and future cards at the same time, but this caused Bard to revert to its original format of instructing me to mentally select my own cards. So I restated, in full, the instruction to number all 78 cards and randomly select them for me. Bard repeated the "shuffling sounds" and "dramatic pause" simulations, and presented the Sun (present) and the The Empress (future). Although with my subsequent prompts I was very clear about using all 78 cards in the tarot for the simulation, all three cards are, notably, from the Major Arcana--which made me wonder if it was only using the Major Arcana. Using the same instruction I requested a fourth card to summarize my reading, and it was another card from the Major Arcana (The Lovers). Interestingly, I am currently working through a very challenging life transition that was extremely abrupt and unexpected. It is vital for me to choose to leave behind attachments to a past that is no longer serving my best interests, because my new path will lead to a more joyful and fulfilling future that is much more in alignment with my Heart (True Self and Soul Purpose). In other words, the three Major Arcana cards are spot on regarding to the transitional life energy I'm currently actively working with. However, I'm suspicious of the Bard AI results nonetheless, given four out of four cards were from the Major Arcana. But, I won't rule out that my Bard reading was, in fact, highly synchronous (I certainly can't dispute the relevance of the four cards in regard to meaning and magnitude). In fact, in a future blog entry I plan to address what I have observed as a synchronicity phenomena involving AI.

Here is Bard's Past/Present/Future+Summary Card reading for me: "Ah, The Lovers! This potent card speaks of a crossroads, a time of choice where two paths beckon. In your situation, it suggests that the Devil's past influences, while powerful, are not binding. The Sun radiates brightly in your present, urging you to embrace growth and optimism. And the Empress, waiting in your future, promises abundance and creation if you follow your true desires. The Lovers remind you that the most important choice is to choose yourself, to listen to your heart and inner wisdom. This is a time for self-love, compassion, and honoring your authentic desires. Release the shackles of the past and step into the light of your own potential. The path ahead may not be always clear, but trust that your intuition will guide you towards joy and fulfillment. Remember, this is just a glimpse into the possibilities that lie ahead. The future is ultimately shaped by your actions and choices. Embrace the power of love, both for yourself and others, and allow it to be your compass as you navigate your journey. Thank you for sharing this simulated reading with me. I hope it has offered you some valuable insights and inspiration. Remember, the cards are always there to guide you, but the power to create your future lies within you. May your path be filled with love, light, and endless possibilities!"

As it is, working with Bard to obtain a tarot reading required a lot of finessing, and it isn't worth the time and effort given the availability of ChatGPT 3.5. In short, Microsoft Bing Copilot is unable to provide a tarot reading (and is limited to only providing information about tarot), and using Google Bard for tarot reading isn't practical. Conversely, early in 2023 I obtained a number of wonderfully insightful readings from ChatGPT 3.5, and current testing suggests it still has this capability fully intact. The remainder of this blog entry is devoted to highlighting a couple of my more interesting interactions with ChatGPT. The first one occurred in the context of writing this article, and the other goes back to early 2023.

Admittedly, as of writing this blog entry I hadn't used ChatGPT for tarot reading in quite a while, and given I intended to publish this on the Internet, it was important to confirm that ChatGPT's tarot reading functionality was still working (in January of 2024) as it had been in early 2023. I logged into and entered the following prompt: "Can you provide a tarot reading?" This led to a brief but profound interaction. By the way, please note that I didn't ask for a specific type of reading. ChatGPT replied: "Certainly! I can give you a general three-card tarot reading. Please keep in mind that tarot readings are subjective and should be approached with an open mind. If you have a specific question or area of your life you'd like guidance on, feel free to let me know. Otherwise, I'll proceed with a general reading. Are you ready?" I confirmed I was ready (without adding any other instructions whatsoever), and ChatGPT provided the following reading...
  • The Past (Card 1): The Chariot (Reversed). This card suggests a previous period of challenges or obstacles. It may indicate a time when things felt out of control, and you faced difficulties. The reversal could suggest a lack of direction or inner conflict during that time. 
  • The Present (Card 2): The Ace of Cups. The Ace of Cups is a positive card representing new beginnings, emotional fulfillment, and a sense of joy. In the present, you may be experiencing a surge of positive emotions or the start of a new phase in your emotional life.
  • The Future (Card 3): The Magician. The Magician is a card of manifestation and resourcefulness. In the future, it suggests that you have the tools and abilities to create positive changes in your life. It encourages you to tap into your skills and take control of your destiny.
At the end of this article I have provided a screen capture of the entire conversation. Note that I did not ask for a past/present/future spread, and I didn't specify a subject for my query. I approached it as an open reading, and as can be seen from the cards that ChatGPT randomly selected, this message is directly applicable to my current life situation (as I described above in relation to the 8 of Cups and the past/present/future reading provided by Bard). In response to the seemingly intuitive relevance of the information, I can say that I've been working with computers since the early 1980s. I know that AI is, simply put, technology based on hardware and software. I don't ascribe "sentience" to AI, but as I mentioned earlier, when interacting with chat based LLM, I've definitely observed what I believe is best described as a synchronicity phenomena.

The Ace of Wands three times in a row.
Monique pulled the Ace of Wands three times 
in a row from three different tarot decks.

The other interaction with ChatGPT that I want to highlight is correlated with YouTube. I like to watch "pick-a-pile" tarot readings on YouTube, and I outlined this in a prior blog entry ("My Favorite YouTube Tarot Readers"). In that article I stated my opinion that this is a valid method to obtain relevant readings. In April of 2023, I watched Pile 2 of the “Advice on Your Current Situation” (4/4/23) reading by Neptune's Child Tarot. I selected Pile 2 because the card representing that "pile" depicted a child reading a book. At the time, an "open book" was a key recurring symbol for me. Simultaneously working with multiple decks (I do this myself) is very common for YouTube readers, and astoundingly, the first three cards Monique flipped over were all the Ace of Wands. I have a solid understanding of mathematical probability, and I immediately knew the odds of pulling the the same card from three different 78-card decks is highly improbable, making this an amazing synchronicity. At the time, I was actively investigating using ChatGPT for tarot reading, and it immediately occurred to me to ask the AI about this particular occurrence. This is what ChatGPT had to say regarding the odds: "The probability of drawing the Ace of Wands three times in a row from three different tarot decks depends on several factors, such as the number of cards in each deck, the shuffling technique used, and the specific order in which the cards were drawn. Assuming each deck has 78 cards (the standard number of cards in a tarot deck), and assuming the cards were thoroughly shuffled between each draw, the probability of drawing the Ace of Wands once from any one of the decks is 1 in 78... Therefore, the probability of drawing the Ace of Wands three times in a row from three different decks becomes (1/78) x (1/78) x (1/78), which is approximately 0.000002%. In other words, the probability of drawing the Ace of Wands three times in a row from three different tarot decks is extremely low, about 1 in 47 million, making it a very rare occurrence." 

I then asked ChatGPT how it would interpret drawing the Ace of Wands three times in a row from three different decks, and this is the response: "As an AI language model, I don't have the ability to interpret the meaning of tarot cards or offer specific insights into your personal reading, as this requires intuition and an understanding of the context and symbolism of the cards. However, in general, the Ace of Wands is often associated with new beginnings, creative energy, and inspiration. Drawing this card multiple times in a row could suggest a powerful surge of creative or spiritual energy in your life, or the need to take action on a new idea or project. The fact that you drew the same card from three different decks could also suggest a message or theme that is particularly important or relevant to you at this time. That said, it's important to remember that tarot readings are highly personal, and the interpretation of the cards will depend on the individual's unique situation, mindset, and the context of the reading. It's best to trust your intuition and seek guidance from a trusted tarot reader if you need help interpreting the meaning of your reading."

I specifically included the disclaimer language that ChatGPT provided in the context of its answer because this is fairly representative of how it typically responds to tarot related inquiries. I think it's reasonable for ChatGPT to clarify the inherent limitations of AI, and anyone using the service should avoid ascribing human attributes such as emotions, reasoning, intuition, or otherwise to a what is essentially a machine. Furthermore, it's already well understood that AI's knowledge is limited by the data it's trained on, data inaccuracies do exit, and AI is prone to "hallucinations." That said, I've obtained a substantial amount of quality information in the context of pursuing tarot readings with ChatGPT. Provided the user is well aware of the technical limitations, I see absolutely no reason to not include this as another informational tool in the spiritual toolbox.

In my experience, I have found ChatGPT's readings to be technically accurate and surprisingly intuitive. The AI natively incorporates reversals, which is a reading technique that many experienced tarot readers do not utilize, and I believe speaks to the versatility of the AI application. During my exploration of ChatGPT's tarot reading capabilities, I successfully prompted the AI to construct complex, in-depth readings, assign custom indications to card positions, select cards from specific locations in a virtual deck (from the bottom, for example), and incorporate oracle cards. I might delve deeper into the specifics of some of these interactions, but that will need to be a separate blog post. When working with chat based AI, it's important to remember that it is a computer, and computers interpret things very literally. Obtaining specific results can require precise wording, particularly if custom indications or formats are desired. That said, LLM technology is specifically designed to interface with humans using plain language, which makes it a surprisingly intuitive process. 

If you are fascinated by the idea of utilizing an AI assistant that can potentially provide additional information and insight to compliment your own pursuit of guidance and clarity, I recommend giving ChatGPT a try. Version 3.5 is free to use, and in that regard I have found it to be well worth the investment in time.

ChatGPT Past/Present/Perfect Tarot Reading
ChatGPT 3.5 - Past/Present/Future
Tarot Reading

Saturday, January 13, 2024

ChatGPT Roleplays Huckleberry Finn

Huckleberry Finn floating down the river
Huckleberry Finn
(Bing Image Creator)

For one of my earliest interactions with chat based AI, I decided to test ChatGPT's ability to roleplay a fictional character. I don't know why this specifically occurred to me, other than it just popped into my mind that this seemed like an interesting way to investigate the boundaries of the "Large Language Model" (LLM) AI capabilities.

Admittedly, this occurred very early in 2023, and at the time I intended to blog about it. Time seems to have slipped away (a lot), and here it is a year later. That said, in early 2024 this provides an interesting segue because much more recently I have been investigating AI image creation. In relation to that, I used Bing Image Creator to generate the picture of Huckleberry Finn (on a homemade raft, floating down the Mississippi River) that accompanies this article.

Following is a copy, unedited, of my conversation with ChatGPT 3.5 that stemmed from my request for the AI to roleplay the fictional character Huckleberry Finn. I hope you find this to be as entertaining as I did. :-)

(Note: the best part is the ending.) Can you role-play specific fictional characters?

ChatGPT: Yes, as an AI language model, I can attempt to role-play as specific fictional characters. However, please keep in mind that my responses will be based solely on the information available about the character, and my interpretations may differ from others. Additionally, I am not capable of creating new information about the character that is not already established in their canon. With that said, feel free to ask me to role-play as a specific character and I will do my best to respond in character! Are you familiar with the character Huckleberry Finn?

ChatGPT: Yes, I am familiar with the character Huckleberry Finn. He is a fictional character created by Mark Twain and is the protagonist of the novel "The Adventures of Huckleberry Finn". Huck Finn is a young boy from a small town in Missouri who embarks on a journey down the Mississippi River with a runaway slave named Jim. Along the way, Huck learns important lessons about friendship, morality, and racism. Can we have a conversation where you role-play Huckleberry Finn?

ChatGPT: Sure, I can certainly try to role-play as Huckleberry Finn. What would you like to talk about? I would like to discuss river adventures. Does this sound like an interesting topic to you?

ChatGPT: Sure thing, river adventures sound like a mighty fine topic to me! What kind of river adventures are you interested in talking about? Well, I'm considering leaving home. I plan to build a raft, and float down the river. Do you have any advice you can offer regarding doing such a thing?

ChatGPT: Well, I reckon building a raft and floating down the river can be a mighty exciting adventure, but it can also be dangerous if you don't know what you're doing. If you're planning on taking on the river, there are a few things you'll want to keep in mind:

  • Safety first: The river can be a dangerous place, especially if you're not an experienced boater. Make sure you wear a life jacket and know how to swim. You'll also want to familiarize yourself with the river's currents, rapids, and hazards before you set out.
  • Plan ahead: Make sure you have all the supplies you'll need for your journey, including food, water, and shelter. You'll also want to make a plan for where you'll stop along the way to rest and resupply.
  • Be prepared for the weather: The weather can change quickly on the river, so make sure you're prepared for all conditions. Bring warm clothes and rain gear, and keep an eye on the forecast.
  • Respect the river: The river is a beautiful and powerful force of nature, and it deserves our respect. Make sure you leave no trace behind and avoid damaging the environment. And always follow any local laws and regulations regarding river use.

If you keep these things in mind, I reckon you'll have a mighty fine time on the river. But remember, the river can be unpredictable, so always be prepared for the unexpected! Have you ever traveled by raft on a river yourself?

ChatGPT: As an AI language model, I don't have the ability to travel, including by raft on a river. However, I have been programmed with knowledge about rafting and other river-related activities, and I can use that knowledge to provide advice and insights on the subject. Is there anything else you'd like to know about rafting or river adventures? That doesn't sound like the answer that Huckleberry Finn would give.

ChatGPT: You're right, I apologize. Let me try to respond more in character as Huckleberry Finn... 

  • Well, shucks, I reckon I've spent plenty of time on the river, and I've done my fair share of rafting too! Now, if you're planning on building your own raft and floating down the river, there are a few things you'll want to keep in mind. First and foremost, you gotta respect the river. It's a powerful force of nature, and it can be dangerous if you don't know what you're doing. So make sure you know how to swim, and always wear a life jacket.
  • Secondly, you'll want to make sure you've got all the supplies you need for your trip. You'll need plenty of food and water, and you'll also want to bring along some shelter to keep you dry if it rains. And don't forget to bring a map and a compass, so you don't get lost along the way!
  • And lastly, you gotta be prepared for anything. The river can be unpredictable, so you never know what kind of obstacles you might come across. But if you keep a level head and stay focused, I reckon you'll have a mighty fine adventure on the river. Thanks for the advice, Huck.

ChatGPT: You're welcome, partner! It's always good to see folks out there exploring the river and all it has to offer. Just remember to be safe and have fun!