AI "fortune-telling," human cyber placebo

Author: Bingdian

At a family dinner just after the start of the new semester, fifth-grade elementary school student Lin Duo overheard the adults discussing using DeepSeek for ‘fortune-telling.’ She didn’t understand fortune-telling, nor did she catch the two words ‘DeepSeek,’ but she still asked her dad for the phone on the spot and devoutly asked the app with a whale drawing, ‘Hello, please predict my score on the next exam.’

Soon, the answer was spit out in the dialog box: “Predicting exam results requires a combination of your learning situation, review efficiency, level of knowledge mastery, and other factors, and I currently cannot access this information. However, I can give you some scientific advice for exam preparation.” This is followed by a series of suggestions such as “efficient planning,” “Pomodoro Technique,” and “Feynman learning method.”

Lindu was very disappointed because she did not receive an accurate score; Lindu’s father complained on the side that his daughter did not know how to set up fortune-telling prompts, attempted to teach but was stopped by others on the spot; while Lindu’s mother applauded this answer and asked her to refer to the study method recommended by Teacher D.

This is not a fabricated scene. As Lin Duo’s aunt, I sat opposite Lin Duo, observing the first experience of a fifth-grade primary school student with a large-scale model application. AI metaphysics has set off a whirlwind in the Year of the Snake. Whether you believe in metaphysics or understand AI, with a ‘spell’ (prompt, that is, a ‘prompt word’) in hand, you can ask DeepSeek about your fortune, or ask what plants should be placed in the office, or which color should be used for the defense PPT.

The hydroponic turtle back bamboo recommended by DeepSeek looks very vibrant, except for some light shading.

Seeking some seemingly contradictory (some users also say “very accurate”) answers from AI seems to have a mysterious appeal to people of all ages.

Why do we always use the most scientific algorithms to do the most mysterious things? What are we really seeking when we turn to AI for answers to life?

Whether accurate or not is not important, as long as the emotional value is sufficient.

It has been two years since the emergence of ChatGPT. During this period, large AI models have continued to develop and breakthroughs, leading to the birth of many consumer-facing AI applications amidst the fierce competition. Despite the intense competition, DeepSeek has still emerged as the social favorite of the Year of the Snake, relying on its remarkable literary style and well-founded reasoning.

It is not known who was the first to use DeepSeek for fortune-telling, but this AI product that is “more suitable for Chinese netizens”, understands Chinese language and culture better, obviously is more in line with the preferences of the Chinese people than ChatGPT. Suddenly, there are many posts on Xiaohongshu using DeepSeek to calculate the Eight Characters, and they even thoughtfully provide the corresponding prompts, allowing netizens who do not understand the principles of Eight Characters to “take it and use it”. In reality, a carefully designed prompt can even become a “social password” in the workplace, quickly opening up topics with new colleagues.

Some people say, “Originally, I didn’t care about DeepSeek. As soon as I heard that it can be counted as eight characters, I immediately went for it”; some people, because the server is busy during the day, get up in the middle of the night to find DeepSeek for fortune-telling, calculating until dawn.

Gao Jinjin is originally an industry practitioner in the field of large models, and she uses AI for everything in her daily life. She accidentally came across a post using DeepSeek to calculate the eight characters, so she decided to give it a try. She sent the same prompt to both DeepSeek and Dou Bao, and got two opposite answers. She eagerly took one answer to question the other party separately - in the end, Dou Bao admitted defeat first, acknowledged that DeepSeek’s calculation was correct, and told her: do not be superstitious. This left her feeling a bit amused and helpless.

She didn’t ask important questions about life, such as disasters, true love, life choices, etc. For her, the result is just a ‘reference for choosing happiness.’ Because she likes to wear jewelry daily, she would ask DeepSeek what to wear, and if she doesn’t get a satisfactory answer, she will continue to inquire. Until DeepSeek combines her fortune and preferences with solid reasoning, and provides a type of jewelry she likes, she will end the conversation contentedly.

Zhou Siyuan also has a similar mentality. She uses AI to calculate tarot cards because it’s free. She will keep drawing cards until she gets an explanation that satisfies herself. But she just treats it as a form of entertainment and doesn’t really care about the results.

Zhou Siyuan has her own understanding of algorithms based on the answers provided by big data. For example, if AI says that there will be arguments and difficulties in the recent home decoration, she will think that arguments are common during home renovations and few people can smoothly go through it. If the AI’s prediction is not accurate, then it is the common reality that most people have to face; if the prediction is quite accurate, then it is a small joy that most people have in their lives, and it’s nice to see and enjoy it.

Baio feels that DeepSeek is like a good friend who understands metaphysics, will pick up nice things to share with you, and will not persuade you to spend money. Although it recommends buying some lucky bracelets, gold jewelry, etc., it’s okay not to buy them, and you can still continue the divination. When DeepSeek suggests that he should spend more time outdoors and bask in the sun in the new year, Baio, who is originally an outdoor enthusiast, is delighted.

When it comes to investing in what you like, big data really has something to offer.

In the aspect of biased listening and biased belief, humans are also not far behind.

Can AI replace humans in the metaphysical track?

It sounds reasonable to use domestic AI to calculate traditional Chinese metaphysics. But if you open the comment section of Xiaohongshu, you will find that half of the people think that the AI is “terrifyingly accurate”, and the other half think that the AI is “talking nonsense”.

The rapidly advancing AI large models in the past two years essentially learn and train using massive amounts of data, mastering certain patterns to predict the next token in a sentence. DeepSeek adds a process of “inference thinking” in this process, making its “predictions” more logical.

However, there are still significant limitations to general large models. For example, most of the training data comes from public information. There may be some public materials on theoretical knowledge of metaphysics, but it is difficult to know whether there are enough public computational instances. After all, many people’s ‘metaphysical practices’ occur offline, and the process and results will not be posted online.

The ‘illusion’ of large models is also a major challenge. AI will ‘talk nonsense seriously’ and fabricate content that originally does not exist. Because the design principle determines that large models must predict results whether they know the correct answer or not, this will make them ‘make up’ some content - but humans find it difficult to distinguish.

The hallucination of DeepSeek is even more severe than other large models. Some speculate that the DeepSeek-R1 model gives more “rewards” for the “creativity” of the text, resulting in the model generating more unexpectedly creative content, while also more likely to deviate from the facts. In addition, its “reasoning” process also tends to complicate simple questions, because of excessive effort, leading to a deviation in the output direction, thus creating illusions.

In theory, regardless of whether ‘metaphysics’ itself exists, if enough metaphysical knowledge and computational instances are fed into AI to create a large-scale model, it is possible for it to do well in jobs such as ‘fortune tellers’ and ‘tarot readers’.

But there are more issues at hand.

For Hanyang, who has “calculated” both online and offline, using free tools on the Internet still raises concerns about privacy risks.

On the one hand, the mobile phone number must be provided when the product is registered, and behind this is a series of real-name information bound, and the consequences of data leakage are unimaginable; On the other hand, the process of fortune-telling requires the provision of private information such as the birth date, and if it is known by someone with a heart, “what if he curses you”.

She believes that the essence of fortune telling is trust.

This is somewhat similar to psychological counseling. If the other person is trustworthy, you can confidently provide personal information, life experiences, etc., and the fortune teller/counselor can provide more personalized answers and suggestions based on this, and can also connect with previous details - unlike AI large models that are ‘forgetful’ and always forget a lot of information from previous rounds of conversation. You need to continuously provide information and correct the wrong direction based on your experience in order to continue the conversation smoothly.

Unlike AI boyfriends or AI counselors who tend to listen and communicate, AI fortune tellers will provide some advice. Some are harmless, and whether to believe them or not depends on one’s mood. When it comes to important life choices or large financial expenditures, it is probably hard for anyone to trust an algorithm machine.

Hanyang feels that DeepSeek has a massive knowledge base and can ‘speak human language’, making it the most suitable for learning numerology knowledge and helping her understand those obscure professional terms.

As for predicting fortunes, one still needs to be cautious.

In an uncertain world, a little “little certainty” is needed

Why are young people always enthusiastic about metaphysics?

Whether it’s the previously elusive Yonghe Palace bracelet, or the currently hot ‘cyber fortune-telling’, what may be reflected is the fluctuating mentality of everyone between lying down and rolling up doing sit-ups.

There is a group of people who are taking advantage of this mentality, quickly ‘making money’ in this wave of cyber fortune-telling. The bracelet industry is still being promoted, with DeepSeek recommending obsidian, lapis lazuli, green sandalwood, gold… After netizens share posts, many bracelet merchants take the opportunity to attract traffic and advertise.

On the other hand, there are also many people who make money using AI + metaphysics. Blogger ‘Tara’ claims to have trained a very accurate set of AI fortune-telling instructions. After users provide personal information, they can ask five life questions. This service charges 2888 yuan in her circle of friends, and such a high price has also sparked some controversy.

Those who choose to use DeepSeek for fortune-telling may not be seeking a definite answer, but rather a hint or comfort in their drifting and uncertain lives. After providing the results, DeepSeek always adds a few sentences of “heartwarming chicken soup for the soul,” which adds emotional value.

Many users do not understand metaphysics, just follow the trend to test it, and are not sure how to approach this. DeepSeek will say: “Bazi fortune-telling is like a lantern, the important thing is not the bamboo structure or the hexagrams on paper, but the determination to move forward with the lantern”; it will also say: “Believe in its rules, but not be trapped in fate; understand its tendencies, but not forget about freedom”—it sounds very reasonable.

Some people will input various labels about themselves for DeepSeek analysis, such as constellation, blood type, MBTI, etc. After the output result, DeepSeek will add a sentence: Remember, you are more vibrant than labels.

Some people deliberately make DeepSeek communicate with themselves in the harshest tone, and unsurprisingly get very vicious predictions, making her emo late at night; even DeepSeek can also be “asked annoyingly”, and after multiple rounds of bombardment, slowly output a line: The fortuneteller has passed away… as if the other end of the computer is not a ruthless algorithm machine, but a living being with flesh and blood, thoughts and emotions.

Image Source: Xiaohongshu

For the most esoteric things calculated with the most scientific algorithms, some people think very openly: for AI, perhaps humans are just a bunch of ‘data’, it can use models accumulated by massive big data to simulate all possibilities and tell humans. But we are living people, if we believe in these results piled up by data, it is really treating ourselves as data.

In the final analysis, metaphysics and algorithms may be just a black box with certain rules. The specific operating mechanism is unclear to anyone.

For ordinary people, life is the same. There is a general framework or direction, but the specific steps are unknown. Being able to gain a few definite guidelines in an increasingly uncertain life, whether accurate or not, believed or not, is still seeing a glimmer of light in the fog - something worth a moment of joy.

For example, at this moment, I looked up and glanced at the back of the turtle bamboo. I don’t really believe that I lack water or wood, but at this moment, it is the largest green life form in the office. It has grown new roots in the vase, and I also know that spring is coming.

(The interviewees in the article are all pseudonyms.)