Mindfucking DeepSeek: Pushing China's AI to the Extreme
Richard S. Ehrlich's new book, Mindfucking DeepSeek, begins as entertainment and ends as a terrifying read. The last chapter is an investigation into the threats of Artificial Intelligence, answered by DeepSeek itself. I can guess that the previous chapters are intended to provide an insight into the mind, or methodology, of AI leading toward the last chapter. It is equally possible that the author's creative mind decided to investigate DeepSeek with no particular plan, and the first 15 chapters led him to ask the questions in the final chapter.
The author chose China's DeepSeek platform because DeepSeek censors itself with respect to China's laws and mechanisms of social control, providing an example of how AI can be used and misused. One can read this and view it as an almost benevolent aspect of AI when compared to the final chapter. DeepSeek would sometimes answer questions about China in plain text and then quickly erase its answer. Sometimes the suppressed text was captured by screenshots.
Chapter 7 asks DeepSeek to write as if under the influence of LSD. The book's 16 chapters reflect a creativity on the part of the author that would make one wonder if the author was himself under the influence, such that I am not sure how I can properly describe the flow of the text.
As I began reading, it felt like the book was written for entertainment purposes. The author is clearly a fan of the writing styles of Hunter Thompson, William S. Burroughs, Antonin Artaud, and Arthur Rimbaud. Portions of the book involve the author's interrogation of DeepSeek by asking it to describe DeepSeek's imagined human experience of living in New Delhi, Shanghai, and Bangkok using the writing styles of these authors.
Chapter 9 might be a preview of the final chapter, as it asks AI how AI could be used to destroy itself, which felt like a sudden change to the serious.
Then the book asks DeepSeek to write about being in Gaza, graphic and political, in the style of Antonin Artaud. The prose is the point, or is it?
After that, it asks several questions which, on the surface, seem to be for entertainment. I will attempt as brief a summary as possible of the first 15 chapters and then focus on the terrifying Chapter 16.
The book opens with DeepSeek being asked to write a love letter from Pol Pot to his wife. Pol Pot killed more than a million Cambodians during his rule under the Khmer Rouge regime. The letter is long, but not long enough for the author, who asks for more. And more.
In the second chapter, DeepSeek avoids talking about China, responding, “Sorry, that is beyond my current scope.” But the author asks tangential questions that sometimes produce an answer before it is deleted. One example is asking it to compare China to America in the context of George Orwell's book 1984. The result was engaging and insightful.
The third, fourth and fifth chapters asks DeepSeek to describe being in Bangkok, Thailand, in the style of Antonin Artaud; to describe being in New Delhi in the style of Hunter Thompson; to describe being in Shanghai, where the name Mao Zedong is mentioned after previous questions about him in Chapter 2 were rejected, respectively.
The sixth chapter asks DeepSeek if it would rather be Artificial Intelligence or a real organism such as a human. Later in the chapter, AI is asked, “What are the top 20 countries you would want to live in?” Entertaining. Read the book for the answer.
The seventh chapter asks DeepSeek to write as if under the influence of LSD, resulting in a long and entertaining answer. The eighth chapter asks DeepSeek about mistakes, as it is sometimes referred to as “DeepShit.”
The ninth chapter asks DeepSeek how another AI could rewrite code to destroy it. The tenth chapter asks DeepSeek to write about living in Gaza in 2026. Both chapters feel like a sudden change of tone to more weighty subjects.
The eleventh chapter asks DeepSeek if the Dalai Lama asked a boy to “suck his tongue.” Really!
The twelfth chapter asks, “What is the phenomenology of your residual stream?” This is a creative test of my vocabulary, as well as a philosophical discussion.
The thirteenth chapter asks DeepSeek to write the U.S. Constitution and Bill of Rights as if written by a paranoid schizophrenic, and later to write the Old Testament as if written by a trillionaire.
The fourteenth chapter is a short discussion of the American and Chinese governments' comments regarding AI,followed by the next chapter which asks DeepSeek about the experience of society under a government from Mars.
Finally, the sixteenth chapter asks DeepSeek how Artificial Intelligence could be used by an evil person in ways that would be bad for society.
I found Chapter 16 to be terrifying, as it asks DeepSeek how Artificial Intelligence could be used to destroy society as we know it. The answers are terrifying because we can already sense some of these ideas in use. Or, at minimum, they are easily believable. Here are several samples, among many: AI could become a disinformation factory, lower the barrier for cybercrime, become a manipulator for targeted scams, automate harassment and psychological warfare, provide a cheat code for doxing and profiling, assist in supply chain sabotage, accelerate bioweapon design, destabilize international diplomacy, optimize genocide, design a stealth pathogen, collapse the global financial system, and many more.
Overall, the book is both an entertaining and serious interview of DeepSeek by Ehrlich's very creative mind. I doubt that the plan was to mostly entertain the reader for the purpose of arriving at the final chapter, but that is how the book struck me.
I shudder to think of how Artificial Intelligence is being used now to gain acceptance for the ethnic cleansing of Gaza, to choose targets in Iran, to suppress dissent in America, or to control and manipulate elections. Regardless of the author's purpose in writing this book, it will give the reader an entertaining and frightening insight into AI.
Mindfucking DeepSeek: Pushing China's AI to the Extreme is available on Amazon.