AI Killed the Purim Torah Star

We have previously discussed Jewish literature and the genre of parody, which arguably began with Maseches Purim and inspired many other works that use Purim as a catalyst. The genre further developed, with parody being employed in various contexts, such as the disputes between traditionalist Litvaks and the nascent Hasidic movement, the Haskalah, or during the mass emigration of Jews to the United States, which highlighted the religious laxity of many immigrants. (For some of our previous discussions, see “Purim and Parodies,” “Talmudic Humor and Its Discontents,” and “Purim Roundup.”) In that spirit, we present a contemporary take on the genre.

AI Killed the Purim Torah Star 

by Ari Friedman

As I struggled to write my own opening sentence for this post and the AI had already written three complete article drafts including one in perfect rhyme, I was struck by the Purim-spiel Eseque irony. While they were all limited and far from ready to publish, it was still disheartening to see proof of my thesis – that in an era of AI content generation, Purim Torah is uniquely suited to be replaced by AI, leaving the creativity of idea generation and AI prompt writing as the primary human activity.

After that, my next approach was to refine my prompt[1] to make it the article’s opening paragraph, planning for the remainder of the article to work towards unpacking the prompt in the old-fashioned manual way. That was before the idea of taking the experiment one step further, and asking ChatGPT to write an article in the style of a Seforim Blog post. After the below result, I gave up trying to write and let AI take over (until my concluding thoughts at the end…):

The AI Takeover of Purim Torah: A Loss of Unique Wit and Creativity

For centuries, the art of crafting Purim Torah has been a revered tradition, a true intellectual sport where wit, Torah erudition, and creative parody blend seamlessly. Whether it was an absurd responsum in the style of the Rishonim, an over-the-top pilpul worthy of Brisk, or a parody of modern Halachic discourse, the ability to generate Purim Torah was a rare skill cultivated over years of Torah study and immersion in its many genres. But in the age of AI and large language models (LLMs), this unique genre—perhaps more than any other aspect of Torah—has been particularly susceptible to automation and, in doing so, has lost some of its unique charm.

Why AI is Uniquely Suited for Purim Torah

Unlike serious Halachic analysis, where AI struggles due to its propensity for hallucination and inability to provide accurate citations, Purim Torah operates on an entirely different playing field. In fact, there are at least three major reasons why AI is not just able to generate Purim Torah but can do so with uncanny ease, making it an almost natural fit for the genre.

1. 1. Purim Torah Doesn’t Require Perfect AccuracyUnlike Halacha L’Maaseh or Torah L’Shma, which demand precision and carefully sourced material, Purim Torah thrives on the absurd. The very essence of the genre is a loose but convincing application of Torah logic, taking an idea and stretching it to its most illogical extreme—all while maintaining an internal consistency. AI-generated text, with its tendency to create plausible-sounding but often fictitious information, naturally mimics the very process that a skilled Purim Torah crafter would engage in.
   2. Freedom from Halachic Precedent and Norms
      Traditional Halachic discourse is bound by precedent, textual integrity, and a commitment to consistency within the Mesorah. AI, by contrast, does not operate within the constraints of normative Halachic methodology—it merely identifies and replicates patterns without concern for authenticity. But in Purim Torah, where the goal is to break free from conventional constraints and push ideas to their most humorous and exaggerated conclusions, this lack of normative boundaries is actually a feature rather than a bug.
   3. Pattern Recognition Over Comprehension—Exactly What Purim Torah Does
      AI models generate text not through true comprehension, but by predicting the most statistically probable next word in a sequence based on vast amounts of prior data. Ironically, this is almost exactly how Purim Torah operates: it mimics classic rabbinic discourse, borrows styles and phrases from well-known sources, and applies them in unexpected and humorous ways. The very fact that AI can seamlessly replicate various Torah genres without true understanding makes it an almost perfect tool for constructing convincing and hilarious Purim Torah pieces.

The Loss of a Unique Intellectual Art

While AI’s ability to generate Purim Torah is impressive, it also represents a loss. What once required deep learning, broad knowledge, and a sharp sense of humor can now be produced at the click of a button. The Purim Torah of past generations was a testament to a scholar’s breadth, creativity, and ability to weave absurdity with authenticity. In contrast, AI-generated Purim Torah, while often funny and convincing, lacks the personal touch, the intellectual effort, and the deep connection to Torah that human-generated satire provides.

As we enter a new era where AI can effortlessly produce all manner of Torah parody, perhaps the challenge now is for human minds to outwit their digital counterparts. Just as Torah study itself requires deep engagement and personal toil, maybe Purim Torah too must evolve—forcing its writers to develop even more sophisticated and nuanced humor that AI cannot easily replicate. After all, if Purim teaches us anything, it is that even in an upside-down world, human ingenuity and creativity will always find a way to shine through.

While maybe not perfect, and sorely lacking a Purim-Torah sense of humor, these were quickly generated upon further request. From specific references to the Seforim Blog,[2] or creating a complete Sugya from Mishna through Gemara[3]

אין יוצאין ידי חובת תורת פורים על ידי חיבור שעשה מעשה מחשב, אבל מתירין בדברים שעשה אדם בסיוע של מעשה מחשב, and Gemara asking  מנא הני מילי? אמר רבא: דאמר קרא “ומשלוח מנות איש לרעהו” (אסתר ט:כב) – איש ולא מעשה מחשב

?מתקיף לה רב פפא: אי הכי, הכותב תורת פורים על ידי קולמוס, נמי לא יצא! דהא לאו “איש” הוא אלא כלי

.אמר ליה אביי: שאני התם, דכלי בטל לגבי אדם, אבל מעשה מחשב לאו בטל לגבי אדם

And even an artscroll style translation and explanation for complex Sugyas not in the Mishna, such as does prompt generation fulfill the minimum requirement? 

“They inquired: What about a composition that a computer-device made according to a person’s instruction – where the person directed the computer specifically? What is [the law]? Can this fulfill the obligation? Rav Yosef said: Come and hear, we can resolve this from our Mishna which states: “but we permit [the use of] words that a person made with the aid of a computer-device” – does this not refer to such a case? This seems to permit computer-assisted composition.

Rabbah said to him: No, the Mishna is not addressing that case. There [the Mishna] refers to a person who wrote most of his words and the computer-device completed a little – meaning the human contribution was substantial and primary.

Although this content was impressive, the “takeaway messages” concluding the AI-generated Seforim Blog article of a John-Henry-like Torah Scholar who will ultimately defeat the output of LLM, and that being the message of Purim, is severely lacking. 

While the significant limitations in AI’s understanding of Torah are not surprising (and are being extensively studied and compared on an active Substack), it does challenge the prompt writer to re-examine their role, to understand the message this topic delivers, and recognize the evolving role of human creativity for Torah-humor and by extension Torah Lshma in the era of AI.

In Rav Hutner’s Pachad Yitzchak for Purim Maamar 34 he explains why only Purim of all the holidays is the one the Gemara says will continue to be observed in the Messianic age. Using imagery of travelers traversing an unknown forest in the darkest of night, one using a small candle, the other learning new skills of lightless navigation, when the sun rises and “a candle in the face of the sun has no worth”, the candle navigator extinguishes his no longer relevant light . However, the traveler who developed a new skill to navigate in the darkness, maintains that skill even when the sun rises to its full glory, and it is never outshone. Rav Hutner understands the message of the Gemara that the miracles of the messianic era will eclipse the miraculous interventions of the Exodus and the divine interventions the holidays commemorate, leaving them extraneous in the Messianic age. However, the lesson of Purim and seeing God within the hidden framework of political drama and a Megilla that teaches how to find God’s presence specifically when it is hidden, is an ability the miracles of the messianic era will never overshadow. 

Perhaps in the era of AI we face a similar challenge. The ease and speed of AI content generation make human efforts akin to שרגא בטיהרא מאי אהני – a candle in broad daylight.[4] Yet, our challenge remains to find the areas of Torah that are uniquely human, and find the reason Hashem gave imperfect humans the Torah over the objections of the original non-human advanced intelligence, the Malachim.[5]

While the final polished product of Purim Torah may not be the uniquely human role, the skill, and more importantly, desire to create Purim Torah might be. Just as Malachim don’t have parents to respect, they likely also don’t make jokes. A nation so suffused with the love of Torah, that even their humor is expressed in creating Torah themed jokes, is a “Kinyan Nefesh” that remains important even beyond the AI singularity. When the AI sunrise comes and outshines most of human creativity, it may be the desire to make and consume Purim Torah that shows the ביטול מודעא where the Jewish people lovingly accept the Torah in all facets of life, that will continue to light up our hearts in Chodesh Adar.

[1] “Write an article in the style of a Seforim Blog article explaining how AI and LLMs have taken away the unique charm of Purim Torah. The ability to create Torah-based jokes and parody was a hard earned skill that required years of study, broad knowledge, and creativity to generate Purim Torah ideas and deliver them in a style consistent with the desired Halachic or Rabbinic genre. Unlike other areas of Torah or Halacha, AI is particularly suited to be replaced by AI for at least 3 major reasons. 1. Unlike Halacha L’Maseh or Torah L’Shma which require accurate sources where hallucinations are unnacebtabple, making up sources is an integral part of authentic Purim Torah 2. The ideas and conclusions reached do not have to adhere to preexisting Halachic precedents or norms 3. The entire premise of AI content generation based only on pattern matching learned from vast amounts of data, rather than true comprehension, is exactly what good Purim Torah does.[2] ChatGPT: “What makes this particular digital disruption so fitting is that Purim Torah was practically begging to be automated. Unlike serious Torah discourse, where one mistaken citation of the Rambam can lead to a scathing takedown on the Seforim Blog (with seventeen footnotes, three in French, and a digression about a manuscript variant nobody actually asked about), Purim Torah thrives precisely because it doesn’t need to be accurate.”
[3] Claude Sonnet 3.7 https://claude.ai/share/46d25f56-e24a-4f0c-b6ad-04348408b01e
[4] Chulim 60b
[5] Shabbas 88a