Combatting antisemitism using dynamic and impressive generative AI methods - opinion

Antisemitism, an age-old prejudice, is experiencing a new surge these days, especially in light of the Israel-Hamas war that started after the brutal attack on Israel on October 7th.

It has taken such sophisticated forms and comes from such "unbiased" sources that it often goes unrecognized, slipping into mainstream conversations under the guise of political discourse or social commentary. Even the activists who label themselves as "anti-racist" sometimes spread these prejudices instead of confronting them.

But today, with the emergence of Generative AI (GenAI,) we have a powerful new tool to detect "red flags," create and distribute content to dismantle anti-Jewish and hate rhetoric, and unlock various educational possibilities.

The new surge of antisemitism: forms and reasons 

Eight months have passed since the horrific massacre in October 2023, and the global surge in antisemitism shows no signs of abating. According to data from the World Zionist Organization, there has been an alarming 283% increase in reported antisemitic incidents in May 2024 compared to the same month last year.

This staggering rise underscores the urgent need for action to combat the growing threat of antisemitism worldwide. This new wave manifests in varied forms - from virulent online harassment to physical threats on campuses like Columbia University, which now endanger Jewish students and faculty.

This alarming trend underscores a broader societal normalization of antisemitism fueled by political tensions and widespread misinformation.

GenAI hunting racism 

GenAI can help "hunt" antisemitic narratives by scanning social media and online platforms for harmful content and identifying and flagging discriminatory language and patterns. It can also generate educational content to counteract these narratives and raise awareness. By analyzing vast amounts of data, GenAI can detect early signs of antisemitism and track its spread, enabling timely intervention and promoting inclusive, respectful discourse.

An example of such use is the Decoding Antisemitism project. This pilot initiative (2020–24), led by Dr. Matthias Jakob Becker, trained large language models to identify explicit and implicit antisemitic content and 'trends' on social media in the UK, France, and Germany.

The project analyzed over 130,000 user comments using a robust annotation system with 160 categories. This data informed the development of BERT models and various experiments with GPT-3.5, GPT-4, and Llama 2, leading to new insights. All d ta was carefully verified. The project's findings are shared with academia, politics, media, education, and security sectors to develop effective counter-strategies against antisemitism.

To illustrate, according to the researchers in the project, various subtle and implicit forms of antisemitic rhetoric found on social media platforms are characterized by coded language, puns, icons, allusions, and even semiotic markers (such as putting Israel in quotation marks or spelling "Israel" with the dollar sign instead of the letter "s": "I$rael"). For example, the term "Globalists with curls" is a veiled reference to Jews, insinuating a hidden global control and playing into long-standing antisemitic conspiracy theories about Jewish world domination.

The phrase "Someone should give Soros a "shower" refers to gas chambers in concentration camps, while Soros represents "rich and powerful Jews." Another subtle instance is the phrase "The Austrian artist was right," which refers to Hitler and the Nazi ideology.

Additionally, terms like "Zionazis" and statements such as "Israel conducts a war against humanity itself" implicitly compare Israeli policies to Nazi actions, a tactic used to demonize Israel and, by extension, Jewish people. Direc  Nazi comparisons are also evident in phrases like "Final Solution of the Palestinian question" and "Fourth Reich Rising," which explicitly provoke by drawing parallels between Israeli actions and those of Nazi Germany.

Creating and spreading content with GenAI

GenAI is a powerful tool for creating and spreading content that combats antisemitic or anti-Israeli narratives and racist rhetoric in general. It can raise awareness and empathy on a large scale, engaging different audiences and languages, all with minimal resources, in a short time and without a technical background.

During the Israel-Hamas conflict, several AI-generated campaigns went viral, some of which unfortunately propagated anti-Israeli sentiments. Instead of being upset or outraged, we can learn the underlying strategies of these campaigns to counter such narratives with similar success. A priceless example of a viral anti-Israeli campaign is the 'All Eyes o' Rafah,' initiated by a young man in Malaysia.

It solved an AI-generated image featuring neat rows of tents with snowy mountains in the background, accompanied by the text "All Eyes o" Rafah," hinting that Israel is targeting innocent and defenseless populations and contributing to the "demonization" that Israel" narrative." 

The image quickly went viral, being shared over 47 million times on platforms like Instagram,​ including by celebrities such as Dua Lipa, Lewis Hamilton, Gigi, and Bella Hadid. The success of this campaign can be attributed to several factors: firstly, the content was graphically appealing yet not overwhelming, featuring an AI-generated image with a simple design.

This simplicity made it easy for users to quickly understand and emotionally connect with the message, encouraging sharing. Additionally, the content was optimized for social media platforms, ensuring it was not blocked or restricted. The campaign achieved extensive reach and engagement by aligning with these platforms' algorithms and rules.

In response, pro-Israel activists created counter-campaigns using AI-generated images. One n-table example depicted a Hamas terrorist standing over a baby with the text, "Where were your "yes on October 7?". This mage aime" to shift the narrative and highlight the attacks on Israel. Still, it faced removal issues on Instagram, leading to a brief controversy before it was reinstated.

Another example of emotionally powerful AI-generated initiatives was "Be Their Voice,"- a campaign I "initiated together with Peres Center for Peace and Innovation and the entrepreneur Danielle Ofek (the founder of Angle Advocacy) in response to the October 7 attacks, aiming to assist in the return of Israelis kidnapped and held in Gaza. We used AI to animate the faces of the abducted children, giving them a virtual voice to appeal to their return home.

We worked with Tal Harari, Mor Biran Morg, and Yoram Reshef to create impactful videos in English and German that captured global attention. The campaign's deep ecampaign'sesonance was further echoed by their grandfather, who shared his torment and underscored the critical role of these efforts in representing his grandchildren (3.5 and 8 years old) during this harrowing time. To avoid profound fake backlash, we include in the creative the explanation that we had no other choice but to use AI to hear the silenced hostages.

The grandfather commented: "The animation vi" eos of my grandchildren hit like an arrow to the heart. We could hug them, hear their laughs, and understand their situation. It's been a month then, and since we can't hear from them, we can't have to speak up for them in any way we can."

These examples "highlight the power and risks of generative AI in modern conflicts. 

GenAI in educating against antisemitism and racism

The scale of impact that advanced technologies may bring to educational projects can not be underestimated.  GenA provides historical context, allowing users to learn about past events and figures more personally and engagingly. It also facilitates meaningful conversations, fostering a better understanding of current issues. 

GenAI offers a versatile toolkit for educating against antisemitism and racism. One effective method is creating personalized educational content, including interactive learning modules tailored to convey the history, impact, and present-day realities of antisemitism and racism. Such customization ensures that users receive content that is most relevant and engaging.

Additionally, GenAI can power chatbots designed to simulate conversations with historical figures or victims of hate crimes, a strategy that helps foster more profound empathy and understanding. Another dynamic approach is scenario-based learning, where users are presented with virtual tours and scenarios that depict real-life instances of discrimination, guiding them through the most appropriate responses and actions. Furthermore, adaptive learning systems with built-in feedback loops allow continually refining these educational models based on user interactions and outcomes, ensuring that the AI remains effective and responsive in its mission to combat hate. 

For instance, last year, together with the International March of the Living and D-ID,   I led a Holocaust commemoration project aimed at the young generation and world leaders using innovative GenAI technology to "bring back to li "e" ten Holocaust-e" heroes, women, and men through realistic videos based on historical photos and authentic testimonies.

Condu ted during the "March of the Liv "ng" events on Holocaust Remembrance Day 2023, the project aimed to make the heroic stories of ghetto fighters and partisans accessible to global leaders, youth, and future generations. These stories, narrated by well-known Israeli actors, were widely distributed via social media, international broadcasts, and public events, reaching millions.

The project was highly successful. It inspired ongoing Holocaust commemoration and reached many young people and leaders from different countries. It deepened their acquaintance and identification with the Holocaust heroes and its lessons.

Another initiative I led was the development of an educational app ahead of Holocaust Remembrance Day 2023, in collaboration with Code for Israel, the Awakener, and D-ID, to accompany youth on learning trips to Poland. The app guides young learners through the Holocaust study experience in a personalized way and features D-ID's GenAI, which animates Holocaust heroes to recount historical events.

Students could dub those original stories themselves or via text-to-speech after an experienced teacher confirms them. This was intended to establish an emotional connection to Holocaust research and enhance learning effectiveness. A successful pilot at the Ghetto Fighters' House MFighters collaboration with the March of the Living organization demonstrated significant educational and emotional impact.

Training generative AI models with the texts, research, and interviews of influential figures have tremendous educational potential for creating uniquely interactive learning experiences. Imagine Martin Luther King Jr. coming back to life straight from the pages of history, smiling at you, and sharing his wisdom and personal story. Such a model would effectively "embody" King, al "owing "users to engage in simulated conversations with him as if he were responding in real-time. Users could explore his views on the significance of establishing the State of Israel, the support of Jews in the civil rights movement, and more. 

For generative AI to have the most positive influence, we should be aware of its drawbacks. For example, as it highly relies on open-source data, such as Wikipedia, it is crucial to keep it validated and updated. Ensuring the accuracy and currency of entries is essential. In the future, GenAI models will be quoted and referenced as definitive sources of truth, making the reliability of their training data vitally crucial to the integrity of the information they provide.

GenAI technology has tremendous potential to combat Israel's delegitimization around the world. Like other tech tools, GenAI demands a high level of responsibility. The ultimate goal is to use it to promote curiosity, empathy, understanding, and tolerance. Similar to nuclear energy, which can be used to generate electricity for civilian use or as a destructive weapon, artificial intelligence is inherently neutral in ethical terms.

It is still up to us, humans, to define its future: either to promote a better and fairer world or the opposite. We must use this advanced technology responsibly, professionally, and sensitively through cross-sector collaborations with professional entities, such as social organizations, governments, and leading technology companies. 

We must remember that the fight against antisemitism is a long and complex process. GenAI must be scientifically evaluated to ensure reliable classification and detection. Antisemitism must be identified in text, images, and audiovisual forms, both explicitly and implicitly, to develop tailored counterstrategies. Let's hope we use GenAI effectively and responsibly in our fight against antisemitism and hatred for our security and a more tolerant world.

Shiran Mlamdovsky Somech is an entrepreneur specializing in ethically and responsibly integrating Generative AI with social impact and cross-sector collaborations. Shira has led various projects to promote and raise awareness of social issues such as domestic violence, education, democracy, and Holocaust commemoration in partnership with Meta, Microsoft, D-ID, March of the Living, and others. Since October 7, she also has been advocating for the release of the hostages. Some of these projects have won international awards and have been presented at esteemed institutions like the UN, NATO, and UN Women. Today, the founder of Generative AI for Good collaborates with tech companies, corporations, NGOs, and international organizations. In the era where AI is reshaping society, Shiran is committed to harnessing the power of technology to effect positive social change. Shira holds an EMBA from the Kellogg School of Management at Northwestern University.