Open vs Closed AI: A Jewish Ethical View

Introduction

The debate between open and closed-source artificial intelligence (AI) development presents one of the most pressing ethical dilemmas of our time. By examining this issue through both historical developments and Jewish ethical traditions, we can gain deeper insights into the interplay between innovation, responsibility, and moral obligation.

Historical Context: The Evolution of AI Development

The trajectory of AI development mirrors broader tensions in technological progress. Initially rooted in academic settings where knowledge-sharing was paramount, AI research followed traditional scholarly models of open publication and peer review. This approach aligns with the Jewish principle articulated in Talmud Nedarim 37a: “Just as I [God] taught you without payment, so you must teach others without payment.”

The evolution of AI development can be categorized into three key phases:

1. Academic Origins (1950s–1990s): AI research initially thrived in universities, reflecting the Jewish value of communal learning. As Pirkei Avot 4:1 teaches, “Who is wise? One who learns from every person.”
2. Commercial Emergence (1990s–2000s): The rise of commercial AI introduced tensions between open knowledge-sharing and proprietary development. This mirrors ancient debates in Jewish tradition about when knowledge should be restricted, as seen in Mishna Chagigah 2:1 regarding the teaching of esoteric wisdom.
3. Modern Landscape (2020–Present): Today’s AI ecosystem includes fully closed systems (such as those from Anthropic and OpenAI), hybrid approaches (like Meta’s AI models), and open-source initiatives (such as EleutherAI). This spectrum reflects the Jewish concept of shvil hazahav—the golden mean advocated by Maimonides, which encourages balance between extremes.

Jewish Ethical Framework

Jewish tradition offers profound insights into how we might ethically navigate AI development:

Pikuach Nefesh and Safety

The principle of pikuach nefesh (preservation of human life) is paramount in Jewish law. Rabbi Joseph B. Soloveitchik extended this concept to include preventive measures that protect human welfare. Applied to AI development, this principle raises critical questions about which approach—open or closed—better safeguards humanity.

Modern AI ethicist Stuart Russell echoes this concern when advocating for cautious AI development. This aligns with the Talmudic concept of “building a fence around the Torah” (Pirkei Avot 1:1)—taking extra precautions to prevent harm.

Transparency and Truth (Emet)

The Jewish value of emet (truth) emphasizes transparency and honesty. Rabbi Elliot Dorff argues that institutions have an ethical obligation to maintain openness when public welfare is at stake. This aligns with modern AI ethicists such as Shannon Vallor, who emphasize transparency in AI development while recognizing the need for responsible safeguards.

Collective Responsibility (Arvut)

The principle of arvut (mutual responsibility) teaches that all Jews are responsible for one another. Rabbi Jonathan Sacks expanded this concept to encompass global human responsibility, urging a moral obligation to protect and uplift all people. This relates to Timnit Gebru’s advocacy for ensuring AI serves diverse communities and Meredith Whittaker’s concerns about concentrated power in AI development.

Practical Synthesis: Jewish Wisdom Meets AI Development

The intersection of Jewish ethics and AI development suggests several key principles:

1. Graduated Access: Following the Talmudic model of teaching knowledge at different levels based on readiness and ethical responsibility, AI development should maintain varying levels of openness based on capability and risk. This aligns with modern proposals for tiered access systems.
2. Communal Oversight: The Jewish emphasis on communal responsibility suggests the need for broad stakeholder input in AI governance. This resonates with Jack Clark’s proposals for controlled transparency in AI development.
3. Balancing Innovation and Safety: Just as Jewish law balances competing principles, AI development must balance open innovation with responsible constraints. This aligns with Luciano Floridi’s argument that different degrees of openness serve different ethical purposes.

Future Considerations

Looking forward, several key areas emerge where Jewish ethics can inform AI development:

1. Global Equity: The Jewish emphasis on tzedek (justice) compels us to consider how AI development affects global power dynamics, as highlighted in Kate Crawford’s research on AI’s societal impact.
2. Knowledge Sharing vs. Responsible Stewardship: The Jewish obligation to teach and disseminate knowledge must be balanced with the ethical imperative to prevent harm. This reflects both traditional Jewish values and contemporary AI safety concerns.
3. Ethical Governance: New frameworks for AI oversight should incorporate both the wisdom of traditional ethical systems and modern technical expertise. Religious and ethical traditions have long served as moral compasses for societal governance, and they can play a crucial role in shaping AI policy as well.

Conclusion

The Jewish sage Hillel’s famous ethical inquiry remains profoundly relevant: “If I am not for myself, who will be for me? But if I am only for myself, what am I? And if not now, when?” These questions challenge us to develop AI in ways that serve both individual and collective interests, requiring immediate action while maintaining ethical vigilance.

The path forward is not a binary choice between open and closed development but the creation of nuanced frameworks that integrate the best aspects of both, guided by timeless ethical principles. As we navigate this complex landscape, the synthesis of ancient wisdom and modern technological understanding can help us ensure AI serves humanity responsibly and justly.

References

• Talmud Bavli: Tractates Nedarim and Chagigah
• Maimonides: Mishnah  Torah
• Soloveitchik, J.B.: The Lonely Man of Faith
• Sacks, J.: To Heal a Fractured World
• Dorff, E.N.: Love Your Neighbor and Yourself
• Russell, S.: Human Compatible
• Floridi, L.: The Ethics of Information
• Crawford, K.: Atlas of AI