AI Sentience—Lessons from the Golem

The rights of zombies, robots, and sentient AI.

1. AI Sentience—The Debate

The Oxford Dictionary defines consciousness as “the state of being aware of and responsive to one's surroundings” or “the fact of awareness by the mind of itself and the world” ¹. Although easily recognized by all, there is arguably no phenomenon as thorny as consciousness from a scientific, philosophical, and epistemological point of view.

Despite the dramatic conceptual gains we are currently witnessing in the neurosciences, we still have virtually no inkling regarding the mechanism(s) responsible for consciousness and qualia—the subjective blueness of the sky, the taste of a strawberry, the pain and joy of childbirth—in human beings. Far less is known about the dissemination and manifestations of consciousness throughout the Earth’s biosphere and whether conscious life may have taken root elsewhere in the universe. 

The possibility of consciousness in artificial intelligence (AI) is a fascinating and controversial topic that has garnered considerable attention of late from scientists, philosophers, and the general public. Recently, a group of experts applied a battery of potential “indicator properties” of consciousness (adduced from neuropsychological studies and theory) to several existing AI systems, including GPT-4, ChatGPT, AlphaGo, and DALL-E. While the study concluded that none of these systems exhibited signs of consciousness, there were “no obvious technical barriers to building AI systems which satisfy these indicators.” ²

Science aside, the bioethical, social, legal, and religious implications of conscious AI are enormous and have galvanized influential leaders in the field to ramp up vigilance and advocate for at least temporary moratoria on AI research and development. ³ In the current article, I present a unique Torah perspective on consciousness and its possible ethical implications for interaction with sentient AI.

Specifically, I begin by summarizing the concept of “Kabbalistic panpsychism,” fleshed out in an earlier issue of Beyond Belief,  ⁴ which I believe provides a cogent framework for appreciating Judaism’s understanding of consciousness. Next, I attempt to inject lessons gleaned from Rabbinic discourse surrounding an artificial life form identified in traditional Jewish literature as the Golem into the contemporary debate surrounding AI sentience and its societal ramifications.

2. Kabbalistic Panpsychism 

In Kabbalistic Panpsychism ⁵ and a previous Beyond Belief installment, ⁴ I outlined four major categories of physical existence recognized by the Torah—the inanimate domain (Domem, Heb.), plant life (Tzomeach), animal life (Chai) and humankind (M’daber, literally “speaker”). The Kabbalah—the Jewish mystical tradition—elaborates an ontology of consciousness that is panpsychist in nature, viz., that all things Created manifest some modicum of consciousness. This assertion is based on a metaphysic that all things and events are brought into existence by the actions/interactions of ten Sefirot (forces or attributes of God) of which the top three (Keter–Chochmah–Binah or Chochmah–Binah–Da’at) impart consciousness. ^{4, 5}

I indicated that in the case of inanimate objects, consciousness informs form and substance; in vegetation, consciousness confers properties of growth and reproduction; in animal life—sentience and emotions; and in humans—self-awareness, rationality, morality, etc. (Fig. 1). As detailed previously, ^4-6 this model understands consciousness to be holographically and hierarchically organized, relativistic, and (contrary to mainstream opinions in neuroscience and philosophy) capable of downward causation. 

 Fig. 1. Kabbalistic Panpsychism. Progressive revelation of transcendent consciousness (Ohr Makif) into immanent consciousness (Ohr Pnimi) by the G”R’s of the various domains of Creation. White denotes consciousness space. G”R = Gimel Rishonot (‘enminded’ three top Sefirot); Z”T = Zayin Tachtonot (“unconscious” seven bottom Sefirot). ⁵

3. Intermediates and Outliers

In addition to the four general domains listed above, there exist “intermediate forms” and “outliers” where categorization of the state of consciousness is problematic. For example, the Etz Chaim, seminal work of the Lurianic Kabbalah (16^th century), considered corals (Almogim, Heb.) as intergrades between the inanimate world (Domem) and plant kingdom (Tzomeach). ⁷ Similarly, the Etz Chaim and the Chassidic master, Rabbi Pinchas Shapira of Koretz (1726-1791) construed “Adnei HaSadeh”—an organism described in Midrashic literature that was tethered to the ground by a cord ⁸—to be part plant (Tzomeach) and part animal (Chaye). ⁹

Situations may be envisioned where, barring any obvious precedent or analogy in Torah scholarship, novel configurations of biology and consciousness may utterly defy pre-conceived schemes at classification. Absent the latter, it may prove difficult or impossible to properly foresee and safely navigate the ethical byways that will inevitably accompany revolutionary breakthroughs in biology and cybernetics. The advent of human brain organoids in 2008 is a fine case in point. Scientists are now capable of generating pea-sized clumps of human embryonic cerebral tissue which can be studied for weeks at a time in culture dishes or after transplantation into living rodent brains.

Remarkably, the stem cells within cortical organoids differentiate into mature neurons which elaborate various neurotransmitters and their receptors, establish highly intricate synaptic networks, and, albeit the source of much controversy, exhibit electrical activity (brain waves) likened to patterns recorded from human newborn infants! ¹⁰ Few doubt the potential of such organoids to provide vital information concerning normal brain embryology and function.

A number of laboratories are already engaging cerebral organoids to study human neurodevelopmental conditions such as autism, schizophrenia, and microcephaly (small head and brain) resulting from Zika virus infection.¹⁰ But alarms have been sounded to ensure that the science does not outpace the stark moral responsibilities demanded by this line of inquiry. The overarching worry here is: given rapid refinements in the field, will human cerebral organoids achieve sentience either in vitro or after implantation and integration within host animal brains?

Are there limits to the complexity of consciousness such organoids may attain? Were they to experience pain and suffering might not such experimentation be tantamount to torture? These concerns would be the same regardless of whether the organoids generate consciousness de novo in the mainstream reductionist (materialist) sense or whether they channel consciousness—antenna-like—from an über-conscious surround as inferred by the Kabbalah and other cosmopsychist positions. ⁵ 

From a Jewish perspective, how consciousness is expressed in such intergrades is not entirely clear in so far as the latter cannot be easily pigeonholed within the Torah’s classification scheme (section 2). Sophisticated bioethics consortia have prudently been assembled in the United States and elsewhere to address these reservations moving forward. But until consciousness can be effectively diagnosed, quantified, and monitored in brain organoids and chimeras, a healthy dose of caution should remain de rigueur. ⁵

And if consciousness in these contrived biological constructs is enigmatic, can anything at all be said about potential sentience in advanced robotics and other purely artificial neural substrates? As discussed in the following section, attempts by Jewish scholars and legislators to define states of consciousness in the ersatz “life form” known as the Golem may provide novel insight and suggest an ethical and legal scaffolding with which to frame ongoing deliberations on AI sentience. 

4. Golem Consciousness

The moniker “Golem” derives from the Hebrew word, גלמי (Galmi) which first appears in the Book of Psalms ¹¹ and connotes something imperfect or unformed. Surfacing periodically over the long course of Jewish history, the Golem is essentially a humanoid (usually male) derived from clay or mud (sometimes wood). The Golem was ostensibly brought to “life” by means of sacred incantations (invoking combinations of the various Names of God) originating with the ancient Kabbalistic text, Sefer Yetzirah (The Book of Formation)¹² and recited by individuals versed in the mystical arts. The creature’s raison-d’être was chiefly to protect Jewish communities from persecution, as was the case for the famous Golem (Fig. 2A) conjured by Rabbi Yehuda Loew (the “Maharal of Prague”; 1512–1609). Occasionally, Golems were formed to serve as domestic laborers or even concubines! ⁵  

Figure. 2. Artificial Life Forms. A. The Golem of Prague. B. Lt.-Commander Data with positronic brain exposed (Paramount Domestic Television). ⁵

But are Golems conscious? The trivial panpsychist response is “yes,” on par with everything else. We’ve seen how the Kabbalah understands all things and events to consist of ten Sefirot of which the first three enable varying degrees of consciousness. But what level consciousness are we talking about in the case of a Golem? Domem (inanimate existence)? Tzomeach (plant growth)? Chaye (animal sentience)? M’daber (self-awareness)? Importantly, is the Golem—and by extension sophisticated automatons, androids and cyborgs—ever sufficiently human-like to warrant protection by basic human rights?

Although still the stuff of science fiction when it comes to robots (see Star Trek’s Data (Fig. 2B) and the 2015–2018 television series, Humans), it is not uncommon for animal activists and others to pose similar challenges regarding highly sentient organisms like chimpanzees, elephants and dolphins. In 2012, the Cambridge Declaration on Consciousness concluded that “the absence of a neocortex does not appear to preclude an organism from experiencing affective states” and that mammals, birds, and some invertebrates have the “neuroanatomical, neurochemical, and neurophysiological substrates of conscious states along with the capacity to exhibit intentional behaviors.” ¹³ 

Rabbinic analyses regarding the status of the Golem in the Creation hierarchy, as recorded in Talmudic and Kabbalistic sources, may help us in refining an ethical approach to the prospect of AI sentience. The discourse may be ancient but the concerns raised and conclusions drawn are starkly relevant to modern debate concerning robotics and AI. ¹⁴ There is fair Rabbinic consensus, if not unanimity, that despite its outward appearance and anthropoid behavior, the Golem is not human and therefore undeserving of any special moral or legal dispensations.  According to traditional Jewish sources, most Golems apparently understood simple commands —with some even opting to disobey! ¹⁵—but were incapable of speech. ¹⁶

The latter, in isolation, cannot be relied upon in Jewish law (Halachah) to ascertain whether Golems are or are not human. Although the Hebrew term for “speaker”, M’daber is used with reference to people, it is not the sole arbiter of the designation “human”. After all, parrots speak and some people are unfortunately mute. The Torah defines “human” as someone “born of woman,” ^{17, 18} clearly disqualifying the Golem and with the same brush, Frankenstein’s monster ¹⁹ and any AI regardless of their performance on the Turing test or other measures of sapience. ²⁰

The Talmud stipulates that a Golem cannot be tallied in the assembly of a Minyan (quorum of ten men required for Jewish public worship). Neither does destruction of a Golem constitute a capital offense because only one who intentionally kills a person “formed within another human being” is liable for murder. ²¹ This latter definition is tantamount to insistence on a natural biological origin and should not be taken literally.  It would certainly not, for example, absolve one of homicide in cases where the victim was a cloned human being or whose embryonic and fetal ontogeny transpired entirely ex utero—hypothetical situations today, perhaps, but plausible scenarios in the not-too-distant future.

Several prominent adjudicators ^{22, 23} citing Kabbalistic texts ^{24, 25} opined that whereas only God can endow persons with human Souls, the power of the Sefer Yetzirah wielded appropriately may sustain Golem consciousness at the level of an animal Soul (Chaye). This may have several significant ramifications in Jewish law regarding “Golem rights” and by extrapolation how we might ideally interact with sentient AI in a futuristic society. If human beings indeed have the wherewithal (be it spiritual or scientific) to create Golems with animal-like consciousness, then this knowledge or skill may, in principle, carry over to the development of advanced robotics.

As such, Jewish legislation concerning the treatment of animals may be a useful guide to our, some would say inevitable, relationships with sentient AI. Relevant injunctions to consider are “Eiver Min Ha’chai”—the prohibition against consuming the flesh of a limb torn from a living animal ²⁶—and “Tzaar Ba’alei Chaim”—the command against causing unnecessary suffering to animals. ²⁷ The latter entails not only physical but also emotional anguish as is made clear by additional prohibitions against removing eggs or chicks from a nest in the presence of the mother bird (“Shiluach Hakan”) ²⁸ and the yoking together of beasts of burden of unequal strength. ²⁹ Along these lines, some contemporary voices advocate avoidance of emotional distress to livestock by shielding them from the slaughter of others. ³⁰ 

5. Conclusion

Whether teaching by experience or allegory (as the Torah is wont to do), Golem lore cautions that we should not dismiss out-of-hand the possibility that artificial neural nets (be they electronic, positronic, or something beyond our current imagination) may one day confer to machines a sense of what it is like to experience pain, joy or frustration—qualia shared by many non-human species as any dog owner will attest to. We may never really know what Chaye-level consciousness, were it to arise, “feels” like to AI.

It has been suggested that a baby step in this direction (although an endeavor fiendishly elusive in its own right) presupposes the delineation of “markers” of consciousness/feelings in a wide spectrum of non-human animals, such as bees, octopi, and crows, in order to distinguish between true sentience and “gaming” (learned behavioral mimicry of sentience). ³¹ To the extent that the Rabbinic analysis of Golem consciousness provides useful guidance and a valid basis for extrapolation (a supposition that at this juncture remains more securely moored to faith than fact), it would behoove us to treat any potentially sentient products of our own design with the same sensitivity, respect, and jurisprudence we courteously afford our most cherished non-human life forms. 

1. Oxford Dictionary. Available at https://en.oxforddictionaries.com/definition.

2. Butlin P, Long R, al. e. Consciousness in Artificial Intelligence: Insights from the Science of Consciousness. arxivorg [serial online] 2023;88. Available at: https://doi.org/10.48550/arXiv.2308.08708.

3. Kleinman Z. AI creators must study consciousness, experts warn. BBC News [serial online] 2023. Available at: https://www.bbc.com/news/technology-65401783.

4. Schipper HM. Love: A Kabbalistic Perspective. Beyond Belief [serial online] 2023. Available at: https://www.beyondbelief.blog/p/love-a-kabbalistic-perspective?utm_source=substack&utm_medium=email.

5. Schipper HM. Kabbalistic Panpsychism — The Enigma of Consciousness in Jewish Mystical Thought. Winchester, UK: John Hunt Publishing, 2021.

6. Schipper HM. Does the Kabbalah Acknowledge Free Will? Beyond Belief [serial online] 2023. Available at: file:///D:/SYNC/Kabbalah/Aish/Kabbalah_Free%20Will/Does%20the%20Kabbalah%20Acknowledge%20Free%20Will_.pdf.

7. Vital C. Etz Chaim 42:1. [1782] 1998.

8. Mishnah Kela'im: 8:5.

9. Leoni E. Rabbi Pinchas of Koretz. 2015. Available at: https://www.jewishgen.org/yizkor/Korets/kor031.html#f31-8r.

10. Cepelwicz J. An Ethical Future for Brain Organoids Takes Shape. Quanta Magazine [serial online] 2020. Available at: https://www.quantamagazine.org/an-ethical-future-for-brain-organoids-takes-shape-20200123/.

11. Psalms 139:16.

12. Kaplan A. Sefer Yetzirah - The Book of Creation. San Francisco: Weiser Books, 1997: 38-40.

13. Verney TR. The Embodied Mind: Understanding the Mysteries of Cellular Memory, Consciousness, and Our Bodies. Pegasus Books, 2023: 138.

14. Triballeau C. Japan roboticists predict rise of the machines 2019. Available at: https://www.afp.com/en/news/15/japan-roboticists-predict-rise-machines-doc-1hu9ki1.

15. Hinchliffe T. AI and Spirituality: Toward the recreation of the mythical, soulless Golem. The Sociable [serial online] 2017. Available at: https://sociable.co/technology/ai-spirituality-soulless-golem/.

16. Babylonian Talmud Sanhedrin 65b.

17. Job 14:1.

18. Job 15:14.

19. Shelley M. Frankenstein: The Modern Prometheus. UK: Lackington, Hughes, Harding, Mavor & Jones, 1818: 280.

20. Marcus G. Am I Human?: Researchers need new ways to distinguish artificial intelligence from the natural kind Scientific American 2017;316:58–63.

21. Responsum Chacham Tzvi 93.

22. Divrei Rabeinu Meshulam 10.

23. Sheilot Yaavetz 2:82.

24. Azulai A. Chesed L’Avraham 4:30.

25. Cordovero M. Pardes Rimonim 24:10.

26. Genesis 9:4.

27. Babylonian Talmud Bava Metzia 32b.

28. Deuteronomy 22:6-7.

29. Deuteronomy 22:10.

30. Love S. Do Animals Understand What It Means to Die? 2022. Available at: https://www.vice.com/en/article/5dg57q/do-animals-understand-what-it-means-to-die.

31. Birch J, Andrews K. What Has Feelings? Aeon [serial online] 2023. Available at: https://www.lse.ac.uk/philosophy/blog/2023/03/03/what-has-feelings-essay/.