The Next Frontier in Jewish Law: Artificial Intelligence

 

With the emerging field of artificial intelligence (AI) becoming increasingly relevant to our everyday lives, Jewish Action hosted a conversation between Rabbi Dr. Ari Z. Zivotofsky, a professor in the Neuroscience Program at Bar-Ilan University and a longtime columnist of Jewish Action, and Rav Yosef Zvi Rimon, rabbinic head of the Jerusalem College of Technology (JCT), to understand some of the implications of this technology and its impact on halachah. In addition to his role at the JCT, Rav Rimon serves as a rav in Alon Shevut Darom and teaches classes in halachah at Yeshivat Har Etzion. He also teaches at the Herzog College and at the Beit Midrash for Women in Migdal Oz. Rav Rimon was awarded the Moskowitz Prize for Zionism on Yom Yerushalayim in 2014 for his monumental project JobKatif on behalf of the Jews of Gush Katif. As part of his work at JCT, Rav Rimon is opening a beit midrash that will research the application of halachah to the newest technological developments. 

Special thanks to Dr. Naomi and Rabbi Dr. Ari Z. Zivotofsky for assisting in the preparation of this article. 

 

Rabbi Dr. Ari Zivotofsky: Webster’s Dictionary defines artificial intelligence (AI) as a branch of computer science dealing with the simulation of intelligent behavior in computers. In recent years, there have been tremendous developments in this area. Computers can actually learn, to the point that they can extrapolate and deduce. AI uses computer algorithms to arrive at conclusions, such as medical decisions, often without direct human input. Will AI limit the role of doctors by making predictions and diagnoses? Halachically speaking, can we rely on a computer to make medical decisions? 

Rabbi Yosef Zvi Rimon giving a talk on AI and halachah at the OU’s Torah New York 2019 event held this past September at Citi Field. Photo: Kruter Photography

Rabbi Yosef Zvi Rimon: This is an excellent question. As you mentioned, there has been extraordinary progress in this area, and AI has become increasingly intelligent over time.  

In 1997, IBM’s chess-playing program Deep Blue beat world chess champion Garry Kasparov. The principles of the game were programmed into the computer, which masterfully won the match by blindly searching through millions of moves. Some two decades later, Google’s AI subsidiary DeepMind devised AlphaZero, a machine-learning algorithm. Given no (“zero”) human input aside from the rules of chess, it beat the world’s best chess-playing computer program after continuously playing chess against itself for a mere four hours; it played the game millions of times and learned from its mistakes. AlphaZero is regarded as the best chess player in the world—human or computer—and when playing, it actually displays evidence of insight, a new kind of artificial intelligence. Nowadays, AI no longer just acquires knowledge; it arrives at conclusions on its own (mevin davar mitoch davar). AI is no longer just feeding information into one big database. As you see with AlphaZero, there is actual learning going on.   

There’s no question that AI is becoming increasingly sophisticated and is able to perform human activities with greater speed and efficiency, often at a lower cost. In the healthcare system, AI is being used to spot lesions on mammograms, and AI systems are being developed to diagnose whether or not one will develop Alzheimer’s disease or other diseases later in life. Scientists are also working on developing algorithms that will help them make decisions about cancer treatment. A doctor may know how to interpret X-rays and can be well versed in the most current research, but computers have access to far more data and therefore arrive at much more precise conclusions. Based on the trajectory of today’s research, one can envision that there will come a day when artificial intelligence will be making most medical decisions.  

Nevertheless, computers are not infallible, and a doctor has something that AI lacks—human intuition.  

RZ: Can AI help us perform mitzvot? What about teaching a robot to do nikkur achorayim (the removal of certain large blood vessels, cheilev [prohibited fats] and the gid hanasheh [sciatic nerve] after a kosher animal is properly slaughtered and inspected)? What about teaching a robot to check articles of clothing for traces of shatnez? Do you see a problem with this? 

RR: I don’t see why checking for the presence of shatnez—wool and linen mixed together in an article of clothing—would be a problem, since it’s a technical process. No special kavanot (concentration of the mind in performance of a religious act) are required. Furthermore, nowhere in the Torah is it written that a human being must do the checking. Similarly, a robot should be able to do nikkur achorayim; however, the actual shechitah (the Jewish religious and humane practice of slaughtering animals) must be done by a God-fearing Jew. The Shulchan Aruch (Yoreh Deah 2:11) rules that shechitah is only acceptable if it is done by a human being. Thus it should be clear that shechitah may not be done by a robot. See also siman 7:1 where the Shulchan Aruch explicitly states that shechitah may not be done by a machine that is not directly operated by a human.  

Could a computer check the sharpness of the shochet’s (ritual slaughterer) blade? The chalif, the knife of a shochet, is his most important tool. It must be exquisitely smooth and exceedingly sharp. The Shulchan Aruch states that one must check the knife b’kavanathalev (with mindful concentration) twelve times (YD 18:9). A laser could be used to check the sharpness of the knife but cannot serve as the the final arbiter because a human being has to do the checking. A laser, can, however, issue a warning to the schochtim indicating the knife is too dull.  

RZ: Video cameras are used for certain aspects of kashrut supervision. For example, on a dairy farm, they are used to verify that the milk produced is cow’s milk and that it was not extracted in violation of Shabbat, et cetera. If a robot could be programmed to learn what’s important in a kosher kitchen, such as not mixing meat and milk and using only kosher-certified products, would it be acceptable to use robotic mashgichim? 

RR: Generally speaking, there is no requirement that a mashgiach has to be physically present. The halachah wants to ensure that a proprietor does not do anything improper. The halachic requirement is “mirtat,” which means that the proprietor has to be afraid that he will be caught and will suffer the consequences if he tries to do something improper. Assuming a robot is observing and recording what is taking place, and a human is monitoring the cameras, I think that would be acceptable. But a human being must be monitoring, because if there are no consequences [for violating kashrut standards], what’s the point of the robot? 

RZ: Our discussion until now centered on today’s technology. Let’s spend a few minutes discussing future technology.  

A new technology, known as brain-computer interface or BCI, enables users to interact with computers via brain activity or “thinking” only. One use of BCI is to enable people with paralysis and other disabilities to control robotic arms or other devices by thinking about such actions. While it seems like science fiction, the wiring together of brains and computers is actually a reality and companies have invested millions of dollars in moving progress along in this area. It is very likely that one day our brains might be able to interact with our smartphones or tablets—and we will be able to turn an oven or light switch on simply by thinking about it. 

The question is, can one theoretically fulfill a mitzvah simply by thinking about it? Conversely, can one violate Shabbat by merely thinking about doing a forbidden act? In a BCI world, is “thinking” defined as “doing”? 

RR: The well-known Shabbat zemer “Mah Yedidut” states, “Chafatzecha assurimv’gam lachashov cheshbonothirhurim mutarimulshadeich habanot—[On Shabbat] your non-Shabbat desires are prohibited, as is performing business calculations/ but pleasant thoughts are permitted, even making matches for the daughters.”  

The question is—what types of “thoughts” are permitted on Shabbat? 

There are several Torah sources that define thought. One such source is found in Tosafot in Masechet Gittin 31a, which discusses separating terumot and ma’aserot (tithes) by “thought” on Shabbat. Can one actually separate terumotand ma’aserot simply by thinking about it? Is thought considered an action? There is a machloketAcharonim on how to understand this Tosafot. According to Rabbi Shlomo Zalman Auerbach it is forbidden to separate terumot and ma’aserot in this manner [on Shabbat]. Thus, according to his view, thought is equivalent to action. Based on this, I don’t think BCI would be permitted for general use on Shabbat; however, if we’re talking about an individual who is paralyzed and BCI [can enable him to move his limbs on Shabbat], perhaps there is room for discussion. 

RZ: What about fulfilling positive mitzvot through brain-computer interface? For example, can one use BCI to lift the arba minim or even to give a ring to a woman in order to betroth her?   

RR: Since one can be mekadesh (betroth) a woman even through a shaliach, using BCI for this purpose may not be a problem.  

However, using BCI to fulfill a positive commandment is another matter. On the one hand, an individual using BCI has not actually performed an action because the computer is an intermediary directing his hand or artificial limb. On the other hand, the individual’s thoughts are controlling the movements, and perhaps that suffices to fulfill the mitzvah.  

Even if a robot were to think on its own, would we be able to rely on it for mitzvot that require the involvement of a human being? 

I believe that with regard to the issue of fulfilling a mitzvah via BCI, each and every mitzvah would have to be discussed and evaluated separately. 

RZ: We’ve touched upon the connection between AI and almost all four parts of Shulchan Aruch. We spoke about kashrut, marriage and Shabbat. What about nezikin [tort law] in Choshen Mishpat? 

Let’s assume, for example, that a self-driving car causes harm to property—or, God forbid, kills someone. Although there may have been a human in the car, an autonomous system—AI—was in full control of the vehicle. Who is responsible? The “driver”? The programmer? The owner of the vehicle? The company testing the car’s capabilities? The manufacturer? 

RR: I was actually asked this very question after a self-driving vehicle did indeed kill someone. According to a teshuvah of the Rosh, it is understood that when a person is controlling a steering wheel in a conventional car, it’s as if he is holding a sword. Therefore, if he turns the wheel and kills someone, he is held responsible. The case is classified as “adamhamazik,” an individual who caused harm. It then needs to be determined whether the driver killed the individual by accident—in which case he would be liable for the Torah’s punishment of exile in an ir miklat [city of refuge]—or if he did so intentionally—in which case, according to the Torah, the beit din would determine his punishment.  

With regard to a self-driving car, however, the car is in control, and therefore the halachic category of adamhamazik does not apply; rather, the category of “mamonhamazik,” property that causes damage, applies. This is similar to a case where a bull gores a person and injures him, which is classified as “mamonshehizik. Assuming the autonomous car has the status of mamonshehizik, who is going to take responsibility? It would seem that the owner of the car would have to take responsibility, since the car is his property. However, unlike the cases of mamon hamazik discussed in the Gemara, the owner of the vehicle could argue, “Why am I to blame? There is someone out there who programmed the car.” One could counter that as the owner, he was obligated to ensure that the programmer’s work was up to standard, and if he did not do so and his property caused harm, he is accountable.  

The underlying question here is how halachah defines the basis of the owner’s responsibility for damage caused by his property. Is an owner inherently responsible for damage caused by his property, unless there are circumstances beyond his control? If this is so, he would be responsible for damage caused by a self-driving car, unless he can prove there were extenuating circumstances. Or is he only responsible for damages that were caused due to negligence on his part, e.g., he didn’t take sufficient security measures to ensure his animal or property would not cause damage? In that case, it would have to be determined that there was some degree of negligence. 

RZ: Do you see a future where a robot or computer will be comparable to a human being? Could a robot, for example, join a minyan? 

RR: Currently, there is a debate in the scientific community as to whether or not we can actually develop the technological know-how to manufacture a robot that will be able to think on its own. Some scientists believe this will happen; others say it will never come to be.  

Even if a robot were to think on its own, would we be able to rely on it for mitzvot that require the involvement of a human being? I don’t think so. A robot could never be considered a human for the purposes of joining a minyan. And even if it were to write in the most beautiful manner and in the most precise way, it could not write a sefer Torah. This is because for such mitzvot, the Torah requires a Jewish person with da’at (understanding). Even if a robot had sechel (intellect/brainpower), it would be lacking da’at. A robot can never attain the status of a human being, and therefore it cannot perform these religious functions. 

RZ: What about future smart homes and Shabbat? Smart homes exist even today, but as AI becomes even more advanced, it will most likely fulfill residents’ wishes before they are even aware of them, presenting new halachic questions.  

Imagine entering your dining room in a future smart home. Your smart home has learned all about you and is aware that you entered the room. It knows the kind of music you prefer and starts playing your favorite songs. After a few minutes, it recognizes that you are thirsty and prepares a cup of coffee. Eventually there may be sensors all over the home—sensors in the refrigerator will detect that you are running low on breakfast foods and will place an order online; sensors in the medicine cabinet will check if you have taken your pills, et cetera. How will we deal with all of this on Shabbat? 

RR: In order to explore this question, let’s begin with sources in the Gemara that deal with “davar she’eino mitkaven,” unintended consequences of a permitted action. In Beitzah 23b, the Gemara presents a disagreement between Rabbi Shimon and Rabbi YehudahRabbi Shimon says: A person may drag a bed, a chair or a bench on the ground [on Shabbat], provided that he does not intend to make a furrow. Rabbi Yehudah says: No vessels may be dragged. Rabbi Shimon permits one to drag a bed, since the furrow that is created is a davar she’eino mitkaven; he had no intention to plow.  

This is similar to another Gemara that discusses one who sees a gazelle in his house. If he closes the door, he traps the gazelle, which is forbidden on Shabbat. Is he allowed to close the door? The Rashba (Chiddushei HaRashba, Shabbat 107a) responds with the following: In Yerushalmi [13:6], it appears that from the outset they permitted him to lock his house, along with the gazelle which is in it, in order to protect his home. This is because even though doing so will trap the gazelle inside, since he needs to protect his home it is permissible, as long as he did not only intend to trap the gazelle. 

It would appear that intent plays a pivotal role in determining whether or not an act constitutes a violation of Shabbat.  

Let’s look at a more contemporary illustration of this principle: Say your neighbor has a light sensor in his front yard and if you pass by the house, the sensor will detect movement and the light will go on. Can you pass by your neighbor’s house on Shabbat? Rabbi Shmuel Wosner writes (Shevet Halevi 9:69) that as long as your intention is not to go there to turn on the light, it is permissible.  

Now let’s examine our original question of smart homes and sensors on Shabbat. The fact that AI sensors are continually learning about you even as you move around your home on Shabbat, does not necessarily constitute a violation of Shabbat (since it is an unintended consequence of your actions). However, suppose you go downstairs to your kitchen on Shabbat morning and after a few minutes the shutters suddenly open and the coffeemaker begins preparing coffee because your smart home “recognizes” that at that particular time and temperature you like the shutters open and a coffee ready—that would be problematic. This is because AI is doing a forbidden melachah (cooking), as a result of your activity.  

There is, however, a greater concept at play here that should be discussed. There are certain activities that the rabbis forbade on Shabbat, even though they are Biblically permissible, because they have the potential to destroy the nature of Shabbat.  

A religious man I know has a very successful Internet reputation management business. One Shabbat, a client in New York was desperate to reach him. It was an “emergency.” She e-mailed him and did not receive a response. She texted him. He didn’t respond. She called him. He didn’t answer.  

After Shabbat, when he saw her urgent messages, he returned the call. Aggravated, she asked him, “Why didn’t you answer my calls? I lost thousands of dollars today!” He explained to her that it was Shabbat. 

“I have Sunday as my day off, but when there’s an emergency, I respond,” she said. 

“Our Shabbat is different. I can’t answer e-mails,” he replied.  

“But I texted you!”  

“I can’t text on Shabbat.”  

“But I called you!” 

“I can’t answer calls on Shabbat.”  

“So what do you do on Shabbat?” asked the woman, intrigued.  

“We go to the synagogue, we pray, and we eat a meal with our family,” he replied. 

“Okay,” she said, “but a meal takes twenty minutes.”  

“No,” he explained. “Our Shabbat meal can take two hours—we sit together with our children, we sing, we talk, we discuss ideas.” 

The woman was taken aback. “From time to time,” she said, “I try to gather my family together for a meal so we can have some quality time, but even if I finally succeed in getting everyone to sit down together, my husband is busy with his phone. Then one child gets a WhatsApp message and is distracted. Another child gets a text and leaves the room. Shabbat seems like an amazing invention. Can you please give me the formula for Shabbat?”  

In truth, while it can sometimes seem as if halachah closes doors, it actually opens up massive gates, allowing one to enter areas that would otherwise be inaccessible. A teenager once told me that by 4:00 pm on Shabbat afternoon he could no longer hold out without his phone. I told him, “You have a serious problem if you cannot be alone with yourself for twenty-four hours. Shabbat is your savior.” On Shabbat, you say no to your computer, no to your phone, no to the Internet; but through Shabbat you open doors to yourself, to your soul, to your family and to Hashem.  

I believe in the derech of Rabbi Avraham Yitzchak HaKohen Kook who embraced progress and felt that technological advancements are positive overall, for the world and for Klal Yisrael. Nevertheless, the fallout of being available all the time is not a positive development; it means that one is always distracted and never focused.  

We must be wary of turning Shabbat into a yom chol. It is up to our generation’s posekim to assess technological advances and to determine what is allowed and what is forbidden, and when heterim are detrimental to upholding the sanctity of Shabbat. Shabbat is the only time we close all our doors and stay focused on the truly important things in our lives. In today’s modern times, Shabbat is more necessary than ever.  

More in this Section: Rabbi Tzvi Ortner: The Solution Finder

This article was featured in Jewish Action Spring 2020.