Finding Smart Applications for AI in Torah Life
You'll never believe what AI has to say about us
Not long ago, while engaged in my professional work, I came across the GPT-3 artificial intelligence engine built by OpenAI (a tech company founded in part by Elon Musk). It didn’t take me long to realize that AI has come a very long way since the last time I’d looked.
How does it work? Well, here’s how GPT-3 itself answered that question:
NLP stands for Natural Language Processing. AI NLP is a subset of AI that deals with the processing of human language data. It involves the use of computers to analyze, understand, and generate human language.
GPT-3 is based on the machine learning model known as a recurrent neural network (RNN), which is specifically designed to understand sequences of words or characters. By analyzing large quantities of text data, GPT-3 can learn how to identify patterns and relationships between words, which allows it to understand the meaning of human language.
In the few weeks since my discovery, I’ve been getting to know GPT-3. Quite well, actually. Early sessions led to the content for the book: “The Wisdom of Artificial Intelligence - Conversations with a cutting edge AI engine”. Subsequent experiments produced a series of short books offering introductions to various technologies.
That’s six books written and published in less than three weeks, for anyone who keeps score.
The real excitement of discovering new technologies comes from finding new ways to apply them. So, taking a break from churning out IT books at a rate of one per day, I thought I’d explore how GPT-3 might be useful in a Torah context.
Hebrew Translation
I began with some Hebrew (and Aramaic) translation. Here’s how the AI handled the Rambam in הלכות שביתת יום טוב ה:א
אף על פי שהותרה הוצאה ביום טוב, אפילו שלא לצורך, לא יישא משאות גדולות כדרך שהוא עושה בחול, אלא צריך לשנות; ואם אי אפשר לשנות, מותר. כיצד: המביא כדי יין ממקום למקום--לא יביאם בסל ובקופה, אבל מביא הוא על כתפו או לפניו. המוליך את התבן--לא יפשיל את הקופה לאחוריו, אבל מוליכה בידו
GPT-3:
Even though there is a holiday, he is not allowed to carry big loads like he does on other days. He needs to change how he does things; if he can't change, then it's okay. How: He should not bring the wine in baskets or boxes, but rather on his shoulder or in front of him. He should not push the cart behind him, but pull it with his hand.
It’s clearly not perfect. But you can see that, rather than attempt a straight translation, the AI is trying to understand the context and meaning of the passage.
Article writing
As I noted, I’ve already used GPT-3 to write entire books. But can it produce credible Judaic content? Here’s one of my prompts:
Write me an article that will convince a 10 year old Jew to keep the laws of the Sabbath.
With nothing more than that, here’s what it gave me:
Dear young Jew,
The Sabbath is a very important ritual in Judaism and it is something that you should definitely try to keep. Here are four reasons why you should keep the Sabbath:
1. The Sabbath is a reminder of our creation. We were created on the sixth day of creation and God rested on the seventh day. The Sabbath is a reminder of God’s creative power and how we should rest on the seventh day as well.
2. The Sabbath teaches us to appreciate our time. We only have one life and we should make the most of it by resting and enjoying time with our family and friends.
3. The Sabbath teaches us about self-control. It is not easy to stop working or doing things we enjoy, but by observing the Sabbath, we learn how to control ourselves and put our priorities in order.
4. The Sabbath connects us to our Jewish heritage. By keeping the Sabbath, we are honoring our ancestors who kept this tradition for thousands of years.
Not too bad. And that wasn’t plagiarized text. I’ve already confirmed that the AI is careful to avoid just copying from existing content.
Now I definitely wouldn’t leave GPT-3 alone with your kids (or in charge of your kitchen). Here’s what it thinks about kashrus:
Kosher laws are a set of dietary laws that are meant to help people observe a healthy and clean diet. There are many specific rules that govern how kosher food must be prepared, but some of the most important rules include avoiding mixing meat and dairy, keeping meat and poultry separate, and cooking foods properly. Kosher law also prohibits eating pork or shellfish, as well as any animal that doesn't have a cloven hoof and chew its cud.
Yep. We all know how hard it can be to cook foods properly. Especially with those angry-looking rabbis standing there and radiating disapproval.
Information resources
My final example for now is focused on using GPT-3 for retrieving hard-to-find information. Here’s a series of questions I posed along with GPT-3’s accurate and time-saving responses:
Where can I find demographic data about orthodox Jewish communities?
There is no comprehensive source of demographic data about orthodox Jewish communities. However, a variety of sources provide some information on the size and composition of these communities.
Which sources?
Sources of demographic data about orthodox Jewish communities include the U.S. Census Bureau, the Pew Research Center, and the North American Jewish Data Bank.
How can I find such data in the U.S. Census Bureau?
The U.S. Census Bureau publishes data on religious affiliation in its American Community Survey (ACS). The ACS is an annual survey that collects information on a wide range of topics, including religious affiliation. You can find data on orthodox Jewish communities in the ACS by using the "Religious Groups" table. This table includes data on the size and composition of orthodox Jewish communities across the United States.
How can I find such data from Pew Research Center?
The Pew Research Center publishes data on religious affiliation in its Religious Landscape Study (RLS). The RLS is a comprehensive study of religion in America that includes data on a wide range of religious groups, including orthodox Jews. You can find data on orthodox Jewish communities in the RLS by using the "Orthodox Jews" table. This table includes data on the size and composition of orthodox Jewish communities across the United States.
Having spent plenty of time searching through the U.S. Census Bureau website, having a quick short cut can be really helpful (assuming the information turns out to be accurate).
At any rate, those were just a few off-the-top-of-my-head examples. I’m sure I haven’t even scratched the surface of what GPT-3 makes possible.
What applications can you imagine?