Chatting with the Bot

ChatGPT has only been with us for a little less than three months, but it has already garnered far more interest—and more nervous criticism—than any earlier chatbot has so far managed to attract. (For the as-yet-uninitiated, a chatbot—also occasionally called a chatterbot—is basically a computer program that knows enough of human language, culture, history, and dialogic technique to participate in a conversation in convincing enough a way to appear to be a human interlocutor.) They’ve been around for a while, too: I wrote a Friday letter five years ago about the efforts of James Vlahos to create a chatbot programmed so successfully with his father’s speech patterns, vocabulary choices, grammatical usages, and idiosyncratic ways of expressing himself that it felt possible for Vlahos to have an actual conversation with his father long after the latter died. (To see that letter, click here. And for a very moving account of the whole project by James Vlahos himself, click here.) At the time, it seemed more amusing than sinister to me, like a kind of living memorial to a late parent that would be a huge advance over just watching that parent in a home movie and being unable personally to step into the action. But it didn’t seem to me as much more than that.

We’ve come a long ways since 2017. Yes, it’s still easy enough to know when you are talking to a machine when you try to phone a bank or an insurance company. And some of those machine-generated partners-in-dialogue actually do a fairly good job of assisting you if you are patient enough to let them do their thing. Still, no one doesn’t prefer to speak to an actual human being, I don’t think. Certainly, I myself do. But no amount of machine-generated robocalls or chatbot-based customer service offerings prepared me even slightly for ChatGPT.

You may have heard about it already. There was, for example, a very long story in the Times just the other day about how the advent of ChatGPT, the latest offering by OpenAI, has transformed the way at least some college courses are taught. (To Kalley Huang’s article, click here.) That in and of itself is pretty amazing, since ChatGPT has only been around since last November. But I was intrigued. It wasn’t until I actually signed on to the OpenAI website and engaged personally with ChatGPT, however, that I realized just how remarkable a door has now been opened to all who dare step across the threshold. So I write today to share some of the details of that specific journey I’ve been on the course of these last few days.

I’m sitting in my study at home on Reed Drive. Joan is at school; it’s just me here. My computer is open and I’ve logged onto the OpenAI site. I can’t decide if I’m being bold or silly in taking this on but I’m also slightly worried. What if the bot really is more capable of intelligent theorizing regarding Jewish law or thoughtful rumination regarding the literary heritage of ancient Judaism than I am? Do I really want to surrender my superior human status that easily to a fancy piece of software? Or would I be acting foolishly not to see for myself how this works? A lifetime ago, I remember snickering when a fellow student in grad school purchased a home computer upon which to write his dissertation and wondering what he can have been thinking by wasting such a huge sum of money on a glorified typewriter. I was a bit hasty in that evaluation, I now realize. So let’s not make the same mistake twice, I tell myself. And so, throwing caution to the wind, I begin.

I figure I might as well start off with some simple stuff.

I ask the bot for the name of Abraham’s father. And the answer comes back instantly:

Abraham's father was Terah.

So that’s easy enough. Who doesn’t know that? But the point here is not just to test the breadth of the program’s knowledge, but to see if it can think. Or appear to think. So let’s go more subtle.

Was Haman a wicked man? I ask. It’s not as simple a question as it sounds and, amazingly, the bot seizes the complexity behind what sounds at first blush like a very simple question.

Haman is a character in the Book of Esther in the Hebrew Bible. He is depicted as a wicked man and the prime antagonist of the story, who plots to exterminate the Jewish people.

I’m impressed. The bot recognizes that Haman is a character in a book, not a historical personality whose existence can be corroborated by scientific evidence. And it seems to realize that what we know of Haman derives directly from the way he is depicted in the Book of Esther. So the bot hedges its bets, calling him a character in a book and then answering my question with reference to the way he is depicted in that book rather than as a simple historical fact.

And now I’m ready to wade into more challenging terrain. So it knows a lot of stuff. But can it think? Or can it appear to think?

What, I now wonder, will it do when the question is by nature unanswerable. Is it justifiable, I now ask, to steal food to feed an infant that will not survive without that food?

This takes a few seconds (the other answers were instant), but the answer, I have to admit, is remarkable:

Stealing is generally considered to be morally and legally wrong. However, there may be some circumstances in which it could be considered justifiable to steal food to feed an infant in order to save the infant's life. This could be considered a form of "necessity defense," in which a person commits a crime because they believe it is necessary to prevent a greater harm. However, it's important to note that whether or not this would be considered a valid defense in a court of law would depend on the specific details of the situation and the laws of the jurisdiction in which the crime was committed.

It's worth mentioning that there is usually help available to those in need, in developed countries, there are many organizations that provide assistance to those facing food insecurity, and it may be possible to obtain food for the infant through these channels without resorting to theft.

 Wow. I wonder what the bot would make of the storyline of Les Mis! (Could a chatbot even go to the theater? Where would it sit?) But the question isn’t only whimsical but also real: is the bot really thinking? It’s hard to say. The answer is intelligent, well-thought-out, reasonable. The bot knows the “simple” answer (i.e., stealing is wrong), but it seems able to weigh that thought against other principles in play (i.e., it can never be right to allow a child to die of hunger if you can prevent that from happening). And it did a pretty good job of weighing those principles against each other and coming up with a thoughtful, rational answer.

I wonder how it will do with questionable premises. Will it follow the question’s lead or challenge its fundamental basis?

If, I now type in, life begins with conception, is aborting a fetus no different than murdering a child? It’s a loaded question. Given the landscape in today’s America, it’s a very loaded question. But since I’ve formulated it as I did, will the bot “just” answer the question or will it attempt to nudge me off my premise?

It took almost half a minute to respond. But then it came up with this:

The question of when life begins is a deeply philosophical and moral one, and opinions on the matter can vary widely. Some people believe that life begins at conception, while others believe that it begins at some point later in pregnancy.

If one holds the belief that life begins at conception, then it could be argued that aborting a fetus is the same as killing a child. However, many people who hold this belief still believe that there are certain circumstances in which abortion may be morally justifiable, such as to save the life of the mother, in cases of rape or incest, or if the pregnancy is not viable.

On the other hand, others may disagree with the statement that life begins at conception and may argue that a fetus is not yet a person, and therefore not equivalent to a child. They may argue that a woman has the right to control her own body and make decisions about her pregnancy.

Ultimately, whether or not one considers abortion to be the same as murder is a matter of personal belief, and it is important to respect and consider the perspectives of those who may hold different views on this issue.

So that’s not a bad answer. If I were asked that precise question in my role as community rabbi, I think my answer would be similar. I’d agree that if the protasis (i.e., the “if” part of an “if/then” statement) is a given, then the apodosis (i.e., the “then” part) would also be correct. But I’d also hasten to point out that that the issue would only be a simple one if you could prove that the protasis were a simple truth and not the matter of intense controversy that it is in the world. And no one, at least so far, has determined definitively when life begins.

And then I think to wonder what the bot might have to say about the issue that concerns me the most in the world: the future of the Jewish people in a world of ever-mounting anti-Semitism. I type slowly, wondering if this is a good idea. But then I remind myself that I’m talking to a machine and I press the “submit” button. This one, it really has to think about. It ruminates for almost 90 seconds. But when it does respond, the answer is more encouraging than almost anything I’ve read on the topic lately.

It is difficult to predict the future, but it is likely that the Jewish people will continue to exist in the future. The Jewish people have a long history and a rich cultural heritage, and there are currently millions of Jews living around the world.

However, the Jewish people have faced significant challenges throughout history, including persecution and violence. Anti-Semitism still exist in some parts of the world and it is something that Jewish communities continue to confront.

The State of Israel, established in 1948, is the homeland of the Jewish people and it is a place where they can live in security and prosperity. The State of Israel is home to a large and vibrant Jewish community and it is likely that it will continue to be an important center of Jewish life in the future.

Overall, while it is impossible to know what the future holds, the Jewish people have a strong sense of community and history, and it is likely that they will continue to thrive and flourish in the years to come.

 I’ll take it! And that concludes my AI session with ChatGPT. (By the way, the GPT part stands for Generative Pre-Trained Transformer, whatever that means exactly, and the AI in OpenAI stands for Artificial Intelligence.) Would you like to take it out for a spin? Just go to https://openai.com, then click on ChatGPT on the lower left corner of the screen. (You’ll have to scroll down a bit to get to the bottom.) And then you can ask away and see what comes back.