Brian Lehrer: It's The Brian Lehrer Show on WNYC. Good morning, everyone. Coming up later in the show, a Brian Lehrer Show editorial board segment on the year in hate, 2022 in hate with three special guests, Princeton professor Eddie Glaude, Rabbi Sharon Kleinbaum and New Yorker Writer Jay Caspian Kang. With all the hate crimes records that are being set this year, there will be a lot to talk about and try to brainstorm our way out of. We will even talk about the year in love as part of that segment.

2022 may also be remembered for the launch of something most of you have probably not even heard of yet. We're leading with this today here on December 9th, 2022 for the historical record. What is it? It's ChatGPT. While discussions about artificial intelligence occasionally bubble up to the general public, regular people haven't been able to experience an AI product very much that is both intelligent, despite AI standing for artificial intelligence, and functional.

At best, AI products have been buggy and not that useful in most applications. At worst, they start spewing hate speech like Microsoft's 2016 chatbot that was shut shortly after launch. Maybe bots are like real people that they're doing that, but that is until ChatGPT. It might be the year 2022 that we look back on and say, "That's when ChatGPT came in and everything changed."

Suddenly, it seems like everyone who's heard about it is not only talking about artificial intelligence but also playing with this and sharing how they're utilizing it to perform a wide range of different tasks. Product designers are using it as a UX writer. Students are using it to write essays that pass plagiarism detectors, they didn't really write them themselves, the chatbot did, but some people are even getting the chatbot to generate movie scripts.

The chatbot has already reached 1 million users since its launch just on November 30th. It might not just become popular. It could change a lot about our world and about ourselves, for better or for worse, if things go certain ways. With us now to discuss ChatGPT and the future of this new technology, potentially, is Ann-Marie Alcántara, internet culture reporter for The Wall Street Journal's Personal Tech bureau. Ann-Marie, welcome to WNYC. Thank you so much for coming on.

Ann-Marie Alcántara: Of course. Thank you for having me this morning.

Brian Lehrer: Listeners, have you tinkered around with ChatGPT yet? If so, what have you used it for? 212-433-WNYC. Did you get it to write you an original movie script that you're now shopping around in Hollywood, or to explain a complicated concept for a second grader, or to write a term paper that you're presenting as having come from you? Are you optimistic about our future with AI on the basis of your experiments with ChatGPT or are you concerned? Call us with your questions and answers at 212-433-WNYC and share your experiences at 212-433-9692 or tweet @BrianLehrer. Ann-Marie, give us some 101. What is ChatGPT and how basically does it work?

Ann-Marie Alcántara: In the most simplest of terms, it's a artificial intelligence model that's trained on the internet, but only on information that it got together in its system before the end of 2021. If you were to ask it a question about Taylor Swift this year, it will not know the answer because it just doesn't have that information. How it works? You create an account on OpenAI, which is the organization behind the product, and then you just start typing away your questions or whatever you want it to do. You can truly ask it anything and everything, and it will give you some sort of answer or some sort of response.

Brian Lehrer: Can you give me an example? Give a developed example because this is not like asking Siri who was elected president in 2020, and it'll pop up with an accurate answer, probably more truthful answer than some Americans would give, but that's a simple kind of question. We're talking about giving it complicated tasks that spit out term papers and movie scripts and all kinds of things. Can you give us a developed example of how somebody has used this that might be emblematic?

Ann-Marie Alcántara: Yes. The story I co-wrote with my colleague Dalvin Brown, and he spoke to a software developer in Portugal Javi Ramirez, who said he used it to solve a complex coding problem. For those of us who are not software engineers, myself included, sometimes it can involve a lot of research on your own and an hour going through Google and responses.

Mr. Ramirez said that ChatGPT was able to answer his question in five minutes. If you want to get a little existential, I've also asked the chatbot who am I and it responded with cannot answer a question about who I am, and only I can know and define myself, so definitely not the type of response you would get from Siri or Alexa or even Google.

Brian Lehrer: It seems like it's become an overnight sensation. So many people that we're hearing about are using this app now. Why is it blowing people's minds just a couple of weeks out of the box?

Ann-Marie Alcántara: I think we, as a society, we've been hearing about this nebulous, "Artificial intelligence is going to take over our lives and replace jobs and do all these crazy things," but we've never really seen it in a consumer way. We've never been able to play with it ourselves. You're imagining a scientist or some researcher working on it. I think it's gone so viral and picked up with so many types of people because you and I can just hop on right now during this call and play with it.

We don't need a background in computer science, we don't need someone helping us, we just literally need an account and a creative mind to ask the chatbot any type of question. I think it's just the possibility of artificial intelligence in our lives finally becoming more tangible and real to us.

Brian Lehrer: Well, people who work in creative fields, language-based fields, even coders, who are always in such high demand, are starting to fear what this technology might mean for their own job stability. What are the capabilities of ChatGPT to even code itself?

Ann-Marie Alcántara: It's really unclear so far what it could do, and the future of work, and how it'll impact that because it's only trained on so much information, so it doesn't always have the most correct answers or some of the answers it provides are still a little vague and sound really smart and might impress your boss, but might not have concrete details that you need. Some experts we spoke to for this story have said that, if anything, it's like a brainstorming tool right now, to get yourself to think differently and to just kickstart your brain and make you get out of that writer's block or whatever mental block you might be having with a project.

Brian Lehrer: Is it a more sophisticated search engine than Google because several reports, Washington Post, New York Post, have said that ChatGPT could replace Google, which is-- I know some people use other search engines, but it is a very dominant search engine in our lives. Does it have the search engine capability that this would be the first thing maybe ever to rival Google's basic function?

Ann-Marie Alcántara: It might. That might be a possibility just based on how it responds to your questions. It does maintain history. If you stay in the same chat session, it will reference previous messages that you've asked it. If you asked it, "Are werewolves real," then the next question, "Are unicorns real," then you wanted to ask it again, "Hey, why won't werewolves and unicorns play with each other," or something like that, It'll remember that you've already asked that and give you an answer based on that, which is really something we haven't experienced and done much with. Unclear to say about [crosstalk].

Brian Lehrer: Let me jump in right on that point because is that different than how Google's and other tech sites algorithms do remember what we've done in the past and searched in the past and sites we visited in the past? We always talk about, in the politics sense, how we get driven further into our echo chambers of people we agree with because these algorithms remember where we've been. Is this different than that even?

Ann-Marie Alcántara: A little bit because there's a slight twist to it. With ChatGPT, it can be sassy, it can refuse to answer questions, it can admit mistakes as well. Unlike most search engines, it won't be too sassy with you. It won't say, we won't answer this question because it's inappropriate. It'll still try to surface you some answer, and just give you warnings about it whereas ChatGPT will respond and say like, "I can't do this, try again," or "Ask it a different way," or "I just refuse to answer this."

Brian Lehrer: It's got morals. Yikes.

Brian Lehrer: It's debatable. I think one person I spoke to for this story said if anything it's immoral. It doesn't know right or wrong, it's just simply answering whatever you want it to do.

Brian Lehrer: What did you mean by, it admits mistakes?

Ann-Marie Alcántara: Depending on your question, it might say, "Oh, I don't have this personal knowledge about this," or "I can't answer this." It's acknowledging that it just can't answer what you want it to do, what you're asking it. It's almost like it has a personality and knows that it can't do it.

Brian Lehrer: It acknowledges its limitations, but Siri will do that too.

Ann-Marie Alcántara: Yes. The smart assistants will do that. Google itself, if you search and your question is too complicated, it'll surface something at the end of the page. "These are the best results I've got," but this has that personality tone with it a little bit which I know Siri and Alexa can have that sassiness to them as well, but it's just different. I'm not an AI expert I don't know how to put it in those wonderful AI terms but it's just--

Brian Lehrer: Somehow next level.

Ann-Marie Alcántara: Yes.

Brian Lehrer: All right, here's Robert in Brooklyn who says he's a college admissions worker and a colleague has been messing around with ChatGPT. Robert, you're on WNYC. Thanks very much for calling up.

Robert: Hey, Brian. Thanks for taking my call. This has been all the talk amongst a bunch of the applicants I work with this week and some other college advisors I know. A good friend of mine tried it out and he sent me the essay and it was beyond bland. There was just no energy, no personality to it because I've read thousands of essays. I read it and my first thought was, "Oh, this just feels like AI created it."

We concluded that this will probably work for students in high school or college writing essays for class, more technical papers, but for a creative endeavor, like a college admissions essay, I don't see it working. Although my applicants are joking about it, this week a bunch of them joked about it and they will get people to use it. Some will create essays with it but I don't see an admissions officer at Yale or Harvard reading one of these essays and not pointing out immediately that it just feels generic.

Brian Lehrer: Could a student, an applicant potentially use it to generate, let's say the shell of an essay based on a question, maybe a really sophisticated question that even includes like, "Write an essay about my life experience," because you want college essays to be somewhat personal often. Write an essay about my life experience, which included this, this, this, and this that I did, and make it persuasive in a way that promotes my worthiness. I think you can even ask ChatGPT to do things like that. Then maybe it'll come up with something bland as you say, but then the student could use it as the basis of something to then touch up. You think that's possible for what you could tell?

Robert: Absolutely, that it is possible to come up with a structure for something and then personalize it. Just like you buy stock plans to build a house and then you use those blueprints and then through the process of building the house, you make it more original. You tailor it to your own needs but anybody, any applicant who would use this software is already giving up.

They're already saying, "I can't write a persuasive admissions essay on my own without the help of AI." I don't know how they take this, in my opinion, bland and generic product of this AI and then turn it into something that feels special or original or memorable. I don't see that happening. If you've already immediately said, "I'm going to give up and just use this AI," then you're not using your own creativity

Brian Lehrer: Then you're not getting in at least to whatever school you are in the admissions office for. Robert, thank you very much. Very, very interesting insight. Listeners, if you're just joining us, we're talking about this new artificial intelligence tool that is all the rage since it was released just I guess a little over a week ago, called ChatGPT that's apparently this next-level thing that can answer much more complicated questions than Siri or Alexa and do complicated searches like Google on steroids and all these things.

We're talking about the potential implications as this becomes very popular very quickly with Ann-Marie Alcántara, internet culture reporter for The Wall Street Journals' Personal Tech bureau who has written about it and inviting your experiences. Tell us about your early tinkering with ChatGPT. 212-433-WNYC, 212-433-9692, or tweet @BrianLehrer. Ann-Marie, first of all, what were you thinking during that brief conversation with a college admissions worker?

Ann-Marie Alcántara: I think it aligns with some of the conversations I've had with AI experts as well about these products. My colleague Daven Brown, he had spoken to Sarah Hoffman, who's the vice president of AI at Fidelity Investment, and she made a similar statement that she also tinkered with it to write a research presentation. While it was impressive, it seemed dated to her. To her, she saw it as more of a brainstorming tool and definitely to lay down the groundwork for whatever you need to do, whatever projects you're on. Then you can obviously build on it but if you're having writer's block or whatever mental block, you can maybe use this to jumpstart your brain.

Brian Lehrer: ChatGPT, could you write me a set of interview questions for Wall Street Journal reporter to talk about ChatGPT? I should have tried that last night but I didn't. Maybe I'll try it for Monday's show. Here's somebody else who has a kind of emotional attribute I think, to attribute to ChatGPT. Edward in Manhattan, you're on WNYC. Hi, Edward.

Edward: Hi. Oh, ChatGPT is very confident in its answers if you read [inaudible 00:17:47].

Brian Lehrer: We're having a terrible connection. I want to be able to hear what you said and repeat it to everybody so they can hear but forgive me for jumping in. You said ChatGPT is very confident in its answers, keep going.

Edward: That if you give it garbage, it can give you garbage regardless. You still have to know what you're talking about. If you're in a professional field, you would still have to review its answers no matter what. I already know other students who have been using it and have cut down their essay writing time in half. Thank you.

Brian Lehrer: Cut down their essays writing time in half and garbage in, garbage out. Nothing new about that, Ann-Marie, but that's interesting. Students using it in the way we were speculating about before, with the caller from the admissions office. Using it to maybe generate a potential structure for something and then running with it. Here's David on the Upper West Side who wants to raise a potential for the flip side of plagiarism using it in a good way. David, you're on WNYC. Hello.

David: Hi. I was just wondering if maybe the opposite case would be, 54% of adults in the US I heard are not considered literate or they don't have pros literacy above what's considered in the 6th-grade level. That's 154 million people. Obviously, we want younger people to work on their literacy, but older people could benefit enormously with this.

Brian Lehrer: Thank you very much. Is that the first you've heard of that Ann-Marie, ChatGPT as a literacy tool?

Ann-Marie Alcántara: Yes, I haven't seen that conversation at all. It's been mostly the opposite of it's going to take away our job, but definitely, the flip side of that is it could help people and then maybe get them into jobs or careers that they might have not had a chance in before.

Brian Lehrer: Here's a post from a listener on Twitter who says, "Already worried that you just called ChatGPT intelligent." Of course, it's artificial intelligence, that's the category but, listener writes, "I really hope you also talk about how problematic it is not to mention how often it gets things wrong." I have read that a few users have pointed out how easy it is to coax the technology into spitting out factually incorrect statements and that the only way to know is if you already know the correct answer. Have you run into that?

Ann-Marie Alcántara: Yes. You've seen this across Twitter as well, where experts in their fields have done precisely that. They've asked something that they know, they've done their research on, and it has the wrong answer or will say something that's part right, part wrong. It's only upon you to know whether or not it's correct or not. I will say that I think the chat GPT team has said when you sign up for it, that this is just like a research project still. I forget the exact language, but it's something along those lines. There's a bit of a disclaimer of this isn't in its final stages whatsoever.

Brian Lehrer: By the way, did we establish what the GPT stands for? It's called ChatGPT. What is GPT?

Ann-Marie Alcántara: That's a great question. I can't recall at this quite moment because my mind is more been focused on the bigger picture of what it is, but I forget what it actually stands for.

Brian Lehrer: Okay. We're going to look that up probably on some old technology platform and try to come up with the answer. Laura in Bedford, you're on WNYC. Hi, Laura.

Laura: Hi, Brian.

Brian Lehrer: You have a story for us?

Laura: Are you able to hear me?

Brian Lehrer: Yes, I can hear you. Can you hear me?

Laura: Yes. As a family, we tried ChatGPT the other night with my high school-aged daughter. She had an in-class essay for English to analyze the poem [unintelligible 00:21:57] and a space of about one minute, the app generated an answer that she felt was better than the one she had generated in class. Then she gave it a second challenge. She had just spent about an hour on an essay for her history class about why Britain was able to succeed in Industrial Revolution.

Again, in the space of about two minutes, the app generated an essay that was probably as good as the one she'd written and she is a good writer. I honestly felt completely demoralized. It was almost like I imagined a machinist in the 1800s would've felt who had spent years learning how to carefully craft small metal pieces. Then suddenly, there was a machine that could stamp them out several a minute at more consistency and higher quality than he could. I just never felt like as a knowledge worker, I guess, I always felt immune to technology overtaking us in that way. It's really preyed on my mind a little bit, to be honest.

Brian Lehrer: Can you give us an example of if you want to get this personal, what you do as a knowledge worker and how you think this technology could be a threat to that field?

Laura: Well, actually, I both code and then I manage coders. To be honest, people management is clearly not there yet. The ability to research, to write, to analyze are all things that I have felt proud of and prized and tried to develop in my kids. Then like I said, watch this AI do something that I just never thought of a machine being able to do. We also use part of the app that will generate art. I asked it to, I can't remember generate some type of picture in the style of Chagall. I have to tell you, the pictures it generated were quite good.

Brian Lehrer: I guess if we're talking about art, maybe it can mimic the style of Chagall, but can it come up with anything original that people might want to keep or view and say, "Hey, this is by my ChatGPT and it's original?"

Laura: Well, I want to say no and I listened with interest to the admissions officer because of course that's something with a high school student we're thinking of, and I'm just imagining all these kids getting on and maybe not for, as you said, personal essays, but for the types of things, history, and even English class, I have to say the quality was surprisingly high and it was awfully fast.

Brian Lehrer: Really interesting. Laura, thank you very much. I guess when it comes to the art, they're already a lot of things you could play with that are random image generators or that will take whatever direction and come out with some images or moving images or still images that might be interesting for people to look at. I don't know if that's so different, but it is interesting Ann-Marie to hear her talk about being a knowledge worker as she put it.

Then it turns out even working in coding and feeling like this could potentially be a threat to that field. The Atlantic has a piece out already with this brand new technology out there only 10 days or whatever it is titled The College Essay is Dead, arguing that this technology will change academia in ways that we're unprepared for. With its coding, data processing, and creation abilities, the New York Times is questioning if ChatGPT is "Coming for the skilled jobs".

I don't know, maybe this is all overblown, it's so new, but that's some of the media reaction. We have a tweet that says, "Hey," this is in the style of a request to ChatGPt. It says, "Hey ChatGPT, would you tell Brian Lehrer about why ChatGPT isn't as great as people think?" You think people are overreacting? You know as an internet culture reporter, that there's such a history of G Whizz journalism that we don't want to be a part of. G Whizz journalism about various new technologies.

Then that technology, whatever it is, either turns out to underperform or turns out to have massive downsides. We don't want to fall into G Whizz journalism. There are already skeptics out there we should acknowledge about its potential even though there are all these articles at the same time about, "Oh my God, look at this potential."

Ann-Marie Alcántara: Yes. I will say, like we were saying earlier that OpenAI the company behind ChatGPT, they've added that disclaimer to their responses and that it's limited by its own material. Some questions that you ask it, it says, "Based on this information that I have, I'm not able to determine my own accuracy or evaluate my own responses." The chatbot itself is also telling people like, "I am not the end all be all, and I will not take away your job just yet."

Similar to what the knowledge worker was saying earlier, the coding website Stack Overflow, which a lot of engineers use to figure out queries and things that they're struggling with in their jobs. They've banned answers created by this chatbot because so many of the answers were incorrect. There's being some guardrails being put in place from other organizations to make sure the information we're all seeing is at least still correct and yes, we're not there yet where our jobs and what we think as humans is at risk yet.

Brian Lehrer: By the way, we did look up what GPT stands for in ChatGPT, and I wonder if you can give us an interpretation of what the answer is. The words are generative pre-trained transformer, so chat generative pre-trained transformer. I guess we know what pre-trained means, that still sounds like some human being programmed it but generative, well, then it can generate things on its own, I guess. Transformer, I guess is just a reference to it being technology. Do you want to say anything about GPT, generative pre-trained transformer?

Ann-Marie Alcántara: Yes, so my understanding of it as well is sort of that phrase is to convey that it's a program that can write like a person and write like you and I would answer something in that-- Like you said, pre-trained, it's using that information we talked about at the beginning of our conversation. Then generative means that it's iterating on every answer and question that you're giving it. I dont know if I meddle that up a little bit, but that's my take on it.

Brian Lehrer: Judah in Queens, you're on WNYC. Hi, Judah.

Judah: Judah.

Brian Lehrer: Judah, I'm sorry. Somebody gave me two Ds.

Judah: That's fine. I get it all the time. It could be Jonah, it could be Joddah. Here's my thinking is that I'm in the humanities. I am a graduate student. I'm a professor or an adjunct, let's be honest, and a poet so I'm really part of the bygone world. I don't really think we should be bemoaning that people aren't going to be writing crappy essays about I don't know, whatever's poet. [crosstlalk] Yes, exactly.

Brian Lehrer: Take that call as example about her daughter. Go ahead.

Judah: Look on my work she might be in despair. I think that that's totally fine because most people do despair at reading those and at writing those. I think as an educator if people are able to generate those essays quickly enough then perhaps we can abandon them and use our time to discuss more interesting things. Like, why do people write poems or what is the nature of language, or I don't know, liberatory potential of language beyond just drudgery? Instead of being knowledge workers in some narrow way, what can knowledge be used for that might go beyond I don't know the simplex of writing a bad history paper that no one wants to write and no one wants to read?

Brian Lehrer: That's a deep thought. Maybe something that does what we take as advanced so easily or parts of what we think of as advanced so easily, makes us reinspect our assumptions about creativity, about the nature of what it is to be human, and somehow do those things better. Am I hearing you right?

Judah: We're also in the context of a classroom. If you're sitting in a classroom and analyzing a poem with people and the only thing that you're doing is essentially, I don't know, a formal analysis, then I'm not sure that that does anything that leads us to question like why have people written poems? Why do people continue to write poems? Why do we care about being in a room with each other?

I think if we can be freed from that drudgery which is not a guarantee. Usually, these sorts of technologies just lead to greater drudgery. I wonder if we can put forward the idea that instead of the person in the 19th century who is a machinist saying, "Oh my God, my job is being replaced." Plenty of people were perfectly happy to have their jobs replaced in the 19th century. The question is are we working toward a better and more equitable world or whether we just end up doing worse and worse and whether we end up being mindless drones?

Brian Lehrer: Isn't this what Andrew Yang and other people talk about that we need to set up a universal basic income system because technology is going to replace so many people's jobs?

Judah: I think Andrew Yang though his version of it is I don't know deeper into the capitalist fold, not like questioning, I don't know, why do people work to begin with.

Brian Lehrer: Judah, thank you. Deep call, thank you very much. If in the capitalist fold, in the capitalist mold, a machine turns out to be a better investment as an employee than a human being then they have to look to maybe something like universal basic income if that happens on such a widespread basis. Let me get two more calls in here that look very interesting as we talk about this new technology that everyone seems to be talking about and people are adopting and playing with at least at a very fast pace since it came out at the end of November called ChatGPT. Brian in Atlanta, you are on WNYC. Hi, Brian.

Brian: Hey, thanks. My wife and I just started a production company and we're trying to raise money for an option we script. I asked this thing how can I raise X dollars for a movie production and it gave me a really bland answer. Things that I would already think about. I'm creating a website for this production company and I'm working with something called SVG Files.

We just got a logo made and it's a pretty complex, there are a bunch of HTML tags and it's not working for me so far. I put in the SVG paths on this thing and it failed. It didn't tell me why. It said it's difficult to know why it's not working. Then I asked it to draw a picture based upon our name which is Funzy Whistle Brit's Productions and it turned me down, it wouldn't draw the picture.

Brian Lehrer: Did it say why?

Brian: It said it's not able to do it. I said are there other systems to do it and what are the names of them? Then it just listed a couple of them. One's called GAN Paint, D-A-L-L-E, but it said itself it could not do it because it says I do not have the ability to have friends or interact with other entities on purely a computational system designed to process generate text-based responses.

Brian Lehrer: It didn't know what a Fungi whistle was.

Brian: It's an F-U-N-Z-Y, actually Funzy Whistles. It was what my grandfather named me as a kid.

Brian Lehrer: It didn't recognize that and interesting you didn't ask it to write your movie. You asked it how to market your movie. How to raise money for your movie. Brian, thank you very much. One more. Chuck in Union City. You're on WNYC. Hi, Chuck.

Chuck: Hi, great segment. I tried it just this morning as you were speaking and I put the same essay question three times and it kicked back the exact same answer. It would seem that a teacher or an admissions officer could easily check this. Just put your essay in and it's going to give the same answer unless a different person putting the same question would get a different answer but when I tried it myself three times, it came back with the identical answer.

Brian Lehrer: Chuck, thank you very much. We thank Ann-Marie Alcántara, internet culture reporter for the Wall Street Journal's Personal Tech bureau for being our guide to these first days and some people's reactions, concerns, hopes, fears to ChatGPT which so many people are using already. Ann-Marie, have you tried to do anything with it? Wall Street Journaling, did you ask it to write this article about itself for you or anything like that?

Ann-Marie Alcántara: I have not though. I have made several jokes with my editor about my actual story when I turned it in. I just don't want to step on those toes just yet. I'm trying to think for myself for a little bit longer while I can.

Brian Lehrer: We'll be the last two holdouts. Ann-Marie, thanks for joining us. We really appreciate it.

Ann-Marie Alcántara: Thank you.

Brian Lehrer: Brian Lehrer in WNYC, much more to come.

 

Copyright © 2022 New York Public Radio. All rights reserved. Visit our website terms of use at www.wnyc.org for further information.

New York Public Radio transcripts are created on a rush deadline, often by contractors. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of New York Public Radio’s programming is the audio record.