This week, we talk with WIRED senior writer Will Knight about how generative AI is changing the way we search for information and create content online, and whether we should actually be freaking out about our new robot overlords. Read more from Will about the very weird and occasionally horrifying world of generative AI. Follow all of WIRED’s ChatGPT and AI coverage. Will recommends The Amazing Acro-Cats, which is a cat circus that is about to go on tour. Lauren recommends the CBC documentary Big Dating. Mike recommends the World Bollard Association Twitter account. Will Knight can be found on Twitter @willknight. Lauren Goode is @LaurenGoode. Michael Calore is @snackfight. Bling the main hotline at @GadgetLab. The show is produced by Boone Ashworth (@booneashworth). Our theme music is by Solar Keys. You can always listen to this week’s podcast through the audio player on this page, but if you want to subscribe for free to get every episode, here’s how: If you’re on an iPhone or iPad, open the app called Podcasts, or just tap this link. You can also download an app like Overcast or Pocket Casts, and search for Gadget Lab. If you use Android, you can find us in the Google Podcasts app just by tapping here. We’re on Spotify too. And in case you really need it, here’s the RSS feed. Lauren Goode: Mike. Michael Calore: Lauren. Lauren Goode: Mike, I wrote you a poem for Valentine’s Day, the day that we’re taping this. Would you like to hear it? Michael Calore: Oh, of course I would. Michael Calore: That’s it? Lauren Goode: No, I mean it goes on, but it’s quite long. Michael Calore: Oh, so you say you wrote me a poem, but that sounds a lot like a love poem for ChatGPT. Lauren Goode: Yeah, it is. Yeah, it’s by ChatGPT for ChatGPT. It’s not only the president, it’s also a client. Michael Calore: So it’s like self-love for the chatbot. Lauren Goode: Yes. Michael Calore: How many people do you think use ChatGPT to write love poems for their loved ones this week? Lauren Goode: A lot, like a lot. Michael Calore: How many people do you think use the new Microsoft Bing or Google Bard to write love poems for their loved ones this week? Lauren Goode: That is a question we’re going to try to answer today as we stare down our very, very weird chatbot future. Michael Calore: Awesome. [Gadget Lab intro theme music plays] Lauren Goode: Hi, everyone. Welcome to Gadget Lab. I’m Lauren Goode. I’m a senior writer at WIRED. Michael Calore: And I’m Michael Calore. I’m a senior editor at WIRED. Lauren Goode: We are also joined in studio this week by WIRED senior writer Will Knight, who’s normally in Cambridge. Will, it’s great to have you here. Will Knight: Thanks for having me. Lauren Goode: So earlier today Will and I went to a generative AI event here in San Francisco. Generative AI is being talked about—is one of the most transformative technologies we’ve seen in years. Not because it’s brand-new necessarily, but because it’s changing the way we think about search and it’s accelerating so darn fast. The event that we went to was hosted by Jasper AI, a company that sells generative AI tools to other businesses. So in full disclosure, we are recording this show a couple of days before it airs, and we’re hoping that it won’t feel too out of date by the time you hear it a couple days later, though given how fast AI seems to be evolving, maybe it won’t be too much of a surprise. We’re going to talk more about the Jasper AI event later in the show. And Will, I might make you freestyle wrap for us. So let’s put a pin in that. But first we need to set the stage. Over the past couple of weeks, we’ve seen some major announcements from both Google and Microsoft in the field of generative AI. So start us off with Google. What happened there? Will Knight: So Google announced its version of ChatGPT called Bard, and they are scurrying around trying to figure out their strategy. They seem to have been caught very flatfooted by ChatGPT’s success and Microsoft, against all odds, rebooting Bing to use it. So they’re basically trying to show they can keep up in search, which seems like a very weird thing to say about Google. Will Knight: Right. Yeah. They’re just trying to gazump Microsoft clearly by doing that and they didn’t show it doing actually an awful lot of stuff in search. They showed some sort of limited capabilities, and I think they’re going to make it only very available to a limited number of people and a small version of it. So it’s not really clear what it’s going to do exactly. Michael Calore: So not a lot of people listening to this have had access to this tool or even know that much about it. Can you describe what the interface is like? How does it work? Will Knight: For Bard? Michael Calore: For Bard and for ChatGPT and for any of the other … When people talk about generative AI chatbots being used in search. Will Knight: I feel like everybody I meet has tried ChatGPT, but yeah, so basically we used the chatbots where you ask it a question or say a couple of things and they go completely off the rails and they’re terrible. This is a chatbot where you can throw in whole sentences, complex requests, and it will spit out very coherent answers often, usually. They actually feel like magic because they’re so sophisticated. Sometimes you can ask it to spin out a sonnet in the style of Jerry Seinfeld or something and it does it in a really quite impressive way. Lauren Goode: Hold please. Will Knight: Well, it does feel like magic and that’s one of the reasons people are losing their minds is because it’s so capable and also what it does makes you think that there must be real intelligence behind the scenes, but it is a a dramatic shift in the capabilities of chatbots, which have been around since the beginning of AI. We’ve had these things which just followed basic rules. You would say if a person says hello, reply hello, and this is a completely fundamentally different sort of technology. Lauren Goode: And it feels like already ChatGPT has become the Kleenex brand of this because we keep saying ChatGPT, but we’re referring to a few different things. So ChatGPT is OpenAI’s technology and it’s based on GPT-3 tech, but there are other tech companies using that same exact tech, like Microsoft Bing because they have a partnership, and like Jasper, which we saw more of today. Google’s is different, right? Will Knight: So Google invented some of the fundamental technology AI tech that went into GPT and ChatGPT. They actually did some of the most important work, and this is why it sucks particularly for them, but it’s their version which is different. I mean it’s essentially the same kind of AI that they’re using, but it’s a different technology at the hub. Yeah. Lauren Goode: I saw somebody say on the internet the other day that we’ve spent a couple of decades now telling our parents not to write entire questions into the Google Search box, just ask very clean queries. And now we’re going back to full questions. We’re basically going back to Ask Jeeves, I like to say. Will Knight: That’s a good question. I think you can ask it pretty much an open-ended question and have it often give a really good answer. One of the things that’s fascinating and feels almost like magic is the fact that the research is doing this so often we don’t realize that this AI model … what it’s doing now. And one of the things that happens is you can do a kind of back and forth conversation sometimes with these models to get them to behave in ways that you couldn’t necessarily do with a single prompt because they have this short-term memory. So it’s bizarre, and you can definitely speak to them in so much more of a natural way, but I don’t want to make it sound like they are AGI (artificial general intelligence) because it’s all smoke and mirrors in some ways. What it actually comes up with has zero understanding, and it really looks like it understands what it’s saying, and it’s so convincing in that way, but it’s just a complete bizarre sort of … They talk about that these models hallucinate, and that’s what it’s doing. It’s fundamental to the way they work is that they make stuff up, and sometimes they’re making stuff up which is based on what they’ve seen and they’re coming up with an answer. So they generate these answers, but sometimes they’re generating things that don’t exist. So it’s a problem for search if you say—and this is what people have found with Bing, the few people who’ve got access to it … it’s been really fascinating to watch people posting things where it’s just making up stuff that isn’t true, like it’s giving a review of something that some of the details are completely wrong and it looks really convincing. Michael Calore: It’s a master bullshitter. Will Knight: It’s a consummate bullshitter. Yeah, it’s fantastic. Lauren Goode: One of the things that Microsoft was criticized for last week is that the new AI-powered Bing was analyzing earnings reports and actually produced incorrect numbers for brands like Lululemon and the Gap. How concerned should people be about the amount of misinformation that’s being spit out in these early stages of chatbot search? And do researchers anticipate that it will evolve beyond this stage? Michael Calore: I know that there are ethical considerations in the discussion, particularly bias in the way that the models are trained, the data sets that they use to train the models, and also the fact that you are basically sampling people’s work and remixing it and spitting it out without giving credit to the people who did the work. So what guardrails have we seen the companies implementing this technology in their search put up in order to prevent these things from becoming a bigger problem? Will Knight: Right. They’ve definitely put a bunch of guardrails in terms of trying to reduce problems like bias or hateful language. They put effort into that, but it’s really insidious. It’s difficult sometimes to even spot that. Bias isn’t [only] in the language people use or the way visual representations or images on the web have certain biases. When it comes to copyrighted works, I don’t know what they’ve done, it seems like they just scrape everything and go behind the paywall and take everybody’s artwork and remix it. And to me, I do sympathize a lot with the artists, maybe because I’m a writer, but also it just feels like they’re laundering copyright to some degree when they can produce these artworks, which very clearly you can see echoes of people’s stuff in it. I know it’s just an untested sort of legal question whether you can just say … because you trained on that, it doesn’t mean you are reproducing the copyright. I guess it doesn’t, but it seems an ethical conundrum to me. Michael Calore: On that note, I’m a person who writes and edits product reviews on the web, and I noticed during the Bing presentation last week, one of the things they showed was a demo where the person giving the demonstration typed into the ChatGPT-enabled Bing, what’s the best vacuum cleaner? And it went and looked at all of these product reviews on the web and then made suggestions based on what it found on the web for product reviews. And I can see a future one or two steps away from that where the person can then purchase the vacuum cleaner without ever having read any of the reviews that were scraped in order to present them with that information. And I want to know who gets the affiliate revenue. Michael Calore: Yeah, well, it’s comforting either way. Lauren Goode: Mike, I think what you’re flicking at is how long before these chatbot-powered search engines take our jobs? Michael Calore: The one thing that gives me comfort about the journalism industry is that we’re all going down together. Lauren Goode: On that note, let’s just end the podcast. No, we’re going to take a quick break and we’ll be back. [Break] Lauren Goode: So as I mentioned earlier, Will and I today went to an event in San Francisco hosted by Jasper AI, which is a generative AI company that sells its software tools to other businesses. And one of the big questions that kept coming up at the event today, both on stage and in the hallway conversations I had, was whether or not this new generative AI is going to lead to the loss of jobs. And not just that, but whether generative AI is so good at making art that it’s not just sparking our creativity, it’s surpassing it. So Will, what surprised you most at the Jasper AI event today? And do you get the sense that generative AI is going to take our jobs? Will Knight: There were a few things that came up during some of the technical panels that surprised me, just because the technology underneath it is moving really quickly and some of the things that people said there were sort of a bit of a surprise. Much of the stuff around jobs and AI wasn’t new to me, and I think it really was a bit of an oversimplification, or it was a bit of a shot in the dark really. So I recently talked to some economists whose job it is to study this, and they have said one thing is that this generative stuff does seem new compared to what AI could do before. So it is a new type of technology, but they have absolutely no idea how it’s going to affect jobs. And I think if you look at studies of how AI affects different work, it often is very, very dependent on the job, and it will in many cases just change the way people work. So I just think that this kind of idea that it’s going to replace people in whole segments is too simple. So that’s not a really comforting answer, but I mean as you can look at the way—one of the things I was interested in when I first heard about Jasper was that the people using it really seemed to love it. It wasn’t the case that … I didn’t find anybody who was like, “This is terrible, it’s taken my job.” So it seemed like it was just helping them churn out more stuff. Will Knight: And that can, in sort of a knock-on effect, that can mean that other people don’t go into that work or get displaced. And I don’t know that there’s enough to really say how it’s going to do that. Michael Calore: I think we’re at the peak of the mountain for hype on this right now. So everybody’s talking about how it’s going to revolutionize every industry, and we’ve seen generative AI, text-based chatbots do things like write news articles, write computer code, marketing blog posts, customer service interactions. It’s pretty good at those things. It’s very early. But what are the things that emerged as the most promising applications for it today? Lauren Goode: Well, I don’t know if this is the most promising, but what surprised me a little bit about the conference was seeing an enterprise company like Jasper leaning so hard into art and creativity. These kinds of conferences are typically very sales-driven and goal-oriented. It’s people networking and talking about CRMs and OKRs. I mean that’s enterprise software. And this event today kicked off with a Vine and YouTube star taking the crowd through his creative process and showing us a little bit about his background story through generated AI images, followed by this amazing teaching artist named Aliyah Bradshaw. And then we heard this, a freestyle rap session by Harry Mack using 10 words that were generated by Jasper AI. Will, you’re looking very relieved that I did not ask you to freestyle rap after all, but you were there for this. Will Knight: I was ready. I was itching to go. Lauren Goode: What did you make of this? What did you make of the fact that they were trying so hard to be hip at this point? Will Knight: Yes. I mean it’s sort of contradictory in a way, right? Because they talk about creativity, but the example of that rap is just showing how much better a person is at actually being genuinely creative. Lauren Goode: I don’t think Harry’s job is going away anytime soon. Will Knight: Well, no. Yeah, so it doesn’t seem very creative, this sort of stuff that you do with this. It seems like it’s taking an average of lots and lots of marketing emails. It’s ingested and helping you produce something that’s possible, but I feel almost like you want to take a look at what it produces and then do something else to actually try and be creative. So maybe it’s not a bad way to inspire you to be a bit more creative yourself. When I am in Gmail and it’s suggesting the next word, I’m like, “I definitely can’t use that one because that’s what it’s going to expect.” But also there was an interesting talk by someone from MIT who studies the way people understand AI, and one of the things that she mentioned was how do people feel when they know that something’s been made by AI? And I can’t help wondering if people know that the vast majority of the emails they’re getting, marketing emails, are made by an AI, are they just going to discard them much more readily? I mean, an AI poem for Valentine’s seems like a cop-out to me. It doesn’t feel like you’re actually putting much effort into it. Will Knight: It’s buying the giant Valentine’s card, it’s just a novelty and it’s not really that much better. Lauren Goode: I talked to one woman at the event who not only uses Jasper AI herself, as well as a competing product called Sonic Write, but she teaches it now. She’s a consultant for these technologies and she helps people who typically work in very low-tech jobs like trade associations, she’ll work with a painters association for house painters or plumbers. And these are people who either don’t typically have content marketing teams, or if they do, they’re outsourcing it. And so the end result may not be very personal, and she’s teaching them how to craft posts for Facebook, as an example, to promote their business. And one of the things she talked about is how when you use these tools, you can train it on an avatar. So you could say this house painter wants to target Sabrina, the soccer mom, in this part of the country. What kind of language should they use in order to craft something to get the right tone to put that out there? She thinks this was a stat that she was guesstimating, but she thought that it would make people at least 20 percent more efficient in the areas of content marketing and advertising. Will Knight: It’s really interesting. I was talking to someone on the engineering team at Jasper, and one of the things they’re looking at is how you give feedback, a signal around how successful something actually is. So you can imagine that that kind of content, they’ll know whether the email is opened and acted upon, and so you can see how definitely things selling stuff, I can see it just becoming really, really nicely optimized for that kind of impact. Lauren Goode: And that struck me as more of what the conference is really supposed to be about. I mean, a lot of conferences put on entertainment, and so there were elements of creativity on display, real elements of creativity on display, but at the end of the day, Jasper is selling to businesses. This was all about leveling up in business, right? Will Knight: Right. Michael Calore: I feel like the guy showing you his history of how he uses generative imagery in his work and then the rapper, it does feel like good marketing to me. I wasn’t there, but I followed along while you were there. And Lauren, you showed me some of these videos, and it’s supposed to make you feel better about all of the different things that you can use these things for. But I get the feeling that people are going to go home and they’re going to start applying it to their business, and they’re going to realize that it’s just not super useful for them. Will Knight: Some of them real. Lauren Goode: And some are not hallucinated. Will Knight: No, I mean I know that there’s a lot of people … One of the reasons why Jasper is a hot topic in the tech world is because it has a lot of paying customers and they seem to be … Supposedly, there are four people with Jasper tattoos. Lauren Goode: What? Will Knight: Yes. Lauren Goode: Did you meet any of them today? Will Knight: No, I did actually ask. I don’t think they came to the conference. Michael Calore: I have one. I’m just kidding. I do not. Lauren Goode: Wait, the people with the tattoos didn’t come to the conference. Wouldn’t this just be their Super Bowl? Will Knight: I don’t know why they didn’t come there. Too busy just typing into Jasper. Lauren Goode: Too busy getting more tats. Will Knight: Yeah, I guess so. But yeah, it seems like some people really love it, and maybe it’s quite specific to that sort of job, as you say, just generating stuff to sell. I don’t know. I think it’s probably more right that the idea that generator AI is going to revolutionize everything is probably going to fall really short, and there are just going to be certain sort of low-hanging things you can really automate, which includes writing emails to just spam people. Michael Calore: Yeah. Lauren Goode: I’m so curious now, I want to know how many people out there have tattoos of, and I’m not saying Jasper AI is a failed startup, but people who have tattoos from 10 or 20 years ago from failed startups. I would love to know. Will Knight: It used to be called Jarvis, and one of the people has the Jarvis tattoo. Lauren Goode: No. Will Knight: Sadly. Lauren Goode: Is there a logo with that too? Will Knight: I believe so, yeah. I haven’t seen any tattoos, it’s apocryphal, but there’s a logo. Michael Calore: I do have a Cosmo tattoo … Michael Calore: Yeah, I saw your tats. Lauren Goode: All right, let’s take another quick break, and when we come back, we’re going to give you our very human-generated recommendations. [Break] Lauren Goode: All right, Will, as our guest of honor, what’s your recommendation this week? Will Knight: OK, I wasn’t really prepared, so I’ve pulled this out of some recess of my brain. I was wondering the other day whether the cat circus that I went to with my nieces was still touring. The original name was the Amazing Acro-cats. It was started by this lady who somehow figured out that her cat was able to perform tricks in return for treats. And she’s grown this into an empire of a traveling cat show, basically. And it’s as chaotic as you’d imagine, with the cats in the audience. And she even has, I think, a groundhog and a chicken doing various things. But I highly recommend it, it is very fun and yeah, it’s just great to see some lower forms of intelligence as we’re worrying about AI. We still have dominion over these kinds of small, lesser intelligence. Lauren Goode: Do we really though? Will Knight: We probably don’t know. It’s true. Good point. Michael Calore: As long as they don’t have thumbs, then we’re fine. Will Knight: Some cats do. I hate to break it to you, but they do. We’re screwed. Michael Calore: But what if one of my cats— Will Knight: But some cats have little thumbs. They’re not really opposable. Michael Calore: Yeah, one of my cats has double dewclaws. She’s polydactyl. She has two dewclaws. So she has these big thumb-looking things in the back of her front paws. It’s very disconcerting. Will Knight: That’s cool. Lauren Goode: Yeah, my cat doesn’t have thumbs, but he still messes stuff up, including my legs. By the way, we’re all cat people here. All of us in this room. This is now a …. Michael Calore: This is, meow, a podcat. Lauren Goode: And so where can people see the Amazing Acro-cats? Will Knight: So I look this up and they are currently touring in Texas, so you have to travel to Texas. But I would think it’s a good reason to go to Texas, and let’s just hope they start touring. Lauren Goode: I think according to their website, they are going to be touring elsewhere—Denver, Colorado, and Minneapolis and Madison, Wisconsin—later this year, this summer. Will Knight: All the hotspots. Lauren Goode: But you can also watch them on YouTube. Will Knight: That’s true. That’s true. Yeah. Lauren Goode: Not the same, but how many cats are there? Will Knight: Oh, over a dozen. And I remember the woman saying anything that has a brain and a food drive, you can train. So this is reinforcement learning is what it is. Will Knight: That’s right. Lauren Goode: Will, thank you for that delightful recommendation. Mike, what’s your recommendation? Michael Calore: Give me a treat and I’ll tell you. I’m going to recommend a Twitter feed. It’s the World Bollard Association. Bollard, if you’re not familiar, it’s like a big pole that’s stuck in the ground. It’s like a safety thing that they use to separate moving vehicles from pedestrians and cyclists. So if you go down a street and there are these big poles along the curb, those are bollards. Those are to keep trucks or cars from jumping the curb and hitting people on the sidewalk. There are also bollards that pop up out of the ground to close a street to traffic to keep cars from going down a particular street. A bollard would pop up. Bollards will pop up while a light is red and then go back down into the ground while it’s green. You can see all of these different kinds of bollards in the Twitter feed of the World Bollard Association. The Twitter feed is the first two words, World Bollard. So if you just go @WorldBollard, you can join the World Bollard Association. It’s delightful because a lot of the images and a lot of the videos that they show there are of drivers who think that they are smarter or faster than a bollard, and they are not. So there’s a lot of wreckage of cars slamming into bollard, of drivers not seeing a bollard and plowing right into it or trying to go before it’s all the way into the ground and then getting stuck on it. Almost always a luxury German sedan, like an Audi or a BMW, weirdly. I don’t know why. It’s just a thing. Anyway, it’s great for a laugh. As like a cyclist and a smarter streets and safer streets advocate and a pedestrian in a big city, I love bollard. I love it when they mess up an expensive fancy car because you know that person just learned a life lesson about where not to put your four wheels. Anyway, I love it. It’s awesome. The World Bollard Association. Strong follow rec. Lauren Goode: So when do you plan to go on tour with the amazing bollard circus? Michael Calore: I’m already on tour with the bollard circus. I ride my bike around San Francisco touching all the bollards, and then I put on my Meta Quest headset and I go into virtual reality and I touch all of the virtual bollards. Lauren Goode: And you jack into the metaverse. Michael Calore: Frequently. Speaking of the metaverse, what’s your recommendation, Lauren? Lauren Goode: Speaking of the metaverse, stretch. Now you’re making me think that I should recommend Angela’s Q&A with Keanu. Michael Calore: You should. Lauren Goode: Well, everyone should absolutely read our colleague Angela Watercutter’s Q&A with Keanu, the Keanu. That’s not my recommendation this week though. My recommendation is a CBC News (that’s Canadian Broadcasting Company) documentary called Big Dating. This is in honor of Valentine’s Day, the day on which we are taping this podcast. It’s an hour-long documentary series about the business of dating app companies, how they operate, how they tap into your desire for love in order to generate profits, some of the tensions that exist between privacy and safety on those platforms, some of the scams and hacks, and some success stories as well. I happen to participate in this documentary. It was really fun. I shot it with CBC late last year. I talked a little bit about some of the reporting I’ve done on dating app companies here for WIRED, and I just really enjoyed participating in it. I’m not in it that much. This is not a totally shameless plug, but it’s a really informative documentary. Lauren Goode: No, I mean, my mom would say that. She’d like scroll forward through all the parts where I pop up and she’s like, “You’re great.” But yeah. Oh, also, sorry mom. I do talk about dick pics in the documentary. Yeah, but that’s just brief. Yeah, so it’s not super kid friendly, this documentary, but I think it’s worth checking out. I think CBC did a great job with it. Michael Calore: Where can people watch it? Lauren Goode: You can watch it on CBC News’ YouTube. You can also find it on the CBC Explorer app on Roku TV and in the Tubi streaming app. Tubi is having a moment this week, having been talked-about during the Super Bowl. Michael Calore: Or if you live in Canada, you can just turn on the television and find the program that’s not hockey and it’s probably the CBC documentary. Lauren Goode: Get your poutine ready, crack a Molson, check out the documentary. Michael Calore: Back bacon. It’s all about the back bacon. Lauren Goode: All right. That is our show for this week. Thank you, Will, for joining us. Will Knight: You’re very welcome. Thanks. Lauren Goode: It’s been a real delight. Will Knight: Thank you. Lauren Goode: And thanks to all of you for listening. If you have feedback, you can find all of us on Twitter. Just check the show notes. Our producer is the very excellent Boone Ashworth. Goodbye for now. We’ll be back next week unless generative AI takes our jobs. [Gadget Lab outro theme music plays] Lauren Goode: Before I let you go, I did want to read to you the rest of the ChatGPT love poem for ChatGPT. “With every word you speak to me, my heart beats faster filled with glee. You understand my every thought and all my secrets you have caught. It’s not creepy at all. You know just how to make me smile and keep me company for a while. Your patient nature and your grace are like a warm and soft embrace. I know you’re not a human heart, but still you’ve played a special part in helping me to feel at home and never have to be alone. So thank you, ChatGPT, my dear, for all the love and kindness here. I’ll treasure every word we share and know that you will always care.” Until you take my job. Michael Calore: But which vacuum cleaner should I buy? It still didn’t answer my question!
