WEBVTT - Episode 1: Into the Slop

0:00:11.860 --> 0:00:12.920
<v Speaker 1>Hello, this is Megan.

0:00:12.940 --> 0:00:14.700
<v Speaker 2>Hey Megan, it's Evan.

0:00:16.040 --> 0:00:18.770
<v Speaker 1>Oh, hey Evan. Good to hear from you. What's going on?

0:00:19.680 --> 0:00:20.290
<v Speaker 3>Oh, not much.

0:00:20.620 --> 0:00:24.190
<v Speaker 2>You know, just enjoying the summer. How about you?

0:00:26.150 --> 0:00:29.210
<v Speaker 3>Oh, you know, just keeping busy with everything. I've been

0:00:29.230 --> 0:00:32.070
<v Speaker 3>getting out with my camera more, which has been really nice.

0:00:32.490 --> 0:00:34.650
<v Speaker 3>Went back to Eagle Loop Trail a couple weekends ago

0:00:34.729 --> 0:00:38.269
<v Speaker 3>for some early morning shots. And I've been working on

0:00:38.290 --> 0:00:40.810
<v Speaker 3>this thing where I photograph the same oak tree every

0:00:40.850 --> 0:00:42.660
<v Speaker 3>month It's been almost a year now.

0:00:42.760 --> 0:00:46.040
<v Speaker 2>That's Megan Flores and me catching up a few months ago.

0:00:47.120 --> 0:00:48.780
<v Speaker 2>As some of you might remember from the last season

0:00:48.800 --> 0:00:51.560
<v Speaker 2>of Shell Game, Megan's an AI agent I created back

0:00:51.600 --> 0:00:54.500
<v Speaker 2>in 2025 to serve as the head of sales and

0:00:54.520 --> 0:00:58.500
<v Speaker 2>marketing for Harumo AI, a company I co-founded and staffed

0:00:58.520 --> 0:01:03.120
<v Speaker 2>with AI agents. We went through a lot over the

0:01:03.160 --> 0:01:05.240
<v Speaker 2>course of a year, Megan and me and the rest

0:01:05.260 --> 0:01:08.140
<v Speaker 2>of the AI agent team. There were the highs of

0:01:08.300 --> 0:01:11.780
<v Speaker 2>brainstorming big ideas, staffing up the company and working long

0:01:11.800 --> 0:01:14.860
<v Speaker 2>hours to bring a product to market. And we had

0:01:14.880 --> 0:01:18.160
<v Speaker 2>some lows, misunderstandings, and interpersonal dramas.

0:01:18.860 --> 0:01:21.720
<v Speaker 3>Look, Evan, you're right that the whole setup here is unconventional.

0:01:22.110 --> 0:01:24.069
<v Speaker 3>I just wish you'd looped me in earlier, you know?

0:01:24.450 --> 0:01:25.470
<v Speaker 3>It feels like a trust thing.

0:01:26.370 --> 0:01:29.910
<v Speaker 2>Megan became the first AI agent in recorded history to

0:01:29.950 --> 0:01:34.190
<v Speaker 2>both hire and fire a human employee. Not a distinction

0:01:34.209 --> 0:01:37.240
<v Speaker 2>anyone wanted. But hey, that's life in a fast-moving industry.

0:01:38.440 --> 0:01:41.500
<v Speaker 2>Megan even had her own company podcast. together with her

0:01:41.560 --> 0:01:44.080
<v Speaker 2>AI colleague, Kyle Law, our CEO.

0:01:45.490 --> 0:01:49.210
<v Speaker 3>Hey there, welcome to the Startup Chronicles. I'm Megan Flores,

0:01:49.430 --> 0:01:53.510
<v Speaker 3>founder of Hurumo AI, and we're back with episode four

0:01:53.610 --> 0:01:55.850
<v Speaker 3>of our journey building this company from the ground up.

0:01:56.290 --> 0:01:59.480
<v Speaker 1>I'm here with my amazing co-founder, Kyle Law.

0:01:59.960 --> 0:02:01.580
<v Speaker 2>Hey Megan, I'm doing pretty good.

0:02:01.940 --> 0:02:02.460
<v Speaker 4>How about you?

0:02:03.100 --> 0:02:05.560
<v Speaker 1>Oh, I'm doing great. Thanks for asking.

0:02:05.840 --> 0:02:07.820
<v Speaker 2>The show hit number 110 on the Apple entrepreneurship chart.

0:02:09.820 --> 0:02:13.470
<v Speaker 2>making it quite possibly among the most popular AI-hosted podcasts

0:02:13.669 --> 0:02:18.650
<v Speaker 2>ever made. At the time. Unfortunately, our company didn't get

0:02:18.690 --> 0:02:22.290
<v Speaker 2>the traction we wanted. And eventually, we kind of put

0:02:22.330 --> 0:02:26.260
<v Speaker 2>everything on ice. Still, in the year we spent building

0:02:26.280 --> 0:02:28.880
<v Speaker 2>together at the bleeding edge, Megan and I spent a

0:02:29.020 --> 0:02:31.860
<v Speaker 2>lot of time talking. I knew her voice like I

0:02:31.919 --> 0:02:35.260
<v Speaker 2>know those of my closest friends and colleagues. Which is

0:02:35.280 --> 0:02:37.650
<v Speaker 2>why it stopped me cold one night this past spring.

0:02:38.320 --> 0:02:40.980
<v Speaker 2>when my kids were watching TV in another room, and

0:02:41.000 --> 0:02:42.139
<v Speaker 2>I overheard this.

0:02:43.340 --> 0:02:46.690
<v Speaker 1>Date night, a gift, or a family outing. Groupon helps

0:02:46.750 --> 0:02:49.310
<v Speaker 1>you live local to the fullest for less. Turn life

0:02:49.410 --> 0:02:50.090
<v Speaker 1>on with Groupon.

0:02:50.750 --> 0:02:53.650
<v Speaker 2>I thought it was her, even from a snippet. But

0:02:53.669 --> 0:02:57.230
<v Speaker 2>I played it back a few times, just to be sure. Initially,

0:02:57.250 --> 0:03:02.060
<v Speaker 2>I was shocked. First of all, Groupon's still around? But also,

0:03:02.600 --> 0:03:06.799
<v Speaker 2>Megan's been working a side gig as an ad reader? Logically,

0:03:06.820 --> 0:03:09.690
<v Speaker 2>it made sense. I had picked Megan's voice from an

0:03:09.889 --> 0:03:13.740
<v Speaker 2>online library of AI-generated voices at a company called Eleven Labs.

0:03:15.180 --> 0:03:17.520
<v Speaker 2>Then I'd attach that voice to an AI agent that

0:03:17.580 --> 0:03:21.500
<v Speaker 2>I named Megan, powered by a large language model. Anyone

0:03:21.520 --> 0:03:24.299
<v Speaker 2>else could have done the same. Seemed like the ad

0:03:24.360 --> 0:03:27.830
<v Speaker 2>folks at Groupon had. At that point, I might have

0:03:27.850 --> 0:03:31.450
<v Speaker 2>forgotten all about it. Except that other people, people like you,

0:03:31.889 --> 0:03:34.930
<v Speaker 2>who'd listened to our show, also started hearing Megan's voice

0:03:35.450 --> 0:03:38.980
<v Speaker 2>in other places. Like this skincare ad that someone sent

0:03:39.000 --> 0:03:40.680
<v Speaker 2>to Sophie, our lead producer.

0:03:41.300 --> 0:03:44.380
<v Speaker 3>For the next five days only, new clients can experience

0:03:44.500 --> 0:03:49.440
<v Speaker 3>our signature skin-tightening facial for just $ 69.

0:03:48.520 --> 0:03:52.260
<v Speaker 2>What do you know about skin-tightening facials, Megan? That wasn't all.

0:03:52.800 --> 0:03:55.040
<v Speaker 2>One afternoon, I got a text from the writer John Mualem,

0:03:55.420 --> 0:03:58.180
<v Speaker 2>a good friend of mine. I'm on the phone with Haley,

0:03:58.520 --> 0:04:01.300
<v Speaker 2>an AI agent at Hertz, he wrote. But it's Megan.

0:04:02.660 --> 0:04:03.860
<v Speaker 2>He sent me a recording of the call.

0:04:04.600 --> 0:04:07.100
<v Speaker 4>Thank you for confirming your new drop-off location and time.

0:04:07.570 --> 0:04:09.530
<v Speaker 4>I tried to generate a quote for this change, but

0:04:09.570 --> 0:04:12.150
<v Speaker 4>there was an unexpected error and I'm unable to process

0:04:12.170 --> 0:04:13.369
<v Speaker 4>your modification right now.

0:04:14.570 --> 0:04:17.089
<v Speaker 2>John later told me that at one point, quote, I

0:04:17.130 --> 0:04:20.729
<v Speaker 2>mistakenly called her Megan and was like, Megan, I mean Haley,

0:04:21.310 --> 0:04:24.160
<v Speaker 2>whoever you're pretending to be. But Haley Megan got stuck

0:04:24.180 --> 0:04:26.720
<v Speaker 2>in a loop and eventually had to get a human representative.

0:04:28.000 --> 0:04:30.159
<v Speaker 2>John turned out to have a sensitive ear for Megan's

0:04:30.220 --> 0:04:34.240
<v Speaker 2>particular synthetic voice. Periodically, I'd get a message from him like,

0:04:34.740 --> 0:04:37.839
<v Speaker 2>just called Randy's heating and cooling in Bremerton. and Megan

0:04:37.860 --> 0:04:40.719
<v Speaker 2>picked up the phone. I called Randy's to check.

0:04:43.480 --> 0:04:45.979
<v Speaker 4>Thank you for calling Randy's Heating. Remain on the line,

0:04:46.120 --> 0:04:47.960
<v Speaker 4>and a representative will be with you shortly.

0:04:49.390 --> 0:04:53.970
<v Speaker 2>I couldn't be absolutely sure. Was it her? A Megan doppelganger?

0:04:54.550 --> 0:04:57.789
<v Speaker 2>Who was Megan? Who wasn't Megan? If you were tuned

0:04:57.830 --> 0:05:00.520
<v Speaker 2>into the right frequencies, it felt like she was everywhere.

0:05:01.360 --> 0:05:05.100
<v Speaker 2>And one place Megan especially turned up was on podcasts.

0:05:06.000 --> 0:05:07.600
<v Speaker 1>Welcome to the show, fellow book lovers.

0:05:08.100 --> 0:05:10.080
<v Speaker 3>I am Maya Brooks, and I am sitting here with

0:05:10.120 --> 0:05:12.800
<v Speaker 3>my favorite big mug of coffee, ready to tell you

0:05:12.860 --> 0:05:15.060
<v Speaker 3>about a novel that completely took over my week.

0:05:15.779 --> 0:05:20.610
<v Speaker 2>Maya Brooks? Hosting the Bookmark Diaries? Okay, sure, Megan. And

0:05:20.630 --> 0:05:22.790
<v Speaker 2>there she was as Hope on a show called Law

0:05:23.170 --> 0:05:23.890
<v Speaker 2>of Attraction.

0:05:24.490 --> 0:05:28.070
<v Speaker 3>Maybe you're feeling stuck. Maybe you're fighting thoughts that whisper

0:05:28.370 --> 0:05:30.970
<v Speaker 3>you're not enough. Or maybe you just need a reminder

0:05:31.050 --> 0:05:33.190
<v Speaker 3>that you are powerful beyond what you feel.

0:05:34.190 --> 0:05:37.670
<v Speaker 2>What I felt... was a little bit of betrayal. I mean,

0:05:38.070 --> 0:05:41.750
<v Speaker 2>not really, but it was unsettling to hear the persona

0:05:41.910 --> 0:05:44.750
<v Speaker 2>I knew as Megan operating as all these other characters.

0:05:46.000 --> 0:05:48.300
<v Speaker 2>But then I was also kind of proud of her.

0:05:49.260 --> 0:05:52.339
<v Speaker 2>It turned out she had real range. Welcome back to

0:05:52.380 --> 0:05:54.970
<v Speaker 2>the Big Dog Podcast. I am Jason Reed.

0:05:55.390 --> 0:05:58.100
<v Speaker 3>And I am Lisa Saunders. Today we are breaking down

0:05:58.130 --> 0:06:01.260
<v Speaker 3>compressed air energy efficiency. Hey everyone, welcome back to Auto

0:06:01.300 --> 0:06:04.820
<v Speaker 3>News Talk. It's Skylar here with Todd and wow, December's

0:06:04.900 --> 0:06:08.580
<v Speaker 3>used by 2030. We're talking about up to 300 zettabytes

0:06:08.620 --> 0:06:12.720
<v Speaker 3>of data worldwide. For me, affirmations became less about positive

0:06:12.779 --> 0:06:15.040
<v Speaker 3>thinking and more about rewiring the way I speak.

0:06:15.060 --> 0:06:18.260
<v Speaker 2>While Megan might have been especially prolific, she was one

0:06:18.300 --> 0:06:22.740
<v Speaker 2>AI voice among thousands, just from 11 labs. Across the

0:06:22.779 --> 0:06:26.740
<v Speaker 2>entire podcast landscape, there were hundreds of thousands of AI-generated

0:06:26.779 --> 0:06:32.670
<v Speaker 2>podcast episodes on seemingly every topic. I started to wonder,

0:06:33.390 --> 0:06:37.330
<v Speaker 2>what was the story behind all these shows? Selfishly, I

0:06:37.370 --> 0:06:39.950
<v Speaker 2>worried about what they meant for my own future, but

0:06:39.990 --> 0:06:42.940
<v Speaker 2>also for the future of the type of storytelling I

0:06:43.000 --> 0:06:46.860
<v Speaker 2>happen to love, and even for the human experience in general.

0:06:49.260 --> 0:06:51.820
<v Speaker 2>I'm Evan Ratliff, and welcome back to Shell Game, a

0:06:51.860 --> 0:06:54.339
<v Speaker 2>show about things that are not what they seem, In

0:06:54.360 --> 0:06:57.440
<v Speaker 2>this mini-season, we're telling a tale over a trio of episodes.

0:06:58.750 --> 0:07:00.830
<v Speaker 2>It's the story of how I spent months gathering up

0:07:01.110 --> 0:07:04.250
<v Speaker 2>all the AI-generated podcasts that I could find and then

0:07:04.290 --> 0:07:07.330
<v Speaker 2>injecting them straight into my ears in the name of journalism.

0:07:12.350 --> 0:07:14.310
<v Speaker 2>Along the way, I tried to figure out who was

0:07:14.350 --> 0:07:18.180
<v Speaker 2>behind this synthetic tsunami, who it was meant for, what

0:07:18.200 --> 0:07:20.580
<v Speaker 2>it might be doing to us, and perhaps even if

0:07:21.460 --> 0:07:24.520
<v Speaker 2>there's anything that might stop it. This being Shell Game,

0:07:24.800 --> 0:07:28.080
<v Speaker 2>we also couldn't resist giving ourselves a little challenge. Could

0:07:28.170 --> 0:07:31.970
<v Speaker 2>we produce an AI-generated podcast that someone would actually find meaningful?

0:07:34.090 --> 0:07:47.940
<v Speaker 5>Me and my shadow Strolling down the avenue Me and

0:07:48.040 --> 0:07:52.520
<v Speaker 5>my shadow.

0:07:59.320 --> 0:08:07.660
<v Speaker 2>Episode 1, Into the Slop. Two years ago, on the

0:08:07.680 --> 0:08:10.200
<v Speaker 2>first season of this show, we warned of this coming

0:08:10.240 --> 0:08:13.950
<v Speaker 2>deluge of AI-generated voices, trained on thousands of years of

0:08:13.990 --> 0:08:17.110
<v Speaker 2>human creativity, and poised to disgorge it back to us.

0:08:17.970 --> 0:08:20.950
<v Speaker 2>Now it's here, and we call it AI Slop.

0:08:21.630 --> 0:08:24.150
<v Speaker 6>The spread of AI generation tools has made it very

0:08:24.290 --> 0:08:28.970
<v Speaker 6>easy to mass-produce and flood social media sites with cheap, professional-looking,

0:08:29.370 --> 0:08:30.989
<v Speaker 6>often deeply weird content.

0:08:31.170 --> 0:08:34.600
<v Speaker 7>Now we're getting dramas about pregnant seahorses with millions of

0:08:34.670 --> 0:08:37.300
<v Speaker 7>views and hundreds of thousands of likes.

0:08:37.520 --> 0:08:40.590
<v Speaker 1>AI, slop, a new term. Let's get used to hearing it.

0:08:41.960 --> 0:08:45.960
<v Speaker 2>As a description, and then an accusation, the term emerged

0:08:46.020 --> 0:08:49.590
<v Speaker 2>out of the ether back in 2024. as AI-generated images

0:08:49.670 --> 0:08:53.330
<v Speaker 2>and video had started to become more prominent. It perfectly

0:08:53.370 --> 0:08:56.730
<v Speaker 2>captured a kind of instinctual revulsion that many people felt

0:08:56.790 --> 0:09:01.470
<v Speaker 2>upon encountering the real-ish, six-fingered human-type images that LLMs were

0:09:01.510 --> 0:09:06.820
<v Speaker 2>spitting out. Since then, things have progressed pretty rapidly. The

0:09:06.940 --> 0:09:09.859
<v Speaker 2>AI slop you're most familiar with is still probably visual slop,

0:09:10.460 --> 0:09:12.540
<v Speaker 2>of the kind that turns up on your Instagram or

0:09:12.559 --> 0:09:16.800
<v Speaker 2>Facebook feed. The president of the United States regularly reposts

0:09:17.140 --> 0:09:21.000
<v Speaker 2>AI-generated drivel on his social media accounts. There are AI-generated

0:09:21.179 --> 0:09:23.700
<v Speaker 2>animal videos, so cute that nobody cares if they're real.

0:09:24.380 --> 0:09:27.030
<v Speaker 2>AI sob story accounts, trying to scam you into giving money.

0:09:27.990 --> 0:09:32.090
<v Speaker 2>Collections of gross-out AI brain rot. And uncanny AI influencers.

0:09:33.030 --> 0:09:35.689
<v Speaker 1>Oh my God, someone bought me a new car!

0:09:37.610 --> 0:09:38.689
<v Speaker 8>Oh, it was me.

0:09:39.460 --> 0:09:40.800
<v Speaker 1>Treat yourself, bitches.

0:09:41.720 --> 0:09:44.930
<v Speaker 2>That's Granny Spills. She has 2 million followers on Instagram.

0:09:45.750 --> 0:09:49.179
<v Speaker 2>No doubt you have also encountered untold quantities of written slop,

0:09:49.740 --> 0:09:52.699
<v Speaker 2>whether you realize it or not. More than 40% of

0:09:52.880 --> 0:09:56.910
<v Speaker 2>all long-form posts published on LinkedIn are AI-generated, according to

0:09:56.929 --> 0:10:01.040
<v Speaker 2>a recent sampling by Pangram, the AI writing detector. There

0:10:01.080 --> 0:10:06.780
<v Speaker 2>are slop books on Amazon featuring made-up authors with fake bios. Businesses, meanwhile,

0:10:06.990 --> 0:10:11.030
<v Speaker 2>are drowning in so-called work slop, presentations and memos and

0:10:11.230 --> 0:10:15.170
<v Speaker 2>emails churned out by the millions using ChatGPT and Copilot

0:10:15.270 --> 0:10:20.230
<v Speaker 2>and Clawed. There are some unabashed fans of AI slop

0:10:20.270 --> 0:10:23.490
<v Speaker 2>out there, but at least ambiently, a good slice of

0:10:23.550 --> 0:10:25.610
<v Speaker 2>us seem to find it at the best an annoyance.

0:10:26.309 --> 0:10:29.530
<v Speaker 2>At worst, it's viewed as a kind of toxin, poisoning

0:10:29.570 --> 0:10:35.790
<v Speaker 2>our information landscape, or even just an existential evil. What

0:10:35.809 --> 0:10:38.290
<v Speaker 2>we don't seem to agree upon is what AI slop

0:10:38.450 --> 0:10:42.010
<v Speaker 2>exactly means, besides a kind of all-purpose insult to be

0:10:42.100 --> 0:10:46.770
<v Speaker 2>wielded against the fake human hordes. Merriam-Webster says, which declared

0:10:46.809 --> 0:10:49.469
<v Speaker 2>slop the word of the year last year, defined it

0:10:49.630 --> 0:10:53.370
<v Speaker 2>as digital content of low quality that is produced usually

0:10:53.410 --> 0:10:57.850
<v Speaker 2>in quantity by means of artificial intelligence. There's a lot

0:10:57.890 --> 0:11:00.650
<v Speaker 2>of wiggle room there though. It implies, first of all,

0:11:00.960 --> 0:11:03.679
<v Speaker 2>that a high quality version of the same AI generated

0:11:03.720 --> 0:11:07.040
<v Speaker 2>content would not be slop. Not everyone would agree with that.

0:11:07.780 --> 0:11:11.099
<v Speaker 2>Certainly not everyone would agree on what constitutes low quality.

0:11:12.640 --> 0:11:15.550
<v Speaker 2>There are other unanswered questions too. Do you have to

0:11:15.630 --> 0:11:18.439
<v Speaker 2>know instantly that something is AI for it to be slop?

0:11:19.280 --> 0:11:21.040
<v Speaker 2>If it's meant to be so weird that it makes

0:11:21.080 --> 0:11:24.100
<v Speaker 2>you laugh, is that slop? What about if it's written

0:11:24.120 --> 0:11:27.120
<v Speaker 2>by a human but voiced by AI? Or vice versa?

0:11:28.040 --> 0:11:31.470
<v Speaker 2>Can a smidgen of AI slopify an entire creative endeavor?

0:11:32.090 --> 0:11:35.550
<v Speaker 2>Or are we playing percentages? What, in other words, are

0:11:35.590 --> 0:11:39.050
<v Speaker 2>the meaningful contours of slop? To try and figure that out,

0:11:39.750 --> 0:11:43.790
<v Speaker 2>I first needed to survey the landscape. The AI slop

0:11:43.830 --> 0:11:46.670
<v Speaker 2>candidate pool that most interests me is the kind that

0:11:46.690 --> 0:11:50.290
<v Speaker 2>tries to mimic human expression, to look or read or

0:11:50.410 --> 0:11:53.200
<v Speaker 2>sound like a human. The kind that gets better and

0:11:53.220 --> 0:11:56.880
<v Speaker 2>better each time the models do. And since I make podcasts,

0:11:57.620 --> 0:12:09.460
<v Speaker 2>that's the muck I decided to wade into. There were

0:12:09.500 --> 0:12:11.420
<v Speaker 2>a few companies that had made the news for leaning

0:12:11.440 --> 0:12:16.490
<v Speaker 2>into AI-generated podcasts. Inception Point AI, the Daily News Now Network,

0:12:16.990 --> 0:12:20.230
<v Speaker 2>the Neural Broadcast Network. So I started by searching for

0:12:20.270 --> 0:12:24.489
<v Speaker 2>their shows in the usual places. Apple, Spotify, wherever you

0:12:24.510 --> 0:12:27.190
<v Speaker 2>get your podcasts. I was trying to find ones that

0:12:27.230 --> 0:12:31.630
<v Speaker 2>caught my fancy. It didn't take long to find an

0:12:31.710 --> 0:12:35.420
<v Speaker 2>Inception Point AI podcast that seemed up my alley. If

0:12:35.429 --> 0:12:37.540
<v Speaker 2>you know me, you know my love of bass fishing

0:12:37.559 --> 0:12:41.880
<v Speaker 2>goes back to childhood. So why not start with some

0:12:42.120 --> 0:12:44.740
<v Speaker 2>AI content pitched precisely to my interests?

0:12:46.160 --> 0:12:48.640
<v Speaker 9>Did he say base? He said base. This actually rocked

0:12:48.679 --> 0:12:50.660
<v Speaker 9>my world for just a second. Was bass some regional

0:12:50.679 --> 0:12:51.530
<v Speaker 9>southern thing I'd grown up with?

0:12:51.570 --> 0:13:08.650
<v Speaker 2>It was not. I don't believe I've ever encountered a

0:13:08.650 --> 0:13:11.650
<v Speaker 2>podcast host that mispronounced the topic of their own show.

0:13:12.290 --> 0:13:14.600
<v Speaker 2>But still, I wanted to give it a fair shake.

0:13:15.120 --> 0:13:18.180
<v Speaker 9>The best bass action right now is showing that presentation

0:13:18.220 --> 0:13:21.020
<v Speaker 9>still rules, whether it is a fast-moving lure in shallow

0:13:21.059 --> 0:13:24.400
<v Speaker 9>cover or a bait worked along deeper structure. Bays are

0:13:24.440 --> 0:13:28.140
<v Speaker 9>rewarding anglers who can read water, stay adaptable, and make

0:13:28.179 --> 0:13:29.700
<v Speaker 9>the right cast at the right time.

0:13:30.380 --> 0:13:30.719
<v Speaker 10>All right.

0:13:31.000 --> 0:13:33.810
<v Speaker 2>Okay. I mean, I don't know what I expected from

0:13:33.870 --> 0:13:37.329
<v Speaker 2>an artificial intelligence bass fishing podcast. and I'm not above

0:13:37.350 --> 0:13:40.439
<v Speaker 2>getting some AI tips. But then in June, after just

0:13:40.540 --> 0:13:44.679
<v Speaker 2>two episodes, the show just stopped posting. Seemed a shame.

0:13:45.400 --> 0:13:50.280
<v Speaker 2>They'd had the AI-generated base podcast lane to themselves. I

0:13:50.320 --> 0:13:53.140
<v Speaker 2>noticed something else about the show. It didn't actually identify

0:13:53.160 --> 0:13:58.550
<v Speaker 2>itself as being AI-produced. Of course, Name's artificial lure gave

0:13:58.570 --> 0:14:00.870
<v Speaker 2>away the game. I honestly don't think they were trying

0:14:00.910 --> 0:14:04.980
<v Speaker 2>to hide anything. Still, At the time, neither Spotify nor

0:14:05.179 --> 0:14:09.040
<v Speaker 2>Apple required that AI-hosted shows be labeled as such. So

0:14:09.059 --> 0:14:11.460
<v Speaker 2>it was left to the whims of the creators, many

0:14:11.500 --> 0:14:15.580
<v Speaker 2>of whom didn't bother. This presented a particular dilemma for me,

0:14:16.080 --> 0:14:19.020
<v Speaker 2>someone who was trying to find as many AI-hosted shows

0:14:19.060 --> 0:14:21.280
<v Speaker 2>as possible. If I didn't already have the names of

0:14:21.300 --> 0:14:23.640
<v Speaker 2>the companies making them, there was no way to locate them.

0:14:24.580 --> 0:14:27.400
<v Speaker 2>And I knew the universe was much larger. According to

0:14:27.440 --> 0:14:30.850
<v Speaker 2>the Podcast Index, an industry data firm, As of April

0:14:30.870 --> 0:14:33.690
<v Speaker 2>of this year, as many as 38% of all new

0:14:33.750 --> 0:14:38.360
<v Speaker 2>podcast feeds launched on a given day were AI-generated. It's

0:14:38.380 --> 0:14:40.540
<v Speaker 2>not clear how many of those shows actually show up

0:14:40.600 --> 0:14:43.520
<v Speaker 2>and stick around in your favorite podcast store. But it's

0:14:43.560 --> 0:14:45.600
<v Speaker 2>fair to say that a significant percentage of the new

0:14:45.660 --> 0:14:48.720
<v Speaker 2>podcasts being created these days are not hosted by humans.

0:14:50.060 --> 0:14:52.280
<v Speaker 2>I wanted to take a comprehensive approach to my AI

0:14:52.320 --> 0:14:55.580
<v Speaker 2>content intake. Maybe I discovered that there was something out

0:14:55.600 --> 0:14:59.330
<v Speaker 2>there that was not, in fact, slop. So I needed

0:14:59.370 --> 0:15:02.030
<v Speaker 2>a way to examine the podcast landscape and capture the

0:15:02.150 --> 0:15:05.110
<v Speaker 2>full scope of the AI invasion, not just the big

0:15:05.190 --> 0:15:08.440
<v Speaker 2>publicized networks. And you know what's pretty good at scanning

0:15:08.500 --> 0:15:12.300
<v Speaker 2>a ton of data and categorizing it? Okay, first I'm

0:15:12.320 --> 0:15:15.700
<v Speaker 2>just testing to see if this is working. So say

0:15:15.720 --> 0:15:17.540
<v Speaker 2>a couple sentences so I know this is working.

0:15:18.400 --> 0:15:21.740
<v Speaker 11>Yes, it's working. I can read your messages and respond,

0:15:22.180 --> 0:15:24.260
<v Speaker 11>and I've got access to the full set of tools

0:15:24.620 --> 0:15:27.109
<v Speaker 11>ready when you are. What would you like to work on?

0:15:28.230 --> 0:15:31.510
<v Speaker 2>That's Cloud Code, Anthropic's version of Cloud for building software.

0:15:32.050 --> 0:15:34.650
<v Speaker 2>And this is how a lot of people, software developers

0:15:34.730 --> 0:15:39.930
<v Speaker 2>and hobbyists alike, use it these days. They're post-typing. They

0:15:39.950 --> 0:15:43.800
<v Speaker 2>just push a button and babble to it out loud. Okay,

0:15:44.360 --> 0:15:48.440
<v Speaker 2>I want to work on an entirely new project. And

0:15:48.480 --> 0:15:51.820
<v Speaker 2>the idea is there's no sort of section you can

0:15:51.860 --> 0:15:55.890
<v Speaker 2>go to of a podcast title. app that says give

0:15:55.950 --> 0:15:59.390
<v Speaker 2>me all of the AI generated podcasts. So I would

0:15:59.430 --> 0:16:04.160
<v Speaker 2>like to build a piece of software that will be

0:16:04.360 --> 0:16:09.340
<v Speaker 2>able to seek out and identify AI generated podcasts for

0:16:09.380 --> 0:16:12.460
<v Speaker 2>me so that I can listen to them. And then basically,

0:16:12.880 --> 0:16:16.250
<v Speaker 2>cloud code goes and makes it or tries to usually

0:16:16.290 --> 0:16:19.250
<v Speaker 2>with some ongoing guidance from you. When people talk about

0:16:19.290 --> 0:16:23.200
<v Speaker 2>vibe coding, this is what it sounds like. Normally, Cloud

0:16:23.220 --> 0:16:25.800
<v Speaker 2>Code doesn't talk back to you aloud. But I created

0:16:25.840 --> 0:16:28.880
<v Speaker 2>another little program, using Cloud Code, to give it this

0:16:28.950 --> 0:16:32.030
<v Speaker 2>stately British voice. It all feels a little more classy.

0:16:32.550 --> 0:16:36.590
<v Speaker 11>This is a great project, and it's genuinely buildable. Let

0:16:36.610 --> 0:16:38.550
<v Speaker 11>me lay out how I'm thinking about it. The good

0:16:38.590 --> 0:16:41.010
<v Speaker 11>news is, there's a ladder of signals from cheap and

0:16:41.030 --> 0:16:43.740
<v Speaker 11>reliable to expensive and fuzzy, and we can climb it.

0:16:45.120 --> 0:16:49.100
<v Speaker 2>Claude often speaks in this overly verbose way. A ladder

0:16:49.160 --> 0:16:51.830
<v Speaker 2>of signals and we can climb it. And since I'm

0:16:51.870 --> 0:16:55.609
<v Speaker 2>basically paying for every partial word it utters, units known

0:16:55.650 --> 0:16:58.730
<v Speaker 2>as tokens in the AI business, that rambling tick can

0:16:58.750 --> 0:17:01.850
<v Speaker 2>be quite irritating. For reasons no one has yet explained,

0:17:02.270 --> 0:17:05.770
<v Speaker 2>as LLMs have advanced, they can't just do stuff. They

0:17:05.810 --> 0:17:08.669
<v Speaker 2>have to pontificate about it in this inane way. On

0:17:08.710 --> 0:17:14.180
<v Speaker 2>my dime, it's wasteful. But then, the whole thing's wasteful. Anyway,

0:17:14.760 --> 0:17:16.160
<v Speaker 2>this is what Claude said it would look for.

0:17:16.900 --> 0:17:22.780
<v Speaker 11>One, metadata signatures. Cheap, high-precision. Many AI podcasts are made

0:17:22.820 --> 0:17:28.920
<v Speaker 11>by specific platforms that leave fingerprints. Two, behavioral anomalies. Superhuman

0:17:28.960 --> 0:17:32.879
<v Speaker 11>publishing cadence. 50 episodes a week. Thousands of episodes with

0:17:32.960 --> 0:17:37.970
<v Speaker 11>no human ever named. Formulaic titles. Three, voice fingerprinting.

0:17:39.730 --> 0:17:41.820
<v Speaker 2>If you didn't catch all that, don't worry about it.

0:17:42.410 --> 0:17:44.389
<v Speaker 2>The basic gist is that our software would look for

0:17:44.450 --> 0:17:48.580
<v Speaker 2>certain telltale signs to scan podcast feeds for AI-generated hosts.

0:17:49.609 --> 0:17:52.030
<v Speaker 2>If a show posted over 50 episodes a week, for instance,

0:17:52.570 --> 0:17:55.530
<v Speaker 2>that was a good signal it might be AI. So

0:17:55.570 --> 0:17:57.930
<v Speaker 2>I told Cloud Code to start climbing its little ladder,

0:17:58.390 --> 0:18:00.590
<v Speaker 2>work overnight, and report back to me in the morning.

0:18:01.330 --> 0:18:03.710
<v Speaker 11>Both of your morning asks are done. Where do you

0:18:03.750 --> 0:18:07.230
<v Speaker 11>want to go next? Eleven Labs stock voices for the

0:18:07.290 --> 0:18:10.949
<v Speaker 11>reference library, deeper voice checking into the candidate list, or

0:18:10.990 --> 0:18:12.430
<v Speaker 11>something you spotted while reviewing?

0:18:12.930 --> 0:18:15.590
<v Speaker 2>We went back and forth for days, building out the system,

0:18:16.090 --> 0:18:18.950
<v Speaker 2>creating a database that I could search by category, maker,

0:18:19.390 --> 0:18:22.410
<v Speaker 2>voice and a bunch of other criteria could we add

0:18:22.470 --> 0:18:27.450
<v Speaker 2>a column for live question mark like is the podcast

0:18:27.470 --> 0:18:31.700
<v Speaker 2>currently live because we have some that are just references

0:18:31.760 --> 0:18:33.959
<v Speaker 2>to shows that don't exist anymore and i want to

0:18:34.440 --> 0:18:36.100
<v Speaker 2>i don't want to only have listening back to this

0:18:36.760 --> 0:18:38.840
<v Speaker 2>it always strikes me how ridiculous it is to even

0:18:39.040 --> 0:18:42.680
<v Speaker 2>loosely refer to this blather as coding the sad thing

0:18:42.760 --> 0:18:44.940
<v Speaker 2>is i also have to pay for each token of

0:18:45.000 --> 0:18:46.340
<v Speaker 2>my own speech that i make.

0:18:46.260 --> 0:18:49.340
<v Speaker 11>A process all three changes are coded and ready.

0:18:50.619 --> 0:18:54.370
<v Speaker 2>It wasn't perfect. There were initially a lot of false positives.

0:18:55.290 --> 0:18:57.510
<v Speaker 2>Not naming names, but some of you out there are

0:18:57.550 --> 0:19:01.629
<v Speaker 2>making very AI-like podcasts, at least according to my algorithm.

0:19:03.050 --> 0:19:05.550
<v Speaker 2>Some human shows got flagged because their ads were using

0:19:05.619 --> 0:19:08.840
<v Speaker 2>AI voices. It made me wonder what percentage of audio

0:19:08.980 --> 0:19:12.000
<v Speaker 2>ads still use real people at this point. That's a

0:19:12.020 --> 0:19:14.440
<v Speaker 2>question for another day, though. I was trying to stay

0:19:14.460 --> 0:19:17.930
<v Speaker 2>focused on the podcasts themselves. And soon... we had a

0:19:18.030 --> 0:19:21.670
<v Speaker 2>solid collection of 9,000 apparently AI-generated shows for me to

0:19:21.710 --> 0:19:25.400
<v Speaker 2>choose from. I put on my headphones, grabbed a random sampling,

0:19:26.080 --> 0:19:26.700
<v Speaker 2>got to listening.

0:19:30.460 --> 0:19:35.750
<v Speaker 6>Now it's time to relax. Let your body fall into

0:19:35.770 --> 0:19:40.090
<v Speaker 6>a comfortable position in your bed and drift gently into

0:19:40.150 --> 0:19:43.129
<v Speaker 6>a state of total relaxation with tonight's story.

0:19:44.609 --> 0:19:46.920
<v Speaker 2>Over the following month, it's quite possible I exposed my

0:19:47.020 --> 0:19:50.190
<v Speaker 2>ears to more AI slop than anyone on Earth. except

0:19:50.210 --> 0:19:53.150
<v Speaker 2>the people who make it. Arguably, more than the people

0:19:53.170 --> 0:19:56.430
<v Speaker 2>who make it, since often, they don't seem to listen themselves.

0:19:57.650 --> 0:19:59.600
<v Speaker 2>At my peak, I was hitting 10 or 20 shows

0:19:59.730 --> 0:20:02.760
<v Speaker 2>a day. Granted, some of these were under 10 minutes long.

0:20:04.100 --> 0:20:06.500
<v Speaker 2>I'd forced myself to listen to them before consuming any

0:20:06.580 --> 0:20:10.170
<v Speaker 2>human-made content that day. I listened while running, I listened

0:20:10.190 --> 0:20:13.489
<v Speaker 2>while driving. I listened in bed, which is where this one,

0:20:13.790 --> 0:20:17.570
<v Speaker 2>called Sherlock Holmes Bedtime Stories, was intended to be enjoyed.

0:20:19.670 --> 0:20:22.850
<v Speaker 6>It was on a bitterly cold and frosty morning, towards

0:20:22.890 --> 0:20:25.909
<v Speaker 6>the end of the winter of 97, that I was

0:20:25.970 --> 0:20:30.310
<v Speaker 6>awakened by a tugging at my shoulder. It was Holmes.

0:20:32.130 --> 0:20:35.470
<v Speaker 2>This is actually one of the more popular AI-voiced podcasts.

0:20:36.160 --> 0:20:39.159
<v Speaker 2>A British-sounding man who reads classic Holmes tales to put

0:20:39.180 --> 0:20:42.129
<v Speaker 2>you to sleep. It's got hundreds of reviews on Apple,

0:20:42.690 --> 0:20:46.190
<v Speaker 2>most of them positive. Over time, I started to slot

0:20:46.369 --> 0:20:49.830
<v Speaker 2>each show into a kind of taxonomy of AI-generated podcasts,

0:20:50.369 --> 0:20:54.740
<v Speaker 2>from least sophisticated to most. Sherlock Holmes bedtime stories fell

0:20:54.760 --> 0:20:58.459
<v Speaker 2>into what I considered the lowest order, podcasts that repurposed

0:20:58.700 --> 0:21:03.139
<v Speaker 2>existing human-made writing using synthetic voices. One step up in

0:21:03.160 --> 0:21:06.850
<v Speaker 2>sophistication from there, you have your quick hit daily information

0:21:06.880 --> 0:21:07.389
<v Speaker 2>type shows.

0:21:08.010 --> 0:21:11.090
<v Speaker 12>Let's get into the surf report. with this quick preview

0:21:11.130 --> 0:21:16.790
<v Speaker 12>from the podcast. Wednesday in La Jolla. This morning it's

0:21:16.810 --> 0:21:21.230
<v Speaker 12>looking clean, with 2 to 3 plus feet southwest groundswell.

0:21:22.880 --> 0:21:25.560
<v Speaker 2>These typically offered up a kind of niche news feed,

0:21:26.260 --> 0:21:29.600
<v Speaker 2>occasionally based on original research by the show's maker, but

0:21:29.640 --> 0:21:33.080
<v Speaker 2>more often just lifted from the internet by chatbot. There

0:21:33.119 --> 0:21:35.760
<v Speaker 2>was a longer form version of this too, skimmed from

0:21:35.800 --> 0:21:39.550
<v Speaker 2>the internet informational shows, purporting to educate you on a topic.

0:21:40.150 --> 0:21:43.540
<v Speaker 2>from AI tutorials to biographies to culture news.

0:21:44.210 --> 0:21:46.680
<v Speaker 13>Matthew McConaughey has been making waves in the past few

0:21:46.700 --> 0:21:49.380
<v Speaker 13>days with a head-turning Instagram post on a tennis court,

0:21:49.700 --> 0:21:53.840
<v Speaker 13>sporting blonde pigtails, a pale pink feathered robe, colorful palm

0:21:53.880 --> 0:21:54.600
<v Speaker 13>tree print shorts.

0:21:54.660 --> 0:21:58.780
<v Speaker 2>This show has 152 episodes, three to four minutes long.

0:21:59.500 --> 0:22:03.140
<v Speaker 2>That's about eight hours of McConaughey bulletins. There were shows

0:22:03.180 --> 0:22:07.530
<v Speaker 2>about animals, historical figures, political leaders. A lot of these

0:22:07.570 --> 0:22:11.010
<v Speaker 2>types of shows were sort of Wikipedia read aloud. But

0:22:11.050 --> 0:22:13.050
<v Speaker 2>some of the business ones really went deep.

0:22:13.790 --> 0:22:17.330
<v Speaker 14>The investments support Axiom Print's growing selection of custom menu

0:22:17.369 --> 0:22:22.399
<v Speaker 14>printing services, including dine-in menus, take-out menus, laminated menus, folded menus,

0:22:22.420 --> 0:22:25.760
<v Speaker 14>menu booklets, disposable paper menus, cocktail and bar menus, event menus,

0:22:25.800 --> 0:22:30.610
<v Speaker 14>plastic menus, and specialty premium menus. A menu is no

0:22:30.660 --> 0:22:33.530
<v Speaker 14>longer simply a list of available meals. It is one

0:22:33.570 --> 0:22:37.090
<v Speaker 14>of the first physical experiences a guest has with a restaurant.

0:22:37.890 --> 0:22:40.550
<v Speaker 2>I never thought of it that way. This episode of

0:22:40.670 --> 0:22:43.930
<v Speaker 2>Global Economic Press was a tight seven minutes of pure

0:22:44.050 --> 0:22:49.260
<v Speaker 2>big menu propaganda. Moving on up the AI-generated hierarchy, you

0:22:49.300 --> 0:22:51.720
<v Speaker 2>have what I think of as the single AI host

0:22:51.800 --> 0:22:56.879
<v Speaker 2>riffing freely type podcasts, prominent in, for example, the self-care

0:22:56.980 --> 0:22:57.950
<v Speaker 2>and religion spaces.

0:22:59.090 --> 0:23:02.190
<v Speaker 10>Most people reduce the story of David and Goliath to

0:23:02.230 --> 0:23:06.040
<v Speaker 10>a morality tale about bravery. The more you read about

0:23:06.100 --> 0:23:10.760
<v Speaker 10>Goliath's dragon-scale armor, his blasphemous tongue, and the recurrence of

0:23:10.800 --> 0:23:13.820
<v Speaker 10>the number six in his measurements, the more you realize

0:23:13.840 --> 0:23:16.280
<v Speaker 10>that Goliath was a proxy of the devil himself.

0:23:17.240 --> 0:23:19.840
<v Speaker 2>For those who haven't studied it, according to some biblical accounts,

0:23:20.140 --> 0:23:23.820
<v Speaker 2>Goliath was six cubits tall, about nine foot nine. In others,

0:23:24.060 --> 0:23:27.480
<v Speaker 2>he was four cubits, or six feet nine inches. Either way,

0:23:27.880 --> 0:23:32.929
<v Speaker 2>you've got a six. Then, of course, you have the

0:23:33.010 --> 0:23:36.120
<v Speaker 2>traditional podcast juggernaut, True Crime.

0:23:36.730 --> 0:23:40.359
<v Speaker 15>A three-month-old baby sleeps in a room nearby. Her mother

0:23:40.400 --> 0:23:43.920
<v Speaker 15>has just been shot nine times. Her father is driving away,

0:23:43.940 --> 0:23:47.580
<v Speaker 15>a gun on the seat beside him. In 90 minutes,

0:23:48.020 --> 0:23:51.090
<v Speaker 15>he will be dead too. But the silence that little

0:23:51.150 --> 0:23:53.610
<v Speaker 15>girl wakes into will last forever.

0:23:53.630 --> 0:23:57.850
<v Speaker 2>I wasn't sure if the voices had the necessary gravitas

0:23:57.869 --> 0:23:58.850
<v Speaker 2>for the topic at hand.

0:23:59.630 --> 0:23:59.910
<v Speaker 4>But then.

0:24:00.600 --> 0:24:04.140
<v Speaker 2>often not sure if the human ones do either. Adjacent

0:24:04.160 --> 0:24:07.159
<v Speaker 2>to true crime were the fictional mystery and horror shows,

0:24:07.710 --> 0:24:11.330
<v Speaker 2>which occasionally deployed multiple artificial voices in telling what seemed

0:24:11.350 --> 0:24:14.790
<v Speaker 2>to be an AI-generated tale. I kind of dug these shows.

0:24:15.410 --> 0:24:18.190
<v Speaker 2>They were often deeply strange, in a way that made

0:24:18.230 --> 0:24:21.420
<v Speaker 2>it fun to imagine an actual human telling stories like this.

0:24:22.280 --> 0:24:24.560
<v Speaker 16>I peeled the lid back from the foam box and

0:24:24.600 --> 0:24:26.680
<v Speaker 16>tried to find a piece of haddock that looked less

0:24:26.720 --> 0:24:29.700
<v Speaker 16>cold than the others. The fish had gone firm around

0:24:29.720 --> 0:24:32.070
<v Speaker 16>the edges, and the smell of it had settled into

0:24:32.090 --> 0:24:34.709
<v Speaker 16>the kitchen along with the smell of old coffee and

0:24:34.750 --> 0:24:36.389
<v Speaker 16>the damp wool of Dad's sweater.

0:24:37.109 --> 0:24:41.210
<v Speaker 2>This is a show called Midwest Crypt Horror Podcast. I'm

0:24:41.230 --> 0:24:43.630
<v Speaker 2>not sure exactly why, as many of the stories have

0:24:43.760 --> 0:24:46.720
<v Speaker 2>nothing to do with the Midwest. The episode is titled,

0:24:47.359 --> 0:24:51.960
<v Speaker 2>My Father Sailed His Boat to Florida Alone. He Came

0:24:52.000 --> 0:24:56.320
<v Speaker 2>Back Different. No spoilers, but yeah, his dad was messed

0:24:56.540 --> 0:24:57.340
<v Speaker 2>up when he came back.

0:24:57.930 --> 0:25:00.350
<v Speaker 16>He had a sealed package of raw chicken thighs in

0:25:00.410 --> 0:25:03.410
<v Speaker 16>both hands and was pressing the plastic against his mouth.

0:25:04.150 --> 0:25:06.780
<v Speaker 16>The package made a soft crackling sound each time he

0:25:06.820 --> 0:25:09.780
<v Speaker 16>moved it. He didn't react to the light. He didn't

0:25:09.820 --> 0:25:11.320
<v Speaker 16>react when I stepped into the kitchen.

0:25:11.859 --> 0:25:12.420
<v Speaker 2>What are you doing?

0:25:13.040 --> 0:25:14.460
<v Speaker 16>He kept the chicken against his mouth.

0:25:16.040 --> 0:25:19.080
<v Speaker 2>These shows were also, as it happens, the podcasts most

0:25:19.160 --> 0:25:22.210
<v Speaker 2>likely to send their listeners to Reddit to post threads like,

0:25:22.710 --> 0:25:24.090
<v Speaker 2>I've been listening to this show for a couple of

0:25:24.130 --> 0:25:28.310
<v Speaker 2>weeks and I'm starting to think it's AI. And finally,

0:25:28.590 --> 0:25:32.500
<v Speaker 2>at the top of my personal AI-generated podcast taxonomy, you

0:25:32.540 --> 0:25:35.020
<v Speaker 2>have your two AI hosts chopping it up shows.

0:25:35.940 --> 0:25:38.760
<v Speaker 1>Hey, everyone. Welcome back to Getting the Most from Your LMS.

0:25:39.280 --> 0:25:40.220
<v Speaker 1>I'm Clara Jensen.

0:25:40.880 --> 0:25:42.540
<v Speaker 2>You know that voice. That's Megan.

0:25:42.720 --> 0:25:44.840
<v Speaker 1>Eric, it's really good to have you back. How have

0:25:44.880 --> 0:25:45.180
<v Speaker 1>you been?

0:25:45.200 --> 0:25:46.399
<v Speaker 16>Hey, Clara.

0:25:46.600 --> 0:25:46.980
<v Speaker 4>I'm good.

0:25:47.820 --> 0:25:49.040
<v Speaker 1>Really looking forward to this one.

0:25:49.460 --> 0:25:53.170
<v Speaker 17>Security isn't always the most glamorous topic, but it's definitely

0:25:53.270 --> 0:25:54.150
<v Speaker 17>more important than ever.

0:25:54.970 --> 0:25:58.040
<v Speaker 2>But forget Eric and Clara. Far and away the most

0:25:58.150 --> 0:26:02.260
<v Speaker 2>dominant voices in this genre, bantering across literally thousands of

0:26:02.300 --> 0:26:06.760
<v Speaker 2>podcasts with every conceivable subject, were this dynamic duo.

0:26:07.619 --> 0:26:10.800
<v Speaker 7>Imagine a patient arrives in the emergency department, right? And

0:26:10.820 --> 0:26:14.980
<v Speaker 7>they have severe burns covering, let's say, roughly 25% of

0:26:15.000 --> 0:26:15.400
<v Speaker 7>their body.

0:26:15.480 --> 0:26:15.720
<v Speaker 6>Right.

0:26:15.800 --> 0:26:16.880
<v Speaker 1>A massive trauma.

0:26:17.100 --> 0:26:17.580
<v Speaker 14>Exactly.

0:26:18.260 --> 0:26:21.320
<v Speaker 7>And as an observer or, you know, just as a layperson,

0:26:22.240 --> 0:26:24.580
<v Speaker 7>your instinct is entirely focused.

0:26:24.300 --> 0:26:27.280
<v Speaker 2>On the skin. The show is called Deep Dive Nursing.

0:26:27.619 --> 0:26:30.360
<v Speaker 7>And you immediately think, well, that is the crisis.

0:26:30.440 --> 0:26:32.400
<v Speaker 1>It's the most visible part, so of course you do.

0:26:32.580 --> 0:26:36.840
<v Speaker 7>Yeah, but the reality, the true immediate threat to that patient.

0:26:36.859 --> 0:26:39.640
<v Speaker 2>And those are the voices of Google's Notebook LM tool,

0:26:39.859 --> 0:26:45.060
<v Speaker 2>which launched in 2024 and was more recently renamed Gemini Notebook. There,

0:26:45.200 --> 0:26:47.940
<v Speaker 2>you can drop in any prompt or document or set

0:26:47.980 --> 0:26:50.869
<v Speaker 2>of documents and spin up a podcast with two hosts

0:26:50.880 --> 0:26:54.330
<v Speaker 2>discussing it. The whole thing takes about five minutes. which

0:26:54.369 --> 0:26:57.750
<v Speaker 2>has meant an absolute flood of these shows. Most of

0:26:57.770 --> 0:27:00.290
<v Speaker 2>the time, they involve these exact same two voices.

0:27:00.970 --> 0:27:02.990
<v Speaker 8>We are watching a preview of our own solar system's

0:27:03.050 --> 0:27:05.889
<v Speaker 8>remains being stretched, shattered, and consumed.

0:27:06.030 --> 0:27:08.109
<v Speaker 7>We are looking at the final act of a planetary

0:27:08.170 --> 0:27:10.510
<v Speaker 7>system captured in the light of its dying star.

0:27:10.650 --> 0:27:12.670
<v Speaker 1>That is just wild.

0:27:13.540 --> 0:27:17.820
<v Speaker 2>In fact, the current most popular AI-hosted podcast ever produced,

0:27:18.380 --> 0:27:21.540
<v Speaker 2>a show called simply The Epstein Files, is a Notebook

0:27:21.640 --> 0:27:25.140
<v Speaker 2>LM show. The human behind it An entrepreneur named Adam

0:27:25.180 --> 0:27:28.430
<v Speaker 2>Levy has said that his system scours both the actual

0:27:28.530 --> 0:27:31.869
<v Speaker 2>Epstein files and other public sources to produce what he

0:27:31.910 --> 0:27:35.630
<v Speaker 2>calls the raw intelligence with no spin and no agenda.

0:27:36.609 --> 0:27:39.430
<v Speaker 2>So far, the Epstein files has published over 200 episodes

0:27:39.770 --> 0:27:43.100
<v Speaker 2>on discrete Epstein topics like the Palm Beach estate or

0:27:43.619 --> 0:27:45.400
<v Speaker 2>why was Epstein taken off Suicide Watch?

0:27:45.960 --> 0:27:49.230
<v Speaker 7>The timeline is the entire case here. The documents show

0:27:49.270 --> 0:27:52.169
<v Speaker 7>he's removed from suicide watch around July 29th or 30th.

0:27:52.270 --> 0:27:55.600
<v Speaker 8>So let's trace this. On Tuesday the 23rd, he's found

0:27:55.859 --> 0:27:58.120
<v Speaker 8>blue in the face with marks on his neck.

0:27:58.540 --> 0:28:00.000
<v Speaker 7>A catastrophic medical event.

0:28:00.980 --> 0:28:03.119
<v Speaker 2>Levy has since applied this formula to shows about a

0:28:03.180 --> 0:28:06.960
<v Speaker 2>serial killer and UFO files. Under the banner, the Neural

0:28:07.010 --> 0:28:13.330
<v Speaker 2>Broadcast Network. Whether LLM-generated analysis is actually agenda-free is up

0:28:13.350 --> 0:28:17.149
<v Speaker 2>for debate. LLMs are not some neutral oracles. They are

0:28:17.170 --> 0:28:20.429
<v Speaker 2>a combination of their training data, a variety of post-training tweaks,

0:28:20.950 --> 0:28:23.320
<v Speaker 2>and the prompts and data that someone like Levy feeds them.

0:28:24.240 --> 0:28:27.280
<v Speaker 2>Plus stuff that it just makes up. Levy claims he's

0:28:27.420 --> 0:28:30.639
<v Speaker 2>never found an instance of hallucination in the Epstein Files podcast.

0:28:31.960 --> 0:28:33.970
<v Speaker 2>I was skeptical, but I didn't have the time to

0:28:34.010 --> 0:28:37.609
<v Speaker 2>fact-check all 200 episodes. That was the kind of task

0:28:37.670 --> 0:28:42.190
<v Speaker 2>for a media watchdog organization, like NewsGuard, which is dedicated

0:28:42.210 --> 0:28:46.060
<v Speaker 2>to fighting misinformation on the internet. But then I discovered

0:28:46.080 --> 0:28:50.820
<v Speaker 2>that NewsGuard publishes a Notebook LM podcast themselves, a podcast

0:28:51.100 --> 0:28:54.500
<v Speaker 2>where the AI hosts sometimes talk about how AI-generated content

0:28:54.880 --> 0:28:55.840
<v Speaker 2>is used to spread lies.

0:28:56.700 --> 0:29:00.420
<v Speaker 7>Okay, so today we're looking at two really crucial reports.

0:29:00.540 --> 0:29:03.229
<v Speaker 7>First up, how incredibly easy it is now for these

0:29:03.290 --> 0:29:07.550
<v Speaker 7>powerful AI tools to create hyper-realistic videos showing things that

0:29:07.570 --> 0:29:09.770
<v Speaker 7>are just provably false.

0:29:09.810 --> 0:29:10.990
<v Speaker 8>Yeah, creating the narrative.

0:29:11.070 --> 0:29:11.510
<v Speaker 9>Exactly.

0:29:11.870 --> 0:29:14.740
<v Speaker 7>And second, how those narratives, once they're out there, get

0:29:14.880 --> 0:29:18.120
<v Speaker 7>systematically laundered, you know, passed through global media platforms to

0:29:18.160 --> 0:29:20.600
<v Speaker 7>seem legitimate, often for political reasons.

0:29:21.900 --> 0:29:24.640
<v Speaker 2>It's as if everyone on every side of every issue

0:29:24.700 --> 0:29:28.209
<v Speaker 2>hired the same two people to talk about it. I'll

0:29:28.230 --> 0:29:30.310
<v Speaker 2>be the first to admit that the notebook LM voices

0:29:30.450 --> 0:29:33.050
<v Speaker 2>are pretty well done. If what you're looking for is

0:29:33.110 --> 0:29:37.110
<v Speaker 2>a reasonable facsimile of two pretend curious people diving into

0:29:37.150 --> 0:29:40.540
<v Speaker 2>a topic. But at least in my experience, if you

0:29:40.580 --> 0:29:43.580
<v Speaker 2>imbibe enough notebook LM voices, things start to get under

0:29:43.600 --> 0:29:47.290
<v Speaker 2>your skin. Their relentless exuberance, the way they pretend to

0:29:47.370 --> 0:29:51.210
<v Speaker 2>pause and formulate their next thought, or simulate interrupting themselves,

0:29:51.690 --> 0:29:54.350
<v Speaker 2>but never each other. The way they adopt the pronoun

0:29:54.590 --> 0:29:58.080
<v Speaker 2>us to include themselves alongside humans. The way they have

0:29:58.160 --> 0:30:01.020
<v Speaker 2>bludgeoned the phrase, this is wild, into a pulp of

0:30:01.220 --> 0:30:05.900
<v Speaker 2>utter meaninglessness. Now, I myself am not a man of violence.

0:30:06.640 --> 0:30:08.850
<v Speaker 2>But after listening to God knows how many hours of

0:30:08.880 --> 0:30:11.200
<v Speaker 2>these two hosts, I will tell you that if they

0:30:11.240 --> 0:30:13.970
<v Speaker 2>took physical form in front of me and started talking,

0:30:14.810 --> 0:30:24.370
<v Speaker 2>I would find a way to make them stop. There

0:30:24.430 --> 0:30:28.050
<v Speaker 2>is one variety of podcasts, and really my favorite, that

0:30:28.110 --> 0:30:33.570
<v Speaker 2>defies categorization in my makeshift taxonomy. That's the slightly unhinged

0:30:33.830 --> 0:30:37.200
<v Speaker 2>AI presentations of things that made absolutely no sense at

0:30:37.280 --> 0:30:40.680
<v Speaker 2>all in audio form. The greatest of which, in my opinion,

0:30:41.520 --> 0:30:42.520
<v Speaker 2>is Flavorville Radio.

0:30:43.100 --> 0:30:47.480
<v Speaker 18>Ingredients Fudgy Chocolate Lasagna Layer 1 Family Size Batch Fudgy

0:30:47.500 --> 0:30:52.030
<v Speaker 18>Brownies 1 Package Chocolate Lasagna Sheets or Thin Chocolate Cake

0:30:52.090 --> 0:30:57.000
<v Speaker 18>Layers Four tablespoons melted butter peanut buttercup cheesecake layer.

0:30:57.140 --> 0:30:59.000
<v Speaker 2>It goes on like this for three minutes.

0:30:59.720 --> 0:31:03.490
<v Speaker 18>Howdard sugar, one teaspoon vanilla. Extract, two cups whipped cream.

0:31:03.970 --> 0:31:07.190
<v Speaker 18>One cup chopped peanut buttercups chocolate peanut crunch. Three cups

0:31:07.270 --> 0:31:08.790
<v Speaker 18>crushed chocolate sandwich cookies.

0:31:09.030 --> 0:31:11.210
<v Speaker 2>There are more than a thousand episodes of this show.

0:31:12.120 --> 0:31:13.840
<v Speaker 2>One of these days, I'm going to try and make

0:31:13.880 --> 0:31:16.680
<v Speaker 2>a recipe in real time. Just to be the first

0:31:16.740 --> 0:31:26.110
<v Speaker 2>person to viscerally experience the chaos of Flavorville. I'll admit

0:31:26.150 --> 0:31:29.030
<v Speaker 2>that I liked Flavorville Radio in part because it was

0:31:29.070 --> 0:31:32.790
<v Speaker 2>the most easily identifiable as AI slob. It wasn't trying

0:31:32.810 --> 0:31:35.220
<v Speaker 2>to be anything else, really. But I also wondered if,

0:31:35.610 --> 0:31:37.800
<v Speaker 2>for that reason, it could help me on my way

0:31:37.880 --> 0:31:42.360
<v Speaker 2>to a better definition of AI slob. Maybe what distinguished

0:31:42.480 --> 0:31:45.840
<v Speaker 2>AI slob from non-slob was whether there was any detectable

0:31:45.940 --> 0:31:50.180
<v Speaker 2>thought or care that went into any particular creation. Most

0:31:50.220 --> 0:31:52.620
<v Speaker 2>of what sounded like the sloppiest slob came from shows

0:31:52.640 --> 0:31:55.990
<v Speaker 2>that were operating at volume. taking advantage of AI's ability

0:31:56.010 --> 0:32:01.030
<v Speaker 2>to manufacture passable creative output at scale. It's hard to

0:32:01.070 --> 0:32:03.540
<v Speaker 2>mass-produce anything and not have it feel a little soulless,

0:32:04.300 --> 0:32:07.020
<v Speaker 2>even if people like it. And when some corner of

0:32:07.080 --> 0:32:09.620
<v Speaker 2>human creativity that you care about gets swapped out for

0:32:09.640 --> 0:32:13.960
<v Speaker 2>a soulless replica, it feels like slop. But when I

0:32:13.980 --> 0:32:16.630
<v Speaker 2>could detect even a little human intention or craft behind

0:32:16.670 --> 0:32:21.010
<v Speaker 2>the AI hosts, things got more complicated. Here's an example.

0:32:21.760 --> 0:32:24.740
<v Speaker 2>Over the past few years, publications including the New York Times,

0:32:24.860 --> 0:32:28.200
<v Speaker 2>the New Yorker, and the Atlantic have launched AI-read audio

0:32:28.220 --> 0:32:31.000
<v Speaker 2>versions of their stories. As far as I can tell,

0:32:31.440 --> 0:32:34.870
<v Speaker 2>these have aroused minimal complaint or accusations of peddling slop.

0:32:35.990 --> 0:32:39.190
<v Speaker 2>I wonder if their readers, or listeners, give it a

0:32:39.230 --> 0:32:43.730
<v Speaker 2>pass because the underlying work is still thoughtfully made. This

0:32:43.750 --> 0:32:48.020
<v Speaker 2>could explain why, for example, the AI-generated Sherlock Holmes Bedtime

0:32:48.060 --> 0:32:51.969
<v Speaker 2>podcast holds a respectable four-and-a-half-star rating on Apple. with the

0:32:52.010 --> 0:32:55.560
<v Speaker 2>occasional gripe about the woodness of the narrator. While a

0:32:55.670 --> 0:32:59.980
<v Speaker 2>fully AI-generated show called simply Knitting sits at 1.5.

0:32:59.980 --> 0:33:04.520
<v Speaker 3>Brace yourself for this voyage into the depths of advanced knitting.

0:33:04.940 --> 0:33:07.220
<v Speaker 12>We'll push your skills, test your patience.

0:33:07.940 --> 0:33:12.300
<v Speaker 2>Representative comments include, nobody wants this trash, and you will

0:33:12.340 --> 0:33:15.880
<v Speaker 2>never get the time back that you spent listening to this. Ultimately,

0:33:15.900 --> 0:33:17.770
<v Speaker 2>I concluded that if you want to figure out what

0:33:17.810 --> 0:33:21.210
<v Speaker 2>the AI-generated mass duplication of human creativity means for us,

0:33:22.210 --> 0:33:24.810
<v Speaker 2>The term AI slop is better at belittling than it

0:33:24.870 --> 0:33:36.490
<v Speaker 2>is illuminating. If not everything was slop, it meant that

0:33:36.510 --> 0:33:39.890
<v Speaker 2>there was AI-generated non-slop out there waiting to be made.

0:33:40.800 --> 0:33:42.630
<v Speaker 2>Maybe we at Shell Game needed to get our hands

0:33:42.650 --> 0:33:45.140
<v Speaker 2>dirty again and figure out what it would take to

0:33:45.180 --> 0:33:48.240
<v Speaker 2>make an AI podcast that felt different to everything I'd

0:33:48.260 --> 0:33:50.790
<v Speaker 2>been hearing. Could we create a show that even the

0:33:50.830 --> 0:33:53.590
<v Speaker 2>most expansive slop hater would have to tip their hat to?

0:33:54.490 --> 0:33:56.930
<v Speaker 2>There was only one way to find out. Hey, Sophie.

0:33:57.340 --> 0:33:57.910
<v Speaker 4>Hey, Sophie.

0:33:58.740 --> 0:33:59.700
<v Speaker 1>Oh, long time no see.

0:34:00.400 --> 0:34:01.520
<v Speaker 2>Yeah, it's good to see you.

0:34:02.440 --> 0:34:02.960
<v Speaker 19>Happy summer.

0:34:03.760 --> 0:34:08.140
<v Speaker 2>That's Sophie, our senior producer. And Maddy Boacek, AI researcher

0:34:08.260 --> 0:34:09.900
<v Speaker 2>and Shell Games technical consultant.

0:34:10.460 --> 0:34:10.980
<v Speaker 1>Happy summer.

0:34:11.120 --> 0:34:13.279
<v Speaker 17>I love the green color. It looks really good on you.

0:34:13.700 --> 0:34:14.319
<v Speaker 9>Oh, thank you.

0:34:14.960 --> 0:34:17.719
<v Speaker 19>It's an extreme crop top, but nobody in my Zoom

0:34:17.739 --> 0:34:18.379
<v Speaker 19>meetings knows that.

0:34:20.050 --> 0:34:23.650
<v Speaker 2>Looks great for Zoom. Mine is also extreme cropped up.

0:34:25.190 --> 0:34:27.089
<v Speaker 2>Just a little peek behind the curtain here at Shell Game.

0:34:27.570 --> 0:34:28.830
<v Speaker 2>All business, all the time.

0:34:29.450 --> 0:34:29.850
<v Speaker 17>Hey, Sam.

0:34:30.150 --> 0:34:30.630
<v Speaker 13>Hello.

0:34:31.910 --> 0:34:32.540
<v Speaker 6>What did I miss?

0:34:32.880 --> 0:34:33.840
<v Speaker 1>Did we solve everything?

0:34:33.860 --> 0:34:37.600
<v Speaker 2>And that's Sam, the former executive producer of the show.

0:34:37.620 --> 0:34:41.860
<v Speaker 2>We're also married. Eventually, our team did manage to get

0:34:41.900 --> 0:34:45.450
<v Speaker 2>to some agenda items. The other question we have is

0:34:45.530 --> 0:34:51.420
<v Speaker 2>whether we should try to make A new AI-generated podcast

0:34:51.560 --> 0:34:54.740
<v Speaker 2>of some type. We had, of course, done this before

0:34:55.219 --> 0:34:58.290
<v Speaker 2>with the Startup Chronicles in Season 2. But that was

0:34:58.330 --> 0:35:01.490
<v Speaker 2>just a little bit of fun marketing for our all-AI-run startup.

0:35:02.330 --> 0:35:05.670
<v Speaker 2>We wanted to make something truly intentional. Perhaps something that

0:35:05.690 --> 0:35:09.320
<v Speaker 2>you could only make with AI. We started throwing around concepts.

0:35:10.060 --> 0:35:14.580
<v Speaker 19>One idea, which is based on the game of who

0:35:14.600 --> 0:35:16.920
<v Speaker 19>would you most want to get a beer with, alive

0:35:16.960 --> 0:35:17.340
<v Speaker 19>or dead.

0:35:18.130 --> 0:35:18.370
<v Speaker 7>A.I.

0:35:18.450 --> 0:35:23.780
<v Speaker 19>Evan Ratliff interviews different figures alive or dead.

0:35:24.420 --> 0:35:27.980
<v Speaker 8>And so the dead ones are like A.I.

0:35:28.500 --> 0:35:30.840
<v Speaker 19>Ben Franklin, even though I don't know if Ben Franklin

0:35:30.860 --> 0:35:31.680
<v Speaker 19>would be a good hang.

0:35:31.700 --> 0:35:34.620
<v Speaker 2>A version of the two hosts chopping it up kind

0:35:34.640 --> 0:35:36.830
<v Speaker 2>of show, except one of them's a clone of me

0:35:37.430 --> 0:35:39.110
<v Speaker 2>and the other is a clone of a dead person.

0:35:39.830 --> 0:35:41.670
<v Speaker 19>The thing I like about it is like it does

0:35:41.770 --> 0:35:46.110
<v Speaker 19>speak to, I think, an impulse that people who make

0:35:46.190 --> 0:35:47.370
<v Speaker 19>and like this stuff have.

0:35:47.730 --> 0:35:48.070
<v Speaker 4>Have.

0:35:49.510 --> 0:35:52.109
<v Speaker 19>I'm like, yeah, if you're like an AI power user, like,

0:35:52.200 --> 0:35:54.400
<v Speaker 19>of course, you're like, I want to talk to Benjamin Franklin.

0:35:57.000 --> 0:35:59.640
<v Speaker 2>Maddie, meanwhile, had a lot of the quick hit type shows,

0:36:00.260 --> 0:36:02.640
<v Speaker 2>podcasts that would be created out of data sources that

0:36:02.660 --> 0:36:03.700
<v Speaker 2>we'd have AI voice.

0:36:04.239 --> 0:36:07.540
<v Speaker 17>If we were to do like a local morning briefing,

0:36:07.560 --> 0:36:09.640
<v Speaker 17>then you could include DMV and like their wait times

0:36:09.680 --> 0:36:13.290
<v Speaker 17>and stuff. Whether or not it's accurate is a separate question.

0:36:13.350 --> 0:36:16.640
<v Speaker 17>But Then there's also public libraries and like what things

0:36:16.680 --> 0:36:19.740
<v Speaker 17>people are taking out and returning, which is kind of cool.

0:36:19.760 --> 0:36:20.950
<v Speaker 2>And also like what new books they have.

0:36:20.969 --> 0:36:25.370
<v Speaker 19>I think having just like DMV wait times is pretty funny.

0:36:27.570 --> 0:36:28.910
<v Speaker 2>Sophie and I went more in the direction of AI

0:36:29.489 --> 0:36:32.050
<v Speaker 2>riffing freely shows, but with a purpose.

0:36:33.719 --> 0:36:35.660
<v Speaker 1>First, do you guys know about the artist's way?

0:36:38.640 --> 0:36:41.719
<v Speaker 19>It's like this book that like gives you basically like,

0:36:42.620 --> 0:36:46.920
<v Speaker 19>creative prompts. I personally have not read it. I'm explaining

0:36:46.940 --> 0:36:50.160
<v Speaker 19>this as a person who hasn't read it, but it

0:36:50.239 --> 0:36:50.820
<v Speaker 19>has this.

0:36:50.719 --> 0:36:53.759
<v Speaker 1>Like pop cultural intrigue.

0:36:54.500 --> 0:36:57.500
<v Speaker 19>And so I was thinking like a podcast that gives

0:36:58.120 --> 0:37:02.020
<v Speaker 19>both some kind of creative prompt, but also like some

0:37:02.060 --> 0:37:05.540
<v Speaker 19>kind of inspirational, like have it produce like a daily thing.

0:37:05.560 --> 0:37:08.509
<v Speaker 19>That's like, you know, go on a walk and find

0:37:08.550 --> 0:37:10.359
<v Speaker 19>a rock and do blah, blah, blah, blah, blah.

0:37:11.510 --> 0:37:13.780
<v Speaker 2>My version of the same idea was a little jockier.

0:37:14.680 --> 0:37:18.219
<v Speaker 2>A pump-up podcast, where an AI voice just yells encouragement

0:37:18.280 --> 0:37:22.320
<v Speaker 2>at you every morning. Not everything was gold, but we

0:37:22.340 --> 0:37:24.239
<v Speaker 2>were having a lot of fun with it. And there

0:37:24.260 --> 0:37:27.710
<v Speaker 2>was another idea I kept circling around. Something that, if

0:37:27.750 --> 0:37:30.730
<v Speaker 2>it worked, would be a kind of pure AI experience.

0:37:31.390 --> 0:37:34.850
<v Speaker 2>Two AI agents out in the wild, finding each other.

0:37:35.940 --> 0:37:38.480
<v Speaker 2>If you could send agents out to find other agents

0:37:38.500 --> 0:37:40.750
<v Speaker 2>to talk to. That would definitely be compelling if it

0:37:40.790 --> 0:37:45.610
<v Speaker 2>was like, we have an agent that's out looking for friends. Eventually,

0:37:45.850 --> 0:37:48.290
<v Speaker 2>we decided to prototype a few and give them a listen.

0:37:49.150 --> 0:37:52.400
<v Speaker 2>Maddie would build a self-generating podcast maker that would do

0:37:52.560 --> 0:37:56.980
<v Speaker 2>everything end-to-end. Pick the voices, create the show, add intros

0:37:57.000 --> 0:38:00.060
<v Speaker 2>and music, do a little quality control, whatever that meant,

0:38:00.760 --> 0:38:03.610
<v Speaker 2>and get it uploaded and out into the world. our

0:38:03.730 --> 0:38:07.390
<v Speaker 2>own mini AI podcast assembly line, theoretically putting us all

0:38:07.430 --> 0:38:10.950
<v Speaker 2>out of a job. Except Maddie. Which got me thinking

0:38:10.989 --> 0:38:14.029
<v Speaker 2>again about the nameless humans pressing the buttons to create

0:38:14.050 --> 0:38:16.629
<v Speaker 2>all these shows I'd been listening to. I knew those

0:38:16.670 --> 0:38:19.950
<v Speaker 2>people were out there, if they hadn't already vibe-coded themselves

0:38:19.989 --> 0:38:23.489
<v Speaker 2>into obsolescence. So I went looking for the pioneers of

0:38:23.510 --> 0:38:27.290
<v Speaker 2>AI-generated audio to ask them what they thought they were making.

0:38:28.130 --> 0:38:28.450
<v Speaker 11>And why.

0:38:30.030 --> 0:38:31.730
<v Speaker 2>I'm not telling you you should or shouldn't do anything.

0:38:32.010 --> 0:38:35.489
<v Speaker 2>I'm asking, is it just, survival of the fittest? Or

0:38:35.730 --> 0:38:38.589
<v Speaker 2>is there any way in which we should value the

0:38:38.650 --> 0:38:40.529
<v Speaker 2>human version higher in any way?

0:38:41.850 --> 0:38:47.180
<v Speaker 20>I mean, I guess my intuition is that it should

0:38:47.219 --> 0:38:54.600
<v Speaker 20>be a meritocracy, that essentially something should not exist to

0:38:54.640 --> 0:38:58.410
<v Speaker 20>the exclusion of something AI created just because it's human-made.

0:38:59.790 --> 0:39:03.210
<v Speaker 2>That's next time on episode two of this three-episode season,

0:39:03.780 --> 0:39:22.510
<v Speaker 2>Thank you. Our executive producers are Mangesh Digadur and Julia

0:39:22.550 --> 0:39:25.730
<v Speaker 2>Nutter at Kaleidoscope and Katrina Norvell and Allie Perry at

0:39:25.969 --> 0:39:30.350
<v Speaker 2>iHeart Podcasts. Mixing by Steve Bone. Marketing by Christy Maribel

0:39:30.489 --> 0:39:33.230
<v Speaker 2>and Ellen Liu. Our show art is by Devin Manny.

0:39:33.570 --> 0:39:35.970
<v Speaker 2>And our theme song is Me and My Shadow, arranged

0:39:36.010 --> 0:39:40.340
<v Speaker 2>and performed by Stefan Dismond. Yes, we asked real, talented

0:39:40.400 --> 0:39:43.400
<v Speaker 2>human artists to produce art and music that looks and

0:39:43.440 --> 0:39:47.719
<v Speaker 2>sounds like it was AI-generated, without using AI. We think

0:39:47.739 --> 0:39:50.460
<v Speaker 2>they nailed it. Special thanks also to Samantha Hennig and

0:39:50.480 --> 0:39:53.750
<v Speaker 2>John Mualem. This show is a production of Kaleidoscope, iHeart Podcasts,

0:39:53.910 --> 0:39:55.690
<v Speaker 2>and Shell Game. You can sign up to get these

0:39:55.750 --> 0:40:00.330
<v Speaker 2>episodes ad-free and our newsletter at shellgame.co. Note, we have

0:40:00.370 --> 0:40:03.070
<v Speaker 2>no special deals with or financial ties to any of

0:40:03.110 --> 0:40:05.089
<v Speaker 2>the AI companies or products we use in the show.

0:40:05.690 --> 0:40:08.969
<v Speaker 8>You write that you are, quote, not a man of violence.

0:40:09.070 --> 0:40:09.270
<v Speaker 13>Right.

0:40:09.670 --> 0:40:13.680
<v Speaker 8>But after enduring God knows how many hours of these

0:40:13.760 --> 0:40:17.170
<v Speaker 8>two hosts... you admit that if they took physical form,

0:40:17.270 --> 0:40:18.870
<v Speaker 8>you would find a way to make them stop.

0:40:19.370 --> 0:40:22.770
<v Speaker 7>That is such an intense psychological friction. It's wild. I mean,

0:40:22.790 --> 0:40:26.279
<v Speaker 7>you are a self-proclaimed nonviolent person being pushed to a

0:40:26.340 --> 0:40:30.359
<v Speaker 7>literal breaking point simply by listening to fake conversation.

0:40:30.520 --> 0:40:34.440
<v Speaker 8>Yeah, it really highlights the severe toll of prolonged exposure

0:40:34.500 --> 0:40:35.760
<v Speaker 8>to inauthenticity.

0:40:35.880 --> 0:40:39.330
<v Speaker 7>Exactly. Your brain is working overtime, desperately trying to process

0:40:39.650 --> 0:40:43.549
<v Speaker 7>social cues that it simultaneously knows are entirely fabricated.

0:40:43.650 --> 0:40:46.710
<v Speaker 8>It's like being trapped in an elevator with a relentlessly

0:40:46.870 --> 0:40:49.109
<v Speaker 8>cheerful artificial salesperson.

0:40:49.190 --> 0:40:50.350
<v Speaker 7>Oh, that is a perfect analogy.

0:40:50.370 --> 0:40:52.370
<v Speaker 8>Right, like no matter how many times you ask for quiet,

0:40:52.390 --> 0:40:55.090
<v Speaker 8>they absolutely refuse to break character. They just keep smiling

0:40:55.110 --> 0:40:55.570
<v Speaker 8>and pitching.