Hosts
About the episode
So many people seem to hate AI, and it’s not just because it might displace millions of jobs, destroy the world, or deepen distrust in big corporations and government. The reason might actually be simpler than we think: People don’t like AI because it is so often used to make things that are fake.
From sham biographies and bogus social media posts to AI-generated journalism, the internet has become inundated with “AI slop.” So how do we begin to clean up the mess? That’s where AI detection company Pangram comes in. Derek talks with Pangram founder and CEO Max Spero about the seduction of AI writing, how his program detects it, and the moral implications of publicly calling out those who use it.
Subscribe to our YouTube channel here.
If you have questions, observations, or ideas for future episodes, email us at PlainEnglish@Spotify.com.
Host: Derek Thompson
Guest: Max Spero
Producer: Devon Baroldi
Additional Production Support: Ben Glicksman
In the following excerpt, Derek and Max Spero discuss bots and dead internet theory, and AI-generated content and its implications for the written word.
Derek Thompson: Tell me a little bit about your background. What were you studying in college? What did you do right after college? And how did you come up with the idea for Pangram?
Max Spero: I studied computer science in college. I loved programming. I did a bunch of things around robotics. I really just loved making things happen in the real world. And then I got into machine learning because I was interested in robotics and that’s the natural next step, instead of hard coding a trajectory, how do you make an algorithm that can learn and be more natural?
Ultimately, what ended up happening is I went to Google, training big machine learning systems basically for ads essentially. And I learned how to do ML there. And then I went to Nuro, self-driving car company. ChatGPT came out during my tenure at Nuro, and I realized this is really going to be important technology and it’s going to change the world in a lot of ways, not all of them necessarily good.
Thompson: And then, tell me what you wanted to do with this company. What did you initially see as the problem that Pangram was going to solve back in 2022 or 2023?
Spero: I think early on I was thinking a lot about bad actors and what they could do with AI as essentially this infinite volume of plausible text. I was thinking somebody could go write hundreds of thousands of fake reviews, somebody could go create totally fabricated news websites. That was political misinformation, disinformation, could just be supercharged with AI. It’s mostly what I was thinking about.
Thompson: It’s funny because most people I think locate the scale of AI risk at the level of apocalypse. AI will destroy all the jobs, AI will kill the people. The idea that AI will publish a lot of dreck on Twitter and LinkedIn doesn’t quite rise to the level of an existential risk. Why was it so important to you that we be able to detect AI writing? What did you see as the real risk here?
Spero: Obviously, humanity is not going to go extinct if we can’t tell AI writing from human writing, but I think what we are doing is we’re trying to save the internet. We’re trying to save the written word and communication. I think if you look at the scale of AI and bot traffic on the internet, it’s gone from a small fraction of human traffic to recently passed 50/50. So, 50 percent of the internet traffic is bots. And I think it’s not that long before it’s going to be 99 percent of internet traffic is bots and not just readers either, but writers, communicators, agents that are trying to get things done on behalf of their owners.
I think we’re very much at risk of this idea of dead internet theory where the internet is just this echo chamber of bots talking to bots and it’s going to be impossible to parse out the signal from the noise when there’s so much AI chatter.
Thompson: I like that answer. I would add that I’m worried about the relationship between humans and humans. Something happens to the way that you can speak to another person on the internet authentically when you plausibly believe that their output might just be ChatGPT or Claude. I mean, we’ve historically had this association whereby if it’s written, then a human wrote it. If it’s written by and signed by Derek Thompson, therefore Derek has in his head an understanding of what he wrote. But this idea that everyone suddenly has an AI ghost writer—
Spero: Yeah, and it’s quite good. It’s quite plausible.
Thompson: … means we don’t know who we’re dealing with when we read certain posts. That to me, scrambles and weakens the relationship between people on the internet, which further deadens the dead internet theory that you’re pointing to.
Spero: Totally. I think previously well-written text was plausible. I see something that’s well-written, looks like a lot of effort was put into it, now that’s like I see something that’s well-written and then I have to question, well, is this AI-generated? Did this person actually put any time into writing this in the first place? I think that sucks in general, that I even have to ask that question.
Thompson: Just in the last few days, The Wall Street Journal reported that, “AI has plunged the book publishing industry into utter chaos.” Just this morning, The New York Times reported that Spotify, LinkedIn, and others are, “… trying to dig out of a digital sewage heap full of low-quality content made by artificial intelligence.” Within that article, they cited Pangram, which reported that AI was used to create nearly half of the posts with more than 50 words on X or Twitter.
That is crazy, this idea that if you read a post with more than 50 words on Twitter, you can flip a coin and heads means it was written by AI or odds worse.
Spero: I’m not sure if that got misquoted somehow, but on Twitter … Our study found that 29 percent of long form-content, so 250 words or more, 29 percent on Twitter was AI-generated. And then of short-form content, 50 to 250 words was about 9 percent.
Thompson: I see. Well, good to have fact-checked The New York Times here in a way that is live. Are there other ways that you can describe how bad the problem is, the problem of more and more of what we’re reading on the internet essentially being written by bots?
Spero: Yeah. It’s not just the social platforms. I think they obviously have this high degree, this big problem. LinkedIn had 41 percent of their long-form content was AI-generated. But if you look at the internet at large, we did this study with the Stanford and the Internet Archive, which found that in May of last year, already 40 percent of the internet was AI-generated. And I think this number is just going to keep expanding.
There’s people trying to pollute the internet for different reasons, whether it’s they want their own narratives in the LLM training data or they’re trying to win at SEO and now it’s easier than ever to produce AI content for search results.
This excerpt has been edited and condensed.
Host: Derek Thompson
Guest: Max Spero
Producer: Devon Baroldi
Additional Production Support: Ben Glicksman

