Skip to content
Drawing the Line: Medium's Policy Shift Bans AI-Generated Content from Distribution episode artwork

Jan 21, 2024 13 min

Drawing the Line: Medium's Policy Shift Bans AI-Generated Content from Distribution

from The Daily AI

In this episode, we explore the nuances of content creation policies as Medium draws a definitive line, banning the distribution of AI-generated content. Tune in for insights into the platform's decision-making process and the broader implications for AI-generated content in online spaces. Invest in AI Box: https://Republic.com/ai-box Get on the AI Box Waitlist: ⁠⁠https://AIBox.ai/⁠⁠ AI Facebook Community Learn more about AI in Music Learn more about AI Models

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

Drawing the Line: Medium's Policy Shift Bans AI-Generated Content from Distribution

0:00 13:40
of matches

Transcript Auto-generated, may contain errors

Medium.com, the article writing website that's been famous with bloggers for over five years is recently putting its put down when it comes to AI. They have announced that AI-generated articles are not welcome on their platform. So today on the podcast, we're going to be diving into why this is an interesting move by the company, and I think more broadly, how this is going to impact AI and the AI space in general, what it says about the sentiments around AI. So without further ado, let's jump into it.

So this news comes as they put out a statement saying, we stand as a platform for human-created content. That's Scott Lamb, who's Medium's content director, and he argues in a company blog post essentially that readers have an inclination towards human-generated text instead of AI, of course. And I think a lot of people, broadly speaking, have the same sentiment. I will say, though, that being said, a lot of times when we read AI-generated content, we may not know.

Like, honestly, I think it'd be kind of crazy to have some sort of way to track, you know, how much of the text we read on the internet is AI-generated that we think is generated by humans. Do I think that's good or bad? I mean, I will weigh in on that. But I will say, you know, humans having that inclination towards human-authored text, I think we do, but I also don't think we always know when something's not human-authored.

And in a blinds test, right, if you put the blindfold on and sip the Coca-Cola and sip the Pepsi, and everyone said they like Pepsi better even though Coca-Cola is more popular, I mean, I think you may find something similar with AI-generated text. Although, like, conceptually, I would much prefer the concept of human-generated text. It's way higher quality. I human-ishly just sit there and put some thinking into it.

But I think that's another kind of debate. In any case, while Medium doesn't actually outright ban AI-assisted writing, I think it's safe to say the platform is really not, you know, extending a warm welcome to it. So, in Lam's view, text produced solely by artificial intelligence is a poor fit for Medium. He said, quote, the very best AI-generated content is still bland, lacking in human insight, and riddled with subtle imperfections.

So he really does not like it. And I think, you know, I'm not saying all the things he says are far off. The thing where it says it's still bland, like, sure, if you just get a stock response from ChatGPT, he might be bland, but if I say write it with an exciting, interesting twist and add puns and do blah, blah, blah, blah, I can make it quite an interesting, you know, piece of work. So I really think that's just up to the prompting.

That I don't think is fair. Then when he says that it is lacking in human insights, perhaps, that might be fair. I mean, really, it's trained on human insight. So you can make an argument for that either way.

I'll give it to him. And then riddled with subtle imperfections. That one definitely could be true, right? We know this thing.

AI hallucinates. It's not perfect. There are imperfections in it. That being said, I would never, you know, recommend someone get something generated by AI content and just spit it out there.

I'd always say, you know, review it as a human, get a copywriter to look over it, make some tweaks, but, like, honestly, a lot of times it's mostly pretty good. But anyways, that's just from my background is that I've used AI to write thousands of articles for blogs on different websites and done, like, seen some really crazy success doing, like, SEO. And again, I know a lot of people think that that's controversial. Some people might not like that.

But regardless, love it or hate it, it's possible, it's viable, and it brought a lot of business to the business I had. So inevitably, you know, my motivation incentive was to continue doing it, and I think that's going to be the motivation incentive for most businesses, right? If they find something that works, people that read it, liked it, you know, got a lot of traffic to those pages. And the most popular pages would actually go send a human writer to read over and add extra context and extra human insights and, you know, spiff up the page, make it higher quality.

But in the initial launch, it was all, you know, almost exclusively AI generated. So, you know, I've seen it be successful there, and of course, I think that's what they're trying to avoid at Medium is to have, like, a ton, like, a flood of AI generated content. And I actually think this is, perhaps this is a smart brand move on their part, you know, saying, like, we're just human generated, so when you go there, you feel like it's, like, kind of this elite, high quality place. Okay, so then, so that's the thing from the branding side.

So I do think it's smart, probably, on their side to say this. What I will say is that it's probably not possible for them to detect what's been written by AI, especially if you do a little tweaking or whatever. But let's get into that, because I do have some insights and thoughts on that. Medium's current filtering system, which essentially uses humans, is designed to intercept a significant amount of AI content and limit its distribution, right?

That's our idea, it's the AI content. I don't think we're going to ban it, but they're just going to kill its distribution and not let it get distributed anymore. The platform is actively developing additional methods to detect AI off their content, it says. But I will also point everyone to the fact that OpenAI, who's literally the creator of ChatGPT, created an AI detection software and then shut it down because it wasn't very good and they didn't do a good job and they found it was too difficult.

So for many reasons, OpenAI shuts down their AI text detector and they're literally the platform that generates the AI text. I think Medium's going to have a hard time. I've seen a lot of tools that, right, a lot of tools that are like AI detectors, but literally if you just grab an article at ChatGPT, throw it into, throw it into a bunch of tools that take your AI generated content and kind of change out some words in there and then all of a sudden it's undetectable by AI. I've also done that and I've seen really high success rates on that.

So I have a hard time thinking Medium's going to be able to, like if I was, I'm not saying this is good, I obviously don't like just tons of spam articles on there, but if I was a spammer trying to get tons of content on there, on Medium for example, I'd run it through ChatGPT, then I'd run it through one of these tools that kind of spins some words in it and then post it and it's undetectable virtually. So I don't think that they're going to be able to actually do that. Maybe the best bet is just saying we're going to limit reach and then hopefully scare people away from stopping to try. Don't know if that'll work, but anyways, I think it's interesting.

If you want to try experimenting with AI generated content on your own personal Medium blog, you're still free to do so, they said. However, they just said, they said like, don't expect it to explode and get super big on the platform if it's just purely AI written article. Medium, they said that they don't think AI is taking the place of human writers and they said that that kind of content is going to only be distributed within the writer's personal network, meaning if you have an account on Medium, unless you've got a thousand subscribers on there, you write it, you know, your thousand subscribers might see it, but they won't show it to anyone else. That's assuming they can detect the AI.

I'm pretty skeptical of that. I'd be impressed. I'm actually probably happy if they can come up with a good tool that could do that. I'm just skeptical of opening.

I couldn't do it that, you know, Medium would be able to. Although that being said, they could train, you know, they have like a massive data set of human written articles to this point. And so they could train an AI model off of all the content there. And I mean, perhaps that would be a way that they could, like essentially, they could train their own tool.

And if they find one that really works, I guarantee there's a lot of other companies willing to license that. So it could be a big money maker if they could get that, if they could solve that problem. Medium does permit the responsible use of AI-assisted technology as part of its kind of quote commitment to transparency and setting reader expectations. The platform actually requires that any AI-assisted story be labeled, so you have to say it's written by AI.

This is actually something I've, you know, recommended for a lot of things, whether that's videos, images, or text. The same way that, you know, the same way that you're required to say sponsored or hashtag ad on like an Instagram post, if there's a sponsor for the product that you're talking about in there. I think, you know, saying hashtag AI isn't a horrible thing to notify that a poster content was generated by AI. Will people do that?

Will people argue about that? Will lawsuits come inevitably? But I mean, I think transparency is always good, so that's my opinion on it, and I'm sure there's lots of different thoughts and theories on that. If AI-assisted text lacks a disclosure in the opening two paragraphs, Medium says that if they don't do that or other AI-generated content isn't properly labeled, the distribution is going to be limited to the person's personal network and no one else on the platform is going to see it.

So as according to Medium's revised distribution guidelines, AI content must now be clearly flagged to be eligible for the distribution. What they're specifically talking about when they say distribution is the fact that your article, if it does really well on Medium, they can be trending on the homepage, they can send it out in emails, as newsletters to people. There's all sorts of ways that they kind of promote trending topics and trending pages, and in addition to that, they also have it where you can make money off of Medium articles and put them behind a paywall, so I'd be curious to see what their policy is there if they don't let you put AI content behind a paywall and whatnot. So I think that this is going to be really, really interesting.

I think that the progress in AI technology and kind of the associated tools are really prompting content platforms to kind of reassess and broaden their policing concerning AI, and I think case in point is image platforms like ArtStation and Coding Community Stack Overflow, right? They both are doing this, and the latter just found itself inundated with low-quality responses generated by ChatGPT, which essentially just led to a stack overflow to ban any AI-generated responses on its platform. So Medium 2, I think, fell victim to a wave of kind of AI-authored posts when Medium sought the community's view on AI. It received kind of a mix of both positive and negative feedback, that's like according to Scott Lamb, that's what he said in a blog post, and according to him, the digital world is currently ongoing a really seismic shift, and the possible implications of AI-generated content are both really far-reaching but also not super clear, and he thinks that they need to be addressed now.

So in 2021, Medium transitioned to a new business model, which essentially was incorporating paid content, and they're doing this so writers can monetize their content, Medium can make some money off of this, and the move at the time, back in 2021, invited a lot of criticism from subscribers who are really reluctant to fund low-cost AI content through monthly subscriptions, allegedly. So Medium has yet to respond to this concern, but made it clear that this is only the beginning, so as an initial measure, the platform is focusing on implementing a transparency policy, and I think this might be the first step. So perhaps the problem, and the reason why they're doing all this is because there's a paywall where you have to pay to see certain articles or you get paid, right, so someone has a subscription, they can see all of the articles, but the author's getting paid, and so I think the idea here is you can go make a Medium account, make a premium one, so everything's paywall, flood it with AI-generated content, and then people, without knowing, might click on your article to read about your topic or whatever, and then you're going to get paid for it, and then maybe halfway through, they're like, oh, this is generated by AI, I feel duped, but then the AI author already made money, maybe that's kind of the whole problem here. Lamb kind of explains their approach, he emphasized that as a technology and its applications continue to evolve, Medium's policies are likely to follow suit, and by fostering a culture of disclosure around AI-generated content, Medium believes it can empower its readers, and I think that even is interesting, right, maybe if they have to put it in the first sentence or something and it's in the description, people won't even click on something they see it's generated by AI, which of course AI-generating authors would be concerned about, but maybe people that don't want to read AI content would appreciate, especially if the author's going to get paid for it, and they don't feel like they want to support that.

So I mean, I always think that more transparency is better, so I don't really see much harm in that, in itself. I think that tech suspected to be AI-generated is essentially going to still be accessible via a direct link, so there is no hard ban, people will still be able to write stuff and share it, but it's just going to have a little bit more disclosure, which I think is probably a good thing. It'll be interesting to see if other companies are doing this in the future beyond just Medium, Stack Overflow, and a few others doing these outright, essentially not allowing or not really wanting AI content on the platform, so it'll be very interesting to follow. Can you imagine if YouTube or some other platform did something similar?

I think this would be a very, very interesting space to follow in the future. If you are interested in investing in a new AI startup, you can go to republic.com slash AI-box, I'll leave a link in the show notes as well to learn more. The minimum investment is $150, and the maximum investment is $100,000. Until next time, have a fantastic day.

Frequently asked questions

How long is this episode of The Daily AI?

This episode is 13 minutes long.

When was this The Daily AI episode published?

This episode was published on January 21, 2024.

Is there a transcript available for this episode?

Yes, a full transcript is available for this episode. You can read the complete transcript on the episode page.

Can I download this The Daily AI episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.