You may think that; I couldn't possibly comment

Plus AI gets into Conde Nast, and films gets weirder

“A woman looking at a website on a computer: the computer is throwing out small pieces of paper with angry faces on them.” Generated by Ideogram 2.0 (see end.)

When you arrive at the bottom of a story on a news site, you will—especially if the story appears in the “Opinion” section—often find yourself effectively being invited to comment on what you’ve just read. Come on, you’re not just a reader now: you’re also one of the people with the power to make your voice heard! At last you can stick it to The Man!

So what’s your reaction when you reach that point and you find something like this greeting you to tell you how many comments have already been left:

Yes, that’s 1,500 comments (or thereabouts) which have accreted below the article. It’s from the New York Times, and I think it was about something vaguely political.

So when you see that number, do you think: wa-hey! This must have some fun going on, it’s time that people heard my view on this too.

Or do you think: there cannot possibly be any value in being the 1,501st comment below this story.

My guess is that you fall into the latter category. And yet there are plenty of people out there who do not. They see an article with 200 comments—arguably more than even a quite devoted reader who reaches that point is going to bother with—and think to themselves what this really needs is my opinion, because that’s what people have been missing in all this.

The real problem with such below-the-line (BTL) input (as it’s known because it’s under the line marking the end of the article) is not really that there’s a lot of it. We’d all love it if the input to a piece on, say, economics all came from economics experts, or something about dark matter had the best cosmologists weighing in with nuanced expansions on the what’s above. But you know that’s not what happens. Instead it’s a sort of demonstration of chaos theory, or a drunkard’s walk which will on average never deviate far from the opinions expressed in the first few comments: even if there are some valuable gems thrown in here and there, they will be lost in the general mélange.

The problem with BTL commentary is that, as with so much of the rest of social media, there’s no good weighting algorithm. Many sites try to use “likes” or similar as a method of implying to people that some remark is perhaps better, in some ineffable way, than another one. This doesn’t, in fact, work, except to show you that what people think on a subject is entirely unpredictable¹.

One does have to ask: what is the purpose of having comments underneath news articles? They’re highly unlikely to lead to people being better informed: even though you will get the occasional knowledgeable commenter, they’ll almost certainly be drowned out by people who are wrong telling them they’ve got it wrong. (What isn’t often pointed out about the Dunning-Kruger curve is that there are lots more people down in the You Are Wrong part than the Yes, That’s Correct part: the ratio’s about 3:1.)

What’s more, these comments don’t look after themselves: in the UK particularly there is the risk of people saying libellous things, or generally causing trouble (especially if they begin doing things programmatically: we had one commenter at The Guardian who would DDOS the comment section of the Technology pages with “comments” consisting of tons of empty lines and the occasional word. Quite mad, but also quite difficult to stop; in the end there was a concerted effort to block various IP proxies that were being used to do the posting, plus rather more careful verification of which domains could be used to register the email address required for a new account. Turns out there are a lot of domains offering completely throwaway emails.) All this means that you need some human moderators, who are given the tedious job of reading the comments to make sure people are in a general sense behaving themselves. At some outlets they get the bonus of picking particular comments to be highlighted—The Guardian, again, does do this, showing such comments above all the other ones in its own space. The NY Times doesn’t, instead opting to let readers do this by “recommending” comments. Here’s what they rated as the top comment on an article titled “He Made a Game About a Joyous Journey. He Also Got a Bit Lost².”, a sensitively told story about a nine-year effort by Anthony Tan, now 25, to produce a video game about a deer:

This was not written by me. Apart from anything, I know how to punctuate and write grammatical sentences.

What can one say, exactly, apart from Are you absolutely certain you’re picking the very best here, NYT readers? In this case, there were 50 comments which were Readers’ Picks. And that’s out of… 55 comments. Which makes it seem that there’s quite a low bar for being Picked. No news organisation that I’ve come across has any sort of algorithm that spots recurring commenters and highlights them on the basis of their useful contributions. (Or, similarly, spots repeated troublemakers and pushes them out of sight.) You’d think that might be a worthwhile investment, if you’re going to have comments?

Anyhow! We’ve established that BTL content is random, troublesome, and incurs expense in some way or another.

But we haven’t answered the question: why have comments at all? Lots of articles don’t, in fact. The Guardian used to have them under almost every story, but I suspect that moderation cost eventually precluded that. Also, the real task had been done, which was the purpose of having comments in the first place: getting people who had come to the site to stay there. No matter that commenters are somewhat less than 1% of readers; the idea is that while they’re writing, they’re spending time on your site rather than somewhere else, and not only that—they’re generating content that the other 99%+ might well read.

All of which means one thing: more opportunities for people to scroll and be shown adverts. The only calculus then for the news organisation is whether the cost of moderation outpaces the revenue from adverts (in which case you choose less moderation or allow fewer comments by closing them earlier on the stories where you do allow them).

What isn’t considered in all this is the broader effect—what economists call the externalities. Mic Wright, when he was a columnist at the Telegraph’s blogs, described comments as “the radioactive waste of the Web” because of the fact they were so often toxic to a greater or lesser extent (Mic’s readers, if they did actually read his pieces, tended not to be fans), and hung around for an indefinite period, during which they might be ingested by search engines and further pollute the wider atmosphere.

You might wonder why the preceding paragraph doesn’t have a link to Wright’s comment, or the Telegraph’s blog. There’s a simple reason: in 2014 the Telegraph deleted everything from the blogs wholesale, including the comments, and shut them down. It’s the web equivalent of putting all the nuclear waste into a rocket and firing it into outer space. (Please don’t try this with the nuclear waste on your home planet).

But! After a bit of spelunking around the Internet Archive, I found the article, written in September 2012:

At their worst, comments are like toxic waste buried under the foundations of an article and irradiating all rational debate with ignorance and aggression. And, like radiation, the effect of the internet commenting culture is spreading. The degradation of discourse online is mirrored in real-world dialogue. Adults who would balk at bullying in school playgrounds are happy to fling snide and often extremely aggressive comments around.

(Side note: the Telegraph blogs were quite a weird space, though perhaps not that different from now. Here’s a page I picked at random from April 2013, via the Internet Archive.)

The toxic nature of so much commentary (apart maybe from that person who got top ranking in the NYT’s story) is definitely an externality which is borne by those outside the paper. For readers, there’s the potential for infuriation reading remarks which show that their author hasn’t understood the topic. (This applies even if the author does understand the topic, and it’s just the reader who doesn’t. Dunning-Kruger applies!) There’s also a more subtle, dare I say, social warming effect: if your comments section is infested with people who say absolutely bonkers things, then you will judge the news organisation by its commenters. (Much the same is happening on eX-Twitter with its blue tick denizens, prompting an exit by lots of people towards Threads or Bluesky.) That’s potentially harmful to the news organisation’s reputation, even in a sideways fashion, because the brand is also about what sort of people read it.

On that basis, comments are potentially one of the most important parts of a site. “Never read the comments”, journalists tell each other, as much as anything because they don’t want to have to think about who is posting. (I was incapable of not reading the comments. And almost incapable of not commenting on the comments. I think I have a naturally feisty nature.)

But there is one thing that could always get me to turn away from a comments section, and which has a similar effect when I see a tweet or other piece of content which seems to be begging for engagement by its rank stupidity. It’s a number. Such as the one on the lower right of this bit of content at The Guardian on Thursday:

OK! Fine! There’s no point in my being the 410th comment there. What the hell difference would it make? Similarly, there are so many tweets and similar that I see which show double-digit responses already there, to which my immediate response is: fine, there’s nothing to be added there. If everyone has the exact opposite opinion to me, one differing view won’t change things; if everyone has the same opinion, it won’t change anything; if it’s evenly split, nothing will change.

I do have one thought, though. The news organisations need the comments so that people will come and dawdle a little longer on the site (adverts!). But equally, they’d prefer them not to be too toxic.

There’s an obvious solution: get a chatbot to generate a slew of comments that can be copypasted under a story in short order. A little light editing—run them through some sort of grammar checker—and you could probably get a decent run of pro- and anti- views. This would render any need for input from actual humans quite redundant, though you might need to allow some of them to go in there and “contribute”. (You could always tell them that their comment is “awaiting moderation” and never publish it.) Minimal moderation cost, because the job essentially shifts to feeding a chatbot, and also you up the reputational appearance of your site, since it won’t have benthic crazies looning up the place.

I’m not saying it’s a good idea. Though it is an idea. Please don’t blame me.

---

Glimpses of the AI tsunami

(Of the what? Read here. And then the update.)

• Ideogram (which generated the picture at the top) got busy. Apps are a thing now, eh: We are excited to release Ideogram 2.0, our new frontier text to image model with industry leading capabilities in generating realistic images, graphic design, typography, and more. Trained from scratch like all Ideogram models, Ideogram 2.0 significantly outperforms other text to image models across many quality metrics, including image-text alignment, overall subjective preference, and text rendering accuracy.

Today's milestone launch also includes the release of the Ideogram iOS app, the beta version of the Ideogram API, and Ideogram Search.

• “No one’s ready for this”: The Verge points out that the easy availability of Google’s Magic Editor on the Pixel 9 mean we can’t really trust any photo that we’re shown. This has been true since Photoshop, of course, but it’s the ease which is different.

• Condé Nast joins “other publishers” in allowing OpenAI access to its content. To train future AI language models. Wonder if those models will be able to write short adversarial comments? It’s a little unclear whether comments on CN articles are infused into OpenAI under this deal.

• This tweet is titled “AI is replacing LSD dealers”, and once you’ve watched the short AI-generated video there, you won’t disagree.

---

• You can buy Social Warming in paperback, hardback or ebook via One World Publications, or order it through your friendly local bookstore. Or listen to me read it on Audible.

You could also sign up for The Overspill, a daily list of links with short extracts and brief commentary on things I find interesting in tech, science, medicine, politics and any other topic that takes my fancy.

• I’m the proposed Class Representative for a lawsuit against Google in the UK on behalf of publishers. If you sold open display ads in the UK after 2014, you might be a member of the class. Read more at Googleadclaim.co.uk. (Or see the press release.)

• Back next week! Or leave a comment here, or in the Substack chat, or Substack Notes, or write it in a letter and put it in a bottle so that The Police write a song about it after it falls through a wormhole and goes back in time.

1. When I worked at The Guardian and used to—probably unwisely—join in the discussion, I was frequently accused of manipulating the number of upvotes my comments received. Which was bizarre, because I didn’t, and didn’t have any idea how or why anyone would want to. Those things are not a popularity contest. But people will find any excuse to be annoyed.

2. I swear someone needs to be ritually killed over Americans’ use of capital letters in headlines. It’s not as if they do it elsewhere in any sentence they write, is it? So why the hell in headlines, where they use up valuable space?