What caught my eye this week.
Working at the end of the age of publishing words has given me a lead on the post-LLM era.
I saw early on how ChatGPT had mined the web for everything ever written – well-enough to spit out answers about anything. And as a writer I had more incentive than most to panic.
It was also clear that Google search would be in trouble – and with it the pipes that had kept independent publishing alive on the web for decades.
My worries soon came to pass. People increasingly now get their knowledge direct from chatbots – whether Google or others. Those who wrote the articles the bots were trained on are withering on the vine.
Another thing I’ve wondered about is when AI spam will overwhelm the Monevator comments. Already on platforms like X, swathes of comments are written by robots.
We have protections in place. But I don’t know how long they will be practical when facing spam like this:
Such spam started appearing in the past month or so. It addresses me or my co-blogger accurately. It references the article.
Only the booby-trap at the end confirms its ill-intentions.
Check mate
You may say there’s something sloppy about this text. (Not to mention that it reads like @TA’s mum had a hand in it…)
Agreed, but remember you’re only have to sanity check one comment here.
I have to parse several hundred spam comments every day as a double-check. Both on spam that gets through our filters or is held for moderation, and also real comments that are incorrectly marked as spam. This is after software has already flagged the obvious offenders.
It’s burdensome, and the reason why I had to close comments on posts over three years old. To keep it vaguely manageable.
Spam comments like the one above stand out because they are still rare. But I imagine they will soon be the norm. (Well, presuming the economics of spamming still works if spammers are somehow paying for AI compute?)
I also expect bots to get clever enough to hide their intentions by posing as real readers, before finally inserting their spam links once they’re trusted.
Incidentally, we can see that’s a spammy link in my example. But if a reader posts a URL to data elsewhere about interest rates, say, it’s not so easy for software.
That’s why comments with links are already often held in moderation, especially from new commenters.
King sacrifice
Long story short: one day only logged-in Monevator members may be able to post comments. (I’m presuming the spammers won’t pay for the privilege!)
I’d be happy for commenting to be another perk for those who kindly support our efforts. It would make general moderation far easier, too.
Really, everyone who comments regularly on Monevator should already become a member. It costs much less than a High Street coffee a month. Even cheaper with annual membership!
With member-only commenting I know we’d lose some good comments, sadly. Although on the flip-side I suspect most discussions would be even more civil than we’re lucky enough to enjoy today.
The real downside would be fencing out non-regulars who bring one-off insights to a discussion. For example, a professional bond trader who arrives here via Google and educates us with a comment on an article about long-dated gilts.
That sort of thing is very valuable. I’m loathe to lose it. So for now the battle against spam continues!
Have a great weekend.
