Vocal has a huge AI problem but it's not what you think

Yesterday, Vocal's Curation Team posted this story where they say:

content generated by AI can never make anything new. It’s all a kind of plagiarism — if we’re being blunt

They say they run 'every challenge entry' through an AI content detector, and they disqualify anything 'generated'.

I'm happy to see this finally confirmed.

Vocal's article doesn't beat around the bush. They're frank, and they cut to the chase.

We know this because they tell us:

Unfortunately, they are very much beating around the bush. Nothing in this story addresses the core problem.

Saying you're being frank, is not the same as actually being frank. This empty rhetoric doesn't go unnoticed, Vocal.

So wait, what's the problem here..?

They say that every 'challenge entry' gets checked for AI content.

Most challenges get a few hundred entries. Those are small enough numbers that some poor Curation Team intern can manually upload the text of the entries into an online AI checker.

This is the absolute bare minimum I would expect, given that challenge winners get real money. I would also expect Top Stories and Leaderboard winners to be checked in this way. And I believe that they are.

No, 'challenge entries' aren't the problem. This is just how the Vocal Curation Team are trying to misdirect us from the actual problem

The problem is everything else.

The fact is that 92% of Vocal's Latest Stories are AI content or spam.

Believe me, I checked:

Vocal doesn't check any Latest Stories for AI or spam.

When we submit a story, Vocal reviews it. The specifics of the review process are opaque, but evidence shows that it doesn't check for word count (e.g. this post of only 61 words despite a minimum word count), or plagiarism (literally a Wikipedia article, literally a text book including the contents page), or spam (you need land clearance in Kentucky? AGMHunting are your guys!).

It might check for profanity, but really, who the fuck fucking knows?

These days, most stories are approved within minutes. Which is great. Except this suggests that no human is moderating them.

Which is strange because Vocal's own FAQ says that "all submissions are reviewed by our moderators":

I don't think this is true. A human moderator could spot AI or spam in seconds. But human moderation is costly, so it's understandable why Vocal probably doesn't use it on all submissions, despite what they say.

Are they just straight up lying to us..?

But here's the thing:

If Vocal really want to tackle AI, they need to use AI.

They need a tool which can check stories automatically.

Vocal have considered tools like this. But, as Justin Maury, Vocal CEO, said in his Spring Update:

they are very expensive at scale and because they are nascent, they produce too many false positives

Hmm. Let's beat this bush a little longer...

CopyLeaks is one of the biggest AI checkers online, doing over 30 million scans per month. (I'm not a shill, I'm just parroting their website.)

The public price to use CopyLeaks API for AI and plagiarism detection is $13.99 for 1300 scans of texts up to 250 words. That's about $0.01 a scan.

Vocal lists roughly 100 new stories a day in Latest Stories. So 100 scans would cost about $1 per day.

Is $1 a day really too expensive for Vocal to solve the single biggest thing that pisses off its users – namely the tsunami of AI spam in Latest Stories?

My figures are rough, and don't account for bulk costs, or indeed bulk discounts. I'm sure Vocal receives exponentially more submissions than what we see in Latest Stories, and scanning masses of auto-submitted AI content would be futile and expensive.

So why doesn't Vocal check just these 100 Latest Stories?

This is the pipeline through which all content gets on the site, so why not plug this leak?

Regarding false positives...

If Vocal implements automatic AI detection, there's lot of ways false positives could be handled.

If a story gets identified as AI-generated and isn't marked as such by the author, it can go back in the author's Drafts, like any story that fails review.

If the story is falsely flagged and the author has used AI as a legitimate writing assistant, they can contact Vocal support and argue their case. A human moderator can step in at this stage, and immediately confirm whether the content is AI or spam.

This happened to me when a story I wrote in the form of product reviews got flagged as spam. I pushed back. It was cleared. No problem.

However if the story is correctly flagged as AI content and it was written by a bot or a spam account, it's unlikely a bot or spam account would spend time arguing the decision.

The vast majority of AI spam is this kind of content – 100% AI-generated with no human editorial involvement and not declared as such. This is not the same as stories written with the assistance of AI.

It is this 100% AI-generated content which I believe must be fully eliminated from the site. Not just flagged or marked. Deleted.

But Vocal don't want to automatically flag or remove fully AI-generated content, because it would lay bare the vast scale of the problem and it would be impossible to hide.

And it's not just individual stories. It's whole accounts.

There are lots of AI and spam accounts posting multiple stories an hour. It seems reasonable that free accounts should not be allowed to do this.

Many of these accounts are brand new. So again it seems reasonable that new accounts should be limited to one post per day. Perhaps only Vocal+ subscribers should be allowed to post unlimited stories.

I'm sure Vocal will say they want to encourage people to tell their stories, and don't want nasty capitalist money to thwart those writers' ambitions. But in reality, they're letting the AI thwart them instead.

There are dozens of quick, cheap and smart ideas which could help reduce the problem. But unfortunately those are not the solutions that Vocal is embracing.

Because Vocal don't want quality. They want churn.

Constant turnover of new content improves their SEO. Vocal isn't funded by advertising, but they get money from investors and from Vocal+ subscribers.

Their 2023 financials show a significant tightening of their belt:

I'm no accountant, but it seems to me like Vocal turned a profit despite stable revenue, by massively reducing costs more than 40%. Their budget strategy, it seems, is to cut corners.

At the same time, Vocal has good site traffic and user engagement metrics. These could persuade investors that Vocal's strategy is working. But investors don't care if the content is sewage.

OK, I might as well 'cut to the chase'...

Is there less AI content since the spring update?

I see no evidence of this.

Are AI spam authors voluntarily flagging their own AI spam?

No. Duh.

Are the AI stories we report being removed?

No. No. No. No. No. No. No. No. No. [Correct at 17 July 2024, 0700 PST]

Re-checked on 22 July, 1300h PST:

Yes. Still no. Still no. Still no. Still no. Still no. Still no. Yes. Yes.

So, not only are Vocal asking us to report AI content to them, they're not taking any action when we do.

Have any of Vocal's actions been effective in reducing the AI problem?

I see no evidence of this.

Instead, Vocal are masking their ineffectiveness with this dishonest, duck-and-weave post filled with empty, patronising rhetoric.

If Vocal insist on sticking their very human brains in the sand, putting out ineffective updates, and posting self-justifying articles, the AI problem will never be solved.

And – to be frank – what's the point of being part of a Vocal community which is 90% robots?

Thanks for reading!

Mostly I write horror and transgressive fiction, but lately I've been venting about online publishing startups and their utter unpreparedness against the AI invasion 😫

Regardless, please buy my eBooks, including the darkly hilarious horror story HEAD CASE and the outrageous feminist splatterpunk METAGOTH, featuring goth antiheroine Rosa Razor.

I really the need the money for aspirin.

Get them from GODLESS (PDF and EPUB) or Amazon for Kindle.

Please buy, share and review – your support means everything!