Anthropic recently announced they will be finger printing generated text so it will be easier to determine if the content you are reading has been produced by AI or a human. This is not like binary content (images/video) where a digital watermark is embedded in the content, this one builds off the underlying way generative AI works.

I have noted many times, generative AI is not “intelligence”. At its heart it is a sophisticated random number generator. It uses all the patterns it has been trained on to guess what the most likely next word is going to be. It is auto-suggest on steroids if you want to boil it right down.

So, given it is guessing the next word, Anthropic have figured out a way to reverse engineer that, to look at a piece of text, and ascertain that it was most likely generated by them. While it is not foolproof, it has put those that enjoy BOVINE (Being Overly Verbose In Needless Expositions) output on notice.

Now the question becomes, if known for sure AI produced the content that you are reading (which can be extended to the source code as well) how does that make you feel about the person passing it off as theirs?

• Multi paragraphed email from your boss, knowing they didn’t write any of it
• 20-page PowerPoint deck from the consultant firm, yet still invoicing you that eye-watering amount just to tell you to either raise prices, or sell more
• Source code for the enterprise platform that was 90% vibed.
• Legal contract from your lawyer, that you hope to be watertight, justifying their hourly rate

The sentiment against AI generated content, has gone beyond its novelty stage and now entered its resentment phase. LinkedIn recently added their “Seems like AI slop” action on each post in a hope to get back to human generated content, after finding over 40% of long form posts are AI generated.

The European Union is forcing the hand of the AI companies to be more transparent in what is AI vs. Human generated content. Advertising networks have woken up, asking advertisers to declare the advert is AI generated. The public sentiment for AI is turning and the governing bodies are responding.

Now, the cynic in me, wonders if there is an upside here for the AI companies. What could they possibly gain from this new level of “transparency”?

I have a theory.

The amount of AI content on the likes of LinkedIn, Reddit, Yahoo, etc. has surpassed human content. Which means, the AI companies are literally running out of quality content to train on. They are training on themselves, or as Kevin Kelly (founding Editor of Wired) put it in his 1994 “Out of Control” book “mentally masturbating”. What this means it that no new information is coming into the ecosystem, to disturb/challenge/stimulate any sort of knowledge or thinking.

They need to flag AI content, and open these standards to the rest of the world, to get us all back to writing raw content by hand again.

While, at first blush, it is an compliance move, I question their real motives behind it, as they are realizing their models are no longer getting the leaps in advancement between each release, and people are quite content with the cheaper slower models.

This has put on notice consultants, lazy journalists, social media creators and every C-level suffering AI psychosis that their BOVINE is about to be called out publicly. I think this is a wonderful step forward for ethical AI.

If you can’t be bothered taking the time to compose and write the content, then do not expect the reader to take the time to read it.

AI Disclaimer: Gemini Nano Banana Pro was used to generate the photo – from the 2017 “The Man who Invented Christmas“ movie.