Why at LUMINANCE we don't use AI for our content

We don’t use AI in our content generation because we believe that future algorithm changes are going to punish AI-generated text, images and video.

​

Here’s the thinking – AI models are prone to what has become known as “hallucinations”. Screw ups basically. Mistakes. They ingest all the info that they find and they spit out a wrong answer.

​

This could be for thousands of reasons, but one reason would be that the original information they learned on was wrong. Really? Incorrect information on the internet… yep, it’s hard to believe, but turns out you can’t trust everything you find online.

​

Now, the AI models are already trying their hardest to distinguish between the trustworthy sources and factually inaccurate ones, and I’m sure they’ll get better at it.

​

But a big red flag is AI models learning from AI generated content. It will introduce what statisticians call compound errors. In other words, an error in the original data, if starting measurements are not re-taken for future calculations, will propagate and compound through future cycles of the calculation (see Propagation of Uncertainty).

​

Let’s say there is an error, or a simple piece of misinformation, on a website. Language Model A takes that website as a source, plus a few other accurate ones, and churns out Answer X. X is then posted to a website. 

​

Language Model B then approaches a similar question and finds the inaccurate website plus Answer X, both offering this incorrect information. Language Model B therefore gives it higher priority. Answer Y is then posted and when Language Model C is asked to tackle the same question it pounces on the original website and Answers X and Y.

​

Now we have multiple sources offering incorrect information with increasing prominence. Misinformation triumphs, the internet goes down the toilet as a source of trusted information, and we might get a few lawsuits thrown in for good measure.

​

Therefore, our prediction is, and we’re fairly confident in some form of this becoming commonplace soon, that algorithms and AI search is going to punish AI generated content. They will be looking for user-generated content and anything which suggests that the content that they’re seeing is AI will get red flagged.

​

For the AI models to thrive, they MUST be able to discount AI-generated content.

​

So, our recommendation is to dodge the AI. Don’t go for those shiny short cuts. Put the hard yards in now, your content will be on a much, much firmer footing in the next few years.