Beware the Great AI “Melting Pot”?

I’ll ask your indulgence as I offer my own take on what seems to be a seemingly endless series of articles talking about AI and whether it is going to lift humanity up to new heights or doom us to oblivion. Fortunately, while I suspect the reality will not find us at either extreme, what prompted me to draft this post is something somewhat more mundane but not a topic I’ve heard spoken about very much if at all.

As AI models become more sophisticated and more people and businesses use them, I believe there’s a real potential for a sort of “averaging out” of content. By this, I mean that as more and more people use AI to generate content, there’s a risk that things will start to sound the same. This is because the AI model is essentially a learning algorithm that processes huge amounts of data in order to generate new content – the old is, effectively, recycled to create the new.

So if the data it’s processing is very similar, whether that be in terms of style, tone, or even language, then AI-generated content is likely going to reflect that sameness. This could lead to a kind of homogenization of the written word, where everything begins to sound like it is coming from the same source, rather than reflecting the unique voices and perspectives of the people creating it. This could be an issue for a number of reasons.

First, it could lead to a lack of diversity in the types of ideas and perspectives being expressed. If everything starts to sound the same, then we’ll be missing out on the richness of different voices and perspectives. This could be particularly troubling in areas such as journalism or opinion writing, where having a range of voices is critical to making sure that all sides of an issue are being represented.

Second, this “homogenization effect” could make it harder to stand out. As an example, look at businesses that rely on content marketing. If all the content starts to sound the same, then it will become harder for companies to differentiate themselves from one another. Ironically, 100% human-generated content may end up being seen as a premium service offering in the future, or maybe there’s an on-going role to having a human editor or writer come in to add some flair or personality to help keep things fresh and interesting.

One thing I haven’t experimented with, but which could be a technology-based solution, would be to instruct your AI model to emulate a specific style or approach. For example, if you want funny content, you could train your AI model using comedy writing as its input; likewise, you could challenge your AI assistant to draft something in the style of someone well-known (e.g., “draft a review of this movie in the style of Mark Twain”). That doesn’t completely solve the averaging of content issue I raise here, but it may permit us to flex things enough to keep the boundaries wider than they may otherwise have been.

Ultimately, the homogenization effect of AI-generated content is something that we need to be aware of and actively be working to address. By continuing to rely on unique human perspectives and by instructing our AI models to emulate specific styles, we can help to ensure that the written word remains vibrant and diverse while taking advantage of the advantages this emerging technology offers.

Or we can just promise to write everything ourselves ;).

Know that I’m pulling for you!

PS. For fun, I used an AI text-to-image tool and requested an image of what it thought an AI melting pot might look like. This didn’t quite nail the concept but it was still pretty cool looking for an article header.