The AI content flood

AI copywriting tools have been around for a few years, but everything changed with the launch of Open AI’s ChatGPT. Like many of you, I was stunned at its well-written responses across a wide range of topics. This isn’t just a writing enhancer — it’s an AI that can write. Use cases are still exploding in my head.

Journalists are notoriously skeptical of AI. You’re right, ChatGPT is not going to replace original journalism or smart analysis (at least for now), and it still “hallucinates” mistakes. But it’s certainly good enough to put to work as a research assistant. For example, feed it a long interview transcript and ask for a bullet-point summary and suggested quotes. That’ll take about 10 seconds.

While ChatGPT won’t win any journalism awards (at least for now), it can certainly automate much of the long tail of content on the internet. Here are just three examples: automatically post ChatGPT-powered summaries of news stories moments after they’re published by major news brands, divided by region. Or pick a topic like cooking, and use ChatGPT to populate a recipe site from scratch. Or pick a city, and have the AI argue in favor or against a political position in context of the local community.

Now multiply this across a million niches — content farms running in hyperdrive at a fraction of the cost. As one AI researcher said when working with an earlier version of Chat GPT, it’s like flooding the internet with “10,000 Wikipedias.”

Copyright and plagiarism are certainly issues that come to mind, and so are propaganda and disinformation. Imagine using ChatGPT to spam human-like comments and social media posts that support an agenda. Or a war. Thinking of running for office? Time to mobilize the AI content army.

Since the AI can write in a variety of styles and lengths, its content is virtually indistinguishable from an average human writer. That’s why OpenAI is working on creating a digital watermark embedded in ChatGPT responses. But even if it’s successful, there are other AI services and plenty of loopholes. For example, many students know that “paraphrasers” are a handy way to change the wording and voice of Wikipedia articles to avoid detection.

If you’re still skeptical, just look at the pace of change. OpenAI is expected to release a significant upgrade of ChatGPT in the next several months. Other companies aren’t too far behind. The more data it consumes and the more “parameters” of the data it learns in training, the better it gets. Humans will write less, AIs will write more. For good and bad, AI-written content will flood the internet, an amalgamation of the work of millions of human writers and journalists who came before it.

Cory Bergman is a co-founder of Factal.

Leave a Reply