A member of our community posted something in Slack I haven't been able to stop thinking about. He's a content marketer and an incredible writer currently working on a project for a client. He got this message from a client:
How do I know you didn't use ChatGPT to write this?
He didn't. But it's a funny question— because ChatGPT is a plagiarism machine. So the question that client asked is:
Did you use the plagiarism tool that plagiarized the entire internet, to, I guess, plagiarize?
Oww, my head.
That's what language models do— they munch on huge data sets often stolen from the internet, content creators, and authors in order to make mathematical predictions about language. Basically, it’s using math to predict what the next most likely word in a sentence is.
AI Language Models don't create anything, they predict.
And it can only predict because it ate up things you created, and things I created, and every scrap of content created by humans in the last hundred years.
And even with all of that content, it's still hungry.
But, here's the problem: AI can't cook.
AI can only eat if we create.
It's only as good as the data it consumes and learns from. AI that's trained on books is much better at predicting text than AI trained on social media posts, *gestures to the internet* for obvious reasons. And authors are getting wise to the fact that their copyrighted work is being used without their consent and suing.
Fanfiction writers are also getting wise to the fact that these models are trained on their work— and are organizing to create mass amounts of content to troll the AIs stealing their work. It's really friggen great.
The "synthetic content" AI generates is often unreliable. It lies. It contains racial and cultural bias. And now that content is everywhere, so when it scrapes the web to learn— it's going to start eating itself. And when it eats its own content it gets sick. Or what researchers call Model Collapse. The outputs become more homogenous, inaccurate, and lose their value.
The biggest risk to AI isn't government regulation or sentience (knocks wood)— its biggest threat is us. AI language models will fail if they can't steal from our creativity.
Despite what reactionary headlines tell us, human-created content is where the real value is. And it's going to become more scarce.
Now, something being valuable and something being compensated for aren't the same thing.
We all need to fully understand how these tools work— and that our writing is being scraped and used to teach them. That's why these copyright lawsuits are so important. It's why the Writers Strike is so important.
It's time for creators to get wise, and fight for our content.
Because if these plagiarism machines can't eat our work, they'll starve.
If you found this useful, please be a helpful work friend and send it to a colleague, or share it on Slack. Did a friend forward this newsletter to you? You can subscribe here.
Expert-led workshops on copywriting, SEO and everything in-between. Hands-on seminars to build a career you love. Skills and templates you can put to work right away while connecting with excellent humans. Oh, and sometimes we make scones.
Are you hiring? Tell us who you're looking for and we'll help spread the word.
Are you looking for your next role? Let us help you land your dream job.
💼 A tool to help you with interview prep.
🙌 Soon you’ll finally be able to download your TikTok videos without the watermark.
🙏 Something to help you keep the social media algorithm gods happy.
🔎 Five steps to help you determine how many backlinks you need to rank.
🔮 Click this mystery link for a surprise.
Got something to share? We want to hear from you!
Share your links with us.