It is difficult to tell if a piece of content is generated by a human or a machine. To know if a piece of content is generated by a human, look for markers within the text itself, like typos, and grammatical oddities. Machines generate text that is usually mechanically perfect and lacks the imperfections of human-generated text. You can also look at the format of the content to determine if a human was involved, as humans are necessary for multimedia formats. If you want to communicate that a piece of content is human-generated, have a human involved in the delivery of the content.
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
- Take my new Generative AI course!
- Got a question for You Ask, I'll Answer? Submit it here!
- Subscribe to my weekly newsletter for more useful marketing tips.
- Subscribe to Inbox Insights, the Trust Insights newsletter for weekly fresh takes and data.
- Find older episodes of You Ask, I Answer on my YouTube channel.
- Need help with your company's data and analytics? Let me know!
- Join my free Slack group for marketers interested in analytics!
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.
In today’s episode, let’s talk about how do we know if a piece of content is generated by a human. It’s really difficult with large language models like the GPT-3, which is the underlying model that powers things like chat GPT, and so many other similar models that are being used in all sorts of content-based AI tools. One of the questions that you legitimately want an answer to sometimes is did a person actually make this? Or is this machine generated? Now for very low stakes stuff? You know, maybe it doesn’t matter. But certainly, if you’re a marketer, maybe you’re paying a writer to write blog content for you, you might want to know, am I paying for a human to write this or am I paying for a machine to write this? And there’s no good way right now to just look at a piece of text and go, “Okay, that came from a human or that came from a machine just on the text alone.” But what you should be looking for are little markers within the text itself, that can give you a hint as to its provenance. Machine-generated text, generally speaking, obeys the general rules of grammar and spelling pretty well actually. And in that respect, content is a lot like diamonds. When you have synthetic diamonds, the way that you determine that a diamond is synthetic is by looking at it through a variety of instruments and looking at the crystal and structure. What you will notice about a synthetic diamond that is different than a natural diamond is that synthetic diamonds are perfect. When you look at them, they are perfect; their crystal and structure is perfect. It’s flawless. When you look at real diamonds, there are flaws. Sometimes the flaws are big, like the diamond is the wrong color. Sometimes the flaws are small, like certain molecules in certain carbon molecules and atoms didn’t line up just right. And so there’s tiny little imperfections throughout it. But it tells you, “Yeah, this was naturally grown. This is not laboratory grown, not assembled quickly.” Content is the same way. When you look at the output of a lot of these large language models, they’re creating mechanically perfect content, good grammar, good spelling, no strange misuses, no weird turns of phrase. And so you can tell in text, “Yeah, this was probably generated by a human” when you spot those unusual constructs; for example, spelling errors, typos. Generally speaking, machines don’t produce typos in generated text, grammatical oddities. For example, you might say, “I got that,” right? Somebody asked, “Hey, what about the set headphones?” I got that. Grammatically, it’d be more correct to say, “I have that. I have that model. I got that model. I have that model” is the grammatically correct way to do it. And so those little quirks of grammar that are unique to you differentiate your text from machine-generated text.
I’ve been doing a whole bunch of writing for my book, Beyond the Velvet Rope, and I what I’ve been doing is I’ve been doing a lot of voice recording and then taking the ugly transcripts and washing them through one of the API’s to clean up the language. And what it spits out is mechanically correct, kind of bland text; it loses a little bit of flavor. And so when I take those summaries, I then blow them back out and and retype an awful lot of them and reintroduce the unique quirks, the imperfections that are my personal voice, the way I speak, the way I write, the way that I use language is unique to me. And unless you are someone who can personally train and fine tune AI models, chances are you’re probably not going to be doing that with one of these large language models to make it sound exactly like you said, you’ll just kind of use the mechanically generated texts. That’s how you know a piece of text is human.
The other thing that you could look for is format, right? Multimedia formats. This piece right here that we’re talking about: is this machine generated? Well, no, clearly not. I’m sitting here talking to you. Right. Do you hear the sound of my voice? You can see me on the screen. Even if I read it off of a script, but I’m not. There’s still a human involved, right? You can still tell, like, this is me. This is I. Am not yet computer generated, maybe someday. But for right now, that’s not the case. That’s the easiest and most prominent indicator that you know a piece of text is human is there’s a human involved right here: I am.
If your customers, if your employees, whoever, if people want that authenticity, they want that certified organic, human generated content, the easiest way to prove it to people is have a human in the mix, right? Have a human delivering the content, even if a machine wrote it. So that okay, well, yeah, there’s, there’s a real person there. Because in a lot of ways, the concern that people are expressing is just a fear, in some ways, that we have been replaced. So if you can show yourself or someone who works at your company, you show a real person behind that. It communicates consciously and unconsciously to somebody else, “Oh, yeah, there’s a person in there. There’s still people behind the wheel somehow within the content that you’re delivering.”
So as you start evolving your content offerings to use artificial intelligence to use generative AI, you may want to deliver that in as human a way as possible. And that means turning on the camera, turning on the microphone, and having a human being be part of the delivery.
Thanks for tuning in. We’ll talk to you soon. If you liked this video, go ahead and hit that subscribe button.
You might also enjoy:
- Almost Timely News, December 10, 2023: Where Generative AI and Language Models are Probably Going in 2024
- What's the Difference Between Social Media and New Media?
- Almost Timely News, December 24, 2023: Why Mistral's Mixture of Experts is Such a Big Deal
- Can Causation Exist Without Correlation? Yes!
- Almost Timely News, December 17, 2023: Improving the Performance of Generative AI Prompts
Want to read more like this from Christopher Penn? Get updates here:
Get your copy of AI For Marketers