Technically, GPT-3 is already here. Every writer – no, every human – will be impacted. But it’s full effects will be felt slowly over time. GPT-3 is a new, disruptive technology that will ultimately (after a generation or two) change the way information is interpreted online by everyone.
What is GPT-3?
In short, GPT-3 is a computer program that writes all new text content. The big difference is that GPT3 writes so incredibly well that 96% of the time it is impossible for a human to tell that a computer wrote the article they are reading. Or more accurately, if a human reads two articles (one written by GPT-3 and written by a human), when the human guesses which was written by a computer they guess right only 52% of the time. A “perfect” algorithm would have the human guess right 50% of the time.
For a bit of background: GPT-3 is the third version of an advanced machine learning algorithm that was trained on an unimaginably large data set. It’s objective is to simply guess the next word that should come in any sentence or paragraph or article. It does this one word at a time to create entirely new text content that no one has seen before.
How Will it Impact my Writing?
First, it is most important to know that it exists, that it is being used, and what it is capable of.
Second, it is important to begin learning to use it yourself. This is somewhat difficult, because the company who created it (OpenAI) decided to commercialize this version of the algorithm. It is not open source, and they are restrictive about who is allowed to use it. In fact, GPT-3 was abused almost instantly.
Third, if you wish to begin (or continue) to be a profitable writer, you should adjust the content you are delivering so that it does not compete directly with the kinds of content that can be generated en masse by GPT-3.
What Can GPT-3 Do Well?
As reported by the GPT-3 user who exploited it to create a hit blog will attest, GPT-3 excels at creative, open-ended, non-discrete concepts. If there is a measure of subjectivity, or if the topic being discussed can be addressed with a series of open-ended concepts, then GPT-3 does very, very well at fooling humans. Think of topics like:
- Health and Fitness
- Travel Tips
- Personal Development
- Creative Writing
- Science Fiction
- Relationship Advice
- Lifestyle Advice Religion Etc.
Be aware, that content fitting these niches (and others like them) will soon be trivially easy to produce at any scale.
What is GPT-3 Bad At?
Topics with specific, discrete answers, or closed-ended questions, is more likely to reveal to the reader that the content is being written by GPT-3. Any technical fields, scientific fields, mathematics, or other highly-niche sub-categories with known technical details will not be seriously threatened by GPT-3. Think of topics like:
- Software Development
- Architecture Design
- Medical Fields
- How-To Instructional Content
- Video Game Guides
- Mechanical / Engineering
It will still be possible for GPT-3 to create content on these subjects, but the content (if it is to be convincing) must necessarily be devoid of any specific technical details that the reader can easily understand and independently verify.
What Should We Do?
Individually, we can place a higher value on content we can personally verify was written by a human. Don’t believe that just because words are written (either in print or on screen) that a human put thought into those words. It could be auto-generated content.
Of course, you should have the same healthy skepticism already for photographs of people thanks to the deep-fake technology that is almost old news at this point.
The same real-time deep fake technology exists for videos as well. Movie-quality CGI alterations can (and have) been used in real time to fool viewers.
Place a higher value on in-person interactions, and on interpersonal communications where you can personally (in the flesh) validate the content that you saw on line. Be gracious to others who are not so well-informed as yourself, and be patient when explaining how easy it is to be manipulated by text, images, and media that may have no basis in reality.
Or just go full Amish and be done with it.