It Started With Pen & Paper. You Won’t Believe What Happens Next.
Writer sits down to finally conquer writer’s block. Then his laptop started talking to him.
Pen and paper.
Keyboard and code.
All tools for writers.
And these tools have never taken any writer jobs, either.
Instead, tools enable the next level of creation.
For writers (and, really, for anyone doing creative work), AI can open up your next level of creation.
For any kind of writing.
With the help of AI (and GPT-3 specifically), you can write poems, short stories, novels, limericks, scholarly articles, blog posts, emails, ads—anything.
In Bionic Writer, you’ll discover what happens with humans and tech meet to write.
What’s the most interesting tool to come in a long time?
Here’s a quick primer:
What is GPT-3 and how does it work?
It consumed all of the internet.
Crunched the numbers.
And figured out how to predict what word should come next.
GPT-3 has read all of Wikipedia, all of Google Books, every month of the Common Crawl dataset from 2016 to October 2019, comprising every piece of writing on the web during that time.
It has read 45 terabytes of text.
This equals a book about 3,375,000,000 pages long.
It has analyzed billions of lines of text, figured out what matters and what doesn’t in language, and then uses that information to write more text.
From everything it consumed, all it cares about is producing the next word that should come after whatever word it just read (or wrote).
That is, it produces different probabilities for different words, and then picks the one with the highest likelihood of being the accurate word.
If I prompt GPT-3 with “Mary had a little…”, it will guess “lamb” and insert it.
Depending on what else I told it, I could also get something like:
You give GPT-3 a prompt—a piece of writing—and then it tries to predict what would come after that, based on all the stuff it has read and the rules it has learned.
It’s like it writes from memory.
Humans do things a little differently.
We build abstract mental models. As we write, we synthesize thoughts and knowledge.
We have an idea or construct of what we want to write formed in our minds. We then write those out in words, sentences, and paragraphs.
At some point, we edit, rewrite, and edit some more.
We use planning, reasoning, problem-solving and other cognitive processes as we write.
GPT-3 doesn’t do that.
But how much does that matter?
Maybe not a lot.
Researchers have given test subjects 200-word news articles written by either humans or GPT-3 and asked them to state which one had written them. On average, they chose right only 52% of the time, which is basically just guessing right.
In some instances, it’s obvious that an AI wrote something.
Other times, you can’t tell.
But we’re not concerned about being able to tell human words vs. robot words apart.
What matters is how GPT-3 can help you write—better.
GPT-3 is very good at writing things that are short and creative. Task it with something long and factual, and it struggles.
So, from now on, think of it this way:
How can GPT-3 help me improve your writing, thinking, and ultimately—your creativity?
In other words:
How do you get the most out of AI as a writer?
You give it a great “prompt”.
The more you put into GPT-3, the more you get.
GPT-3 writes better if you give it examples of what you want.
It takes the context and the patterns in the prompt and uses them to narrow down its model of language to things that match.
Then, it generates text that fits the context and pattern.
Your prompt is critical. The more information you give the model in the prompt, the better the output will be. If you give it the context and three-to-five examples of what you want, you’ll get a better output than if you give it little information.
For writers, of any kind of text, what you’re seeing is a blend between:
Software developers (who write ‘prompts’ in the form of code—very specific ‘prompts’ that tell software what to do).
And…
Creatives (who conjure up ‘prompts’ in their mind, work them out in real life, and create a piece of art).
Putting it simply, we’re all dealing with ‘prompts’ one way or another.
Prompt craft is how you want to think about this.
How can you give AI and your imagination better direction with prompts?
Now, when I sit down to write anything, I start with a vague prompt and then through my writing and use of GPT-3, it develops into a clearer prompt.
In turn, my imagination feeds it back into GPT-3, and out comes text.
I got some amazing links for you below.
Plus what tools are best.
Next time: Watch over my shoulder as GPT-3 writes me good stuff and make me gooder writer.
“So Sam, which tool is best?”
There’s no clear answer to this.
Because it all depends on your use case.
There are over 100+ tools available. And more are coming.
Most of the time, I’ll tell you to just use GPT-3 from OpenAI.
Why?
Almost all AI writing tools are layers on top of GPT-3.
(This includes popular apps like Jarvis, Copy.ai, Anyword, and many more).
These tools tap into the vast ocean of words that GPT-3 can generate and fine-tune what you get back to suit your needs.
Writing fiction and stories? Sudowrite is better for this (and not, for example, for writing Facebook ads).
Writing ads, emails, and other copy assets? Anyword can help with that.
Writing blog posts and other content market assets? Copy.ai can help with that.
Want to use GPT-3 for SEO? Keyword Insights AI can help.
And on it goes.
There’s overlap between uses for all of these tools.
But for the most part, tools are fairly specific about use cases and will tailor what you get along those lines.
In upcoming weekly issues, I will walk through several tools that people seem to love and show you what’s possible.
For now, just remember that almost all apps use GPT-3 for text generation.
Here’s a list of about 41 tools (no affiliate links, just honest opinion).
Talk again soon,
Samuel Woods
The Bionic Writer