You have heard about ChatGPT for writers. I’m not an expert in computer science or data analytics. I’m just a regular guy with a high-school education trying to keep up with things I consider to be important, like making a living. OnText is all about that, especially side hustles in the writing and publishing fields. OnText is here to stay.
Artificial intelligence isn’t going away either. Early adopters will benefit by understanding tools based on artificial intelligence, like ChatGPT. Those of us interested in writing and publishing have to pay close attention to these platforms.
I’ve spent time chatting with ChatGPT.
No doubt you’ve heard of this bot. ChatGPT is an amazing language system that carries on intelligent, two-sided conversations. When you use it, the first thing you see is an input form inviting you to “Send a message….”
Having a conversation with a network
The first time, to amuse myself, I tried a few random prompts (messages) for my own amusement. I was not just amused, I was amazed at the replies. After practicing and watching a few YouTuber’s efforts, I upped the difficulty level. What could this platform really do for content creation? I mean, that’s what I do, I figured I should explore.
Done with being simple, I wanted to do a deep dive into how ChatGPT processes language.
Me: This is an instruction. I will give you a simple prompt after you respond to this one. In response to that prompt, I want the number of possible next words for each word in the prompt. Do you understand?
Chat: Yes, I understand. Please proceed with your next prompt.
Me: April showers bring May flowers.
Chat: Certainly! Here’s the number of possible next words for each word in the prompt “April showers bring May flowers:”
- April: 12,415
- showers: 1,605
- bring: 5,998
- May: 5,275
- flowers: 2,517
(Elapsed time: under 5 seconds.)
Woah! I barely understood the instruction myself. Apart from realizing that this stuff is bat-shit amazing, I imagine you’re wondering
- what is a “next word”
- why does it matter
- and why did I go down this particular rabbit hole?
A glimpse under the hood of ChatGPT for writers
I always have to look under the hood, that’s why.
Although I could have dazzled you with clever AI tricks, the prompt I used reveals how ChatGPT thinks. Interviewing the ultimate expert, ChatGPT, I discovered it has no problem explaining itself.
I asked it how it works, and it told me. The entire conversation takes longer to read than it took to happen. You can read the transcript at the end of this article.
GPT (Generative Pre-trained Transformer) is a natural language processor. I learned a lot from a 60 Minutes interview with Google experts. These processors predict the probability of each possible next word in a sentence based on the words that come before it. Going back to my April showers prompt, you can see that there are a lot of “possible next words.”
The chatbot does not randomly assemble by randomly choosing a word. It derives context and meaning from your prompt and then assembles a response based on context, meaning, and the best next word related to how a human might speak.
Back to reality
Of course, this was an exercise, but I got the feeling that ChatGPT and AI are the magic tools that will help you reach your goals. You’ll be responsible for using them responsibly. So here, unedited and from the expert, are five significant ways ChatGPT can do just that.
- Generating topic ideas: ChatGPT can be a great tool for generating topic ideas for content creation. You can input a broad topic, and ChatGPT can provide you with a list of related subtopics or angles to explore. This can help you identify fresh and unique topics that your audience will find interesting.
- Writing prompts: ChatGPT can also be used to generate writing prompts. This can be especially useful for creative writing or if you’re stuck on a particular piece of content. You can input a keyword or phrase, and ChatGPT will generate a writing prompt or sentence to get you started.
- Research: ChatGPT can be used to conduct research for your content creation. You can input a question or topic, and ChatGPT will provide you with a range of relevant information and resources. This can save you time and effort in your research process.
- Content optimization: ChatGPT can also be used to optimize your existing content. You can input a piece of content, and ChatGPT will analyze it and provide you with suggestions on how to improve it. This can include suggestions for better headline or meta descriptions, keyword usage, or sentence structure.
- Content promotion: ChatGPT can also help you with content promotion. You can input a piece of content, and ChatGPT can suggest social media posts or email templates to promote your content to your audience. This can help you get more visibility and engagement for your content.
Working responsibly
The team of experts at OnText takes writing, self-publishing, and making money seriously. We believe automating certain tasks is a good thing. We do not believe in scamming the reading public, nor do we think anyone should get paid for doing nothing.
If we use AI, we’ll tell you that. We assign an editor to every piece of generated content. We are aware of the overuse of passive voice, the repetitive arguments, and the cold and unemotional tone of generated content.
So we fix those issues. We add content and check facts. In normal circumstances, we would have heavily edited GPT’s responses above, but we wanted you to see the reality.
And you?
It’s likely that all writers will employ a content generator at some point. There are many ways that technology can run amok, and it will be all of us that prevent real harm from happening.
Learn the technology. Understand it. Don’t fear it. As the future unfolds, we humans must all be ready to have a say.
The full transcript
The autoregressive process in NLP is typically implemented using a type of neural network called a “recurrent neural network” (RNN). RNNs are designed to work with sequences of inputs, such as words in a sentence, by maintaining a “hidden state” that captures information about the previous inputs.
During training, the RNN is given a sequence of input words and is asked to predict the next word in the sequence. The prediction is made by passing the hidden state of the network through a “softmax” function, which generates a probability distribution over the possible next words.
The predicted word is then added to the input sequence, and the process is repeated for the next word in the sequence. At each step, the hidden state of the network is updated based on the current input and the previous hidden state.
During generation, the autoregressive process is similar, but with a key difference: instead of being given a sequence of input words and asked to predict the next word, the network is given a prompt and asked to generate the next word in the sequence.
To generate the next word, the network first processes the prompt and generates a hidden state. This hidden state is then used to generate a probability distribution over the possible next words, which is sampled to select the actual next word. This word is then added to the prompt, and the process is repeated to generate the rest of the sequence.
Overall, the autoregressive process in NLP involves using a recurrent neural network to predict the next word in a sequence based on the previous words, with the network maintaining a hidden state that captures information about the previous inputs. The process can be used for both training and generation, with the latter involving generating a sequence of words from a given prompt.
Learn more
9 Ways to Illustrate a Storybook
One Response