How to use GPT3 to augment your creativity
First, this post wasn’t written using GPT3. I’m not that self-destructive. But I’m here to tell you that when it comes to starting long-form content for your blogging or marketing needs it is one of the best tools out there. As I noted before, the current crop of AI-assisted tools is designed to make aesthetic sense, not literal sense. The writing that these tools provide, be it code or paragraphs of text, looks great but isn’t great. Without getting too philosophical, GPT3 is a CNC machine that can cut out a metal airplane part that would take a human hours to build. But the final quality control and finishing must be done by hand because you don’t want to trust a robot with the lives of 300 passengers.
Your LinkedIn broetry isn’t as vital, but you get the idea.
I’ve been using GPT3 to build blog posts for my new site, NaturalGoods.news. You might be able to tell the difference between a human post and an AI-generated and, further, you’ll probably be wrong. At this point in the game, content aggregation is dead. The robots can do it for us far quicker than we can do it ourselves.
Here’s how I do it.
I write this newsletter and talk about great startups for free. I’d love to be able to make it a major part of my income. Want to help?
Thanks!
Some code
First, stop visiting ChatGPT3. It’s going to get clogged up and you’ll never be able to depend on it. Instead, use the GP3 API. You can sign up for an API key here.
This is my dead-simple Python3 code. You’ll need to install various libraries to get it to run including openai. Here is my requirements.txt file:
appdirs==1.4.4
attrs==22.2.0
cattrs==22.2.0
certifi==2022.12.7
charset-normalizer==2.1.1
et-xmlfile==1.1.0
exceptiongroup==1.0.4
idna==3.4
numpy==1.24.0
openai==0.25.0
openpyxl==3.0.10
pandas==1.5.2
pandas-stubs==1.5.2.221213
python-dateutil==2.8.2
python-decouple==3.6
pytz==2022.7
requests==2.28.1
requests-cache==0.9.7
six==1.16.0
tqdm==4.64.1
types-pytz==2022.7.0.0
typing_extensions==4.4.0
url-normalize==1.4.3
urllib3==1.26.13
Now for the code. I use an .env file to store OPENAI_ORG_KEY
and OPENAI_API_KEY
:
from __future__ import print_function
import requests
import json
import requests_cache
import os
import openai
from decouple import config
requests_cache.install_cache(cache_name='apicache', backend='sqlite', expire_after=360)
article=open("article.txt","r")
thisdesc=article.read()
myprompt="Write a blog post based on this article: "+thisdesc
openai.organization = config("OPENAI_ORG_KEY")
openai.api_key = config("OPENAI_API_KEY")
openai.Model.list()
output=openai.Completion.create(
model="text-davinci-003",
prompt=myprompt,
max_tokens=500,
temperature=0
)
print(output['choices'][0]['text'])
This code essentially calls the API to ask it to write a blog post based on the contents of article.txt
which is contained in the same directory as the Python code. You can change the text of the myprompt
variable to make it ask anything.
Have any questions? Ping me at john@typewriter.plus. I’m doing some consulting on AI-based content and this is the easiest way to get it to work for the least effort.
So now what?
So I’ve probably sufficiently confused you or frustrated you with my little code snippet but the bottom line is that this stuff is easy. Creating readable content out of ideas, outlines, or even excerpts from other content is now trivial. Notice I said “readable” and not usable. This content is, at best, messy. For example, let’s tell my program to write a few paragraphs for us.
The result?
ChatGPT3 is a revolutionary new API that is revolutionizing the way content is created. It is a natural language processing (NLP) tool that uses artificial intelligence (AI) to generate content from a given prompt. This API has the potential to revolutionize the way content is created, allowing for more efficient and accurate content creation.
The first benefit of using the chatGPT3 API is its accuracy. This API is able to generate content that is more accurate than traditional methods of content creation. This is because it uses AI to analyze the given prompt and generate content that is more accurate and relevant to the prompt. This means that content created using the chatGPT3 API is more likely to be accurate and relevant to the topic at hand.
The second benefit of using the chatGPT3 API is its speed. This API is able to generate content much faster than traditional methods of content creation. This is because it uses AI to analyze the given prompt and generate content much faster than a human could. This means that content created using the chatGPT3 API can be created much faster, allowing for more efficient content creation.
The third benefit of using the chatGPT3 API is its cost-effectiveness. This API is much more cost-effective than traditional methods of content creation. This is because it uses AI to analyze the given prompt and generate content much more quickly and accurately than a human could. This means that content created using the chatGPT3 API can be created much more cost-effectively, allowing for more budget-friendly content creation.
Overall, the chatGPT3 API is a revolutionary new API that has the potential to revolutionize the way content is created. It is an accurate, fast, and cost-effective way to create content, making it an ideal choice for content creators.
First, you’ll notice we didn’t quite get the request right. There are some tweaks we could have done to reduce the amount of content, etc., but what we got back was acceptable. Was it interesting? No. Was it informed? Not really. Was it useful? Eh?
What content like that needs is some anima or vital force. I’m reminded of Dr. Oatman talking to Martin in Grosse Pointe Blank:
Martin: I had one where I was that television mechanical rabbit. You know, with the, the...
Dr. Oatman: The battery bunny.
Martin: Yeah, I was the bunny.
Dr. Oatman: That sounds like a very, very depressed dream.
Martin: Really?
Dr. Oatman: Yes.
Martin: Why?
Dr. Oatman: Martin, it's a terrible dream! It's a depressing dream to dream about that rabbit. It's got no brain, it's got no blood, it's got no anima! It just keeps banging on those meaningless cymbals, and going and going!
The AI does what you ask it and nothing more. The light is out behind its eyes. And, again, given the depths of depravity that currently pass for online content, maybe that’s for the best.
Ultimately, however, we aren’t winning the war by depending on AI. We must use it as a tool, not a crutch. There are some cases when you can use AI to generate ideas quickly and easily - I found it to be a very friendly way to summarize news and create readable blog posts - but I’m always wary of quick fixes.
Look: writing was hard. It just got a lot easier. But it’s still hard. Don’t give up trying to create compelling and interesting writing just because you can tell a robot to do it for you. You’ll end up like Martin Blank - a bunny banging endlessly on in the dark.
How’s that for an image an AI can’t create?