Like most of you, I’ve been using (although probably more aptly called “playing with”) Chat GPT and DALLE-2. You may have noticed that for the last few months, the images on this blog have been created by DALLE-2. In the past, I’d be using Pexels or Pixabay to look for images that might align with the topics I cover on this blog, so most of the time, after a lot of searching, I might find an image that is pretty close to the topic. But more often than not, I don’t.
I’m more of a writer. In many cases, I consider the image superfluous to the writing. Still, in this hypervisual age that we are living in, even writing must be accompanied by some compelling image.
Case in point: I wrote a post I published on LinkedIn called “I Fell Asleep In A Meeting” about a time when I met at a client’s site. We were trapped in a small stuffy, hot room in the basement of their Silicon Valley campus, listening to the client drone on and on about their issues. I didn’t need to be there, but our CEO at the time thought it would be a good idea that I attend. I nearly fell asleep in my chair, so I thought I’d write about the experience. I found an image of an office worker lying on a carpeted office floor face first to accompany the post.
It was the single most popular thing I’d ever posted. I think it got over 11,000 views in a few hours. I wasn’t sure if the image or the title drew the most folks (I’m sure many office workers have probably felt the same).
But I digress. Nowadays, visual and auditory elements make a massive difference in content visibility. Every blog post must have a relevant image, preferably an eye-catching one, even better if it’s eye-catching and funny in some way.
Surfing the free image sites didn’t help me. So when DALLE-2 became available, I started using it to create images for my blog posts. In some cases it created compelling images (I especially like this one from a previous post). In others, it took tweak after tweak to develop something that I thought was compelling enough.
Artists who are concerned that DALLE-2 or its siblings will take their jobs away need not worry. The work is OK, but if you ask me, it lacks soul. It’s “hollow.” It might be fine for throwaway images to accompany a blog post, but as real art – it is simply a mirror of what we have done before. It’s a mirror of humanity but a ghost – content without real substance.
So when ChatGPT-3 was released, the media storm around it turned into a full-fledged tornado; I felt that I had to try it out to see if it would create something that could approximate a human-written post. I have used AI in the past to generate blog posts (for example, this post which is pre-ChatGPT -3), but it still isn’t up to human standards. It’s just like the DALLE-2 images – serviceable – but soulless.
You can tell that a machine wrote them. They might be able to generate excellent content marketing copy full of keywords to help attract search engines and improve SEO, but mimicking the creativity and insight of a human writer? Not there yet.
However, that doesn’t mean that OpenAI and its various tools are pointless. If you ask me, it’s a great tool to help you extend your thinking – to assist you in developing new concepts that you may have not thought about. I said before that humans depend on logic, emotions and intuition to make decisions and create new ideas because we simply can’t know as much as an AI, who has instant access to a massive corpus of data. AI can only create anything that approximates human speech and art because it has so much data to choose from that even semi-random combinations of things seem original.
For example, we tried a few things with ChatGPT to see if it could tell a creative story. It repeatedly told us that “No, a pickle and a glass of water cannot go out on a date as they are both inanimate objects”. But when we asked it to ask us “impossible questions”, I got some great writing prompts that you could turn into a story:
- What would happen if unicorns could fly to the moon and play soccer with aliens made of marshmallows?
- What would happen if a giant robot made of spaghetti fought a dragon made of jelly beans?
- What would happen if a time-traveling toaster went back in time and toasted bread for dinosaurs?
- What would happen if a giant, talking cactus decided to run for president of the United States?
What does this mean? OpenAI is an essential tool that can help you innovate. Working in conjunction with OpenAI tools, we can have it create new connections that we may not be able to see, due to our limited data sets. The ideal content creator is a centaur: a human and machine working together to deliver a new experience.
Sure, many humans might use OpenAI tools to get out of work (I’ve said before that humans are resource-preserving machines – although some might call it laziness), substituting sub-optimal outputs of the tools to ease their burdens, but if you want the best possible output, working together is the ideal solution.
OpenAI does the grunt work, suggests new pathways, and helps straighten your thoughts. You take its output and iterate on it until it gets closers to what you need.
You, the human, need to add the soul.