Is AI coming for our blogs?

One question that seems to be on the mind of creatives everywhere is the coming AI revolution, and whether it’ll remove all our productive capabilities, preventing us from reaching employment? I certainly don’t think so. Here’s why!

First, AI’s writing is bad. When I tried to get the latest version of chat-gpt to review my cover letter, and improve it, I realized that even if one ignores hallucinations, the products it produced were too wordy and lacked pizazz. Its strengths are building common structures, grammar, plagiarizing, reformatting data, and summarizing work, but if even the reliability of that is in question, AI will need those able to troubleshoot or ask important questions.

I wrote a standup bit that got a bunch of laughter a while back, and included a couple other bits from my ouvre that I’d tested on my friends. When I asked GPT to produce another one in a similar style, it pretty much replaced a few words and maintained the original bit. Frankly, it was embarrassing that it couldn’t produce anything noteworthy or artistic.

AI also sucks at following directions. I was trying to reformat and simplify a list of medical codes for a previous job, and it did the first ten and stopped. Despite instructing it repeatedly to continue, we were only able to get 10 codes at a time. This leads me to think that certain parts needed to be jail-broken in order to begin to be functional. All the biases added in by developers like OpenAI, Google, and Microsoft make a worse product, albeit maybe a safer one.

One statement that exemplifies the limits of AI is: junk in, junk out. The training sets used to improve writing are necessarily going to be filled with nonsense in order to get loads of information accessible in the first place. Unfortunately, due to its unmonitored nature, AI is not getting sent the best and brightest. Those developing AI could try and rectify this in the other direction by combing through all material beforehand, but that would strictly limit its breadth of knowledge in the first place, and introduces human biases that may not be perfect themselves.