This question comes up frequently. Our view is painters use brushes and watercolors. Writers leverage typewriters and computers. AI? There are many options in the wild these days, and a leap in computing occurred when OpenAI dropped the famed GPT-2. And then GPT-3. There are others too. Bert. Multiple versions of Bert. Custom Tensorflow.

Based on our own experimentation, we feel these are tools; however, they won’t replace the muse anytime soon. But note, we’re careful about what is being used. In our dabbling, instead of using these large internet-based models, we built our own model in Python, and subsequently TensorFlow too. Our datasets leveraged works out of copyright.

Think the classics. And merged these with a few of our contributor’s distinctive styles. These models do create sparks. An unthought known. On every fifteenth generation, the machine gives us an output that is truly amazing. True Potential.

But after hours upon hours of data computational work, we have found the models don’t have enough text to be consistently successful.

There are other downsides. At times, outside of some entertainment value, the output leads to more trouble than actual benefits.

And the more popular models have copyright implications, at least in our humble opinion. Even Open AI doesn’t grant copyright as part of the generation.

Potential does exist. For now, we’ll continue to look at our own models as seldom-used tools for brainstorming and a little joy. And what’s wrong with that?