I got to know about this RecurrentGPT the other day and decided to give it a try. It is designed to have unlimited context length, yes, unlimited. And works based on Long-Short Term Memory (LTSM) mechanism.
But let’s keep that aside and talk about what I was able to create by using RecurrentGPT:
I provided a 1-paragraph description of the story I wanted, the description read:
A child is abducted and taken to an alien land where he learns their ways and finally returns home after a long time.
And as it was just for a test, I didn’t put significant effort into writing the prompt, as you might already be noticing.
The program kept running for almost an hour and then I had a 15,000-word story. I didn’t have to click any buttons or do anything else during that 1-hour.
I also converted the story to audio format by using text-to-speech features at NaturalReaders. Here’s the long 1 hour 16 minutes audio for the same:
About the output, I think, it’s not a superb story. It keeps repeating certain things and feels like everything is going in a loop. I remember the mention of how their love grew stronger more than 10 times.
Even if you read/hear the entire story, you may still not completely understand it, especially after the second half. It just sounds too technical and doesn’t have the human touch.
However, I am always optimistic and I am sure it can be further improved.