re: #24 dat_said

Training data is going to be a concern, especially due to latent bias and quality of data sets. Companies that have high quality data, are going to have an easier time applying that data to the model and getting better results. The key will be minimizing hallucinating activities and what will be considered acceptable percentage of such hallucinations.

If you have 99% certainty, that means you’ll get an incorrect result, that seems great, but if you’re getting 100000 searches a day, you’ll be spitting out a bunch of incorrect outputs, and depending on the use, it could have deadly consequences.