There's simply not enough data for a neural network to assist...

There's simply not enough data for a neural network to assist someone to reproduce Bene's iPPS and many have tried, and failed, to reproduce it in the past.

One key point is iPPS is not a single molecule, but a complex heterogenous mixture.

With regards to the mixture, not all of the inputs have yet been fully characterised. All we know is the composition of the mixture (partial or fully) and the fingerprint. People have tried to characterise the mixture in the past and managed to gather some information, but failed to characterise it fully. I also cannot comment on whether a bunch of similars has been produced in the past. While other forms of PPS have been created, I think the mixture is completely different to Bene's iPPS. The mixture and composition changes the biological properties.

While I doubt AI can be used to reproduce Bene's product, I think it is theoretically possible to reproduce the product in the laboratory if one were able to fully characterise the mixture. In this case, the issue will come down to cost; how much it costs Bene to produce vs the alternative option. I think anyone trying to copy the product will find the costs to produce to be prohibitive.

My views only, dyor