Your hypothetical neural network models don't exist. They won't be used to test SMTs or anything else.

Yes, I know - simulation, quality assurance and testing processes need to be developed. Frankly, it's shocking they weren't, and explains why after two and a half years the network is more broken than ever. (OK, hyperinflation was worse.) Simulations are never ideal, but they do greatly help develop sustainable, if flawed, systems, that are easier to iterate over. I'm not a software developer, but the same tenets have worked across most professions since the dawn of the industrial age.

As far back as 2016, I have been asking for proper testing processes and peer-review of Steemit Inc's code. If the agenda is to continue whack-a-mole experiments without any evaluatory processes - I suppose it is now time for me to give up and move on. All the best, though, I'd be delighted to be wrong and bidbots dramatically decrease in influence (adjusted for the natural downtrend over the last few months) or disappear in a few months' time.