You are right that a transition to the cloud is not going to help. Bad software will still be bad in the cloud, good software can be just as good (and often better on several axes) when not in the cloud. Though from reading some of the complaints in the thread you link here, Amazon does have very good OCR software (called Textract), so that could help.
But on this point:
> And while search technology has improved, it hasn't become orders of magnitude better.
In general it has, it is just that the USPTO's software hasn't. I imagine a team of the right 2-5 people could make something better than you would wish for in your wildest dreams. But how to actually make that happen is another kettle of fish. Anyone capable of fixing this would most likely be better remunerated doing something else, but nerds are easy to snipe and even just reading your gripes here and in the subreddit has made me want to solve this problem if for no other reason than the humanitarian one.
> In general it has, it is just that the USPTO's software hasn't.
On the contrary, USPTO's internal search tool (PE2E Search) is overall good and has many features that I wish public search engines like Google had. The main difference is that it's designed for power searchers, not the general public. Yes, PE2E Search has a lot of issues, but the USPTO contractor who commented here has probably never used it to search for patents and thus has no basis for comparison. You should take what they said with that caveat in mind.
My basic point is that no existing search technology makes up for the sheer increase in documents to search. The last significant innovation in patent search was switching to computerized search in the 90s. The changes since then have been relatively minor, but the number of documents to search since then has increased dramatically. Maybe some AI based search will eventually be a game changer, but for now it's not (I've tried 5+ AI search tools and they usually aren't good) and I don't see that changing any time soon.
But on this point:
> And while search technology has improved, it hasn't become orders of magnitude better.
In general it has, it is just that the USPTO's software hasn't. I imagine a team of the right 2-5 people could make something better than you would wish for in your wildest dreams. But how to actually make that happen is another kettle of fish. Anyone capable of fixing this would most likely be better remunerated doing something else, but nerds are easy to snipe and even just reading your gripes here and in the subreddit has made me want to solve this problem if for no other reason than the humanitarian one.