[dbcurtis]

Couple of thoughts.

My wife was a law firm junior associate in the banking law group of a major firm during the S&L crisis. She spent a lot of time on the due diligence work involved in merging sick S&L's into each other. I suspect that if faced with a similar Augean stable, having auto-summarization and auto-triage would be a hugely beneficial. True, need for fewer junior associates since they are looking only at exception documents and otherwise tuning the "queries", for lack of a better term. But I think it would help even out the work load and the net impact is a healthier organizational structure. (That experience made her rethink her career path, she ended up in corporate practice doing software licensing.)

In the area of patent law, claims have a very regular structure and I think their analysis would yield to analysis given the current state of machine learning. Given the stakes involved, it seems to me that automated claims checking and claims analysis would help everyone. Patent firms could produced better work product in less time. There is no great surplus of patent attorneys, and given the time and cost constraints, many companies pursuing patents tend to establish a budget and attack patents in priority order until the budget is gone. I don't think patent firms will end up billing any less in total, but simply bill the same for more patents each completed for less money. As a small-time inventor, this would be good, because if you only want one patent, you see your costs reduced.

[polskibus]

And then you end up writing text in a way that fools auto summarizing AI. War against AI starts in law.

[entee]

This. So much this. One big problem with AI/ML as is currently eating the world is that it's just a really fancy averaging engine. It memorizes (incredibly, beautifully, superhumanly) but it doesn't really actually understand. There's a whole spate of "lets trick the AI" in image processing, I have to believe that there are easy ways to do the same with text.

That said, one counterargument could be that obfuscating an AI will lead to more confusing contracts which could actually end up making them harder to enforce. So perhaps in this case there's a counter-force.

[wpasc]

I read a great, great paper about training systems against this. Obfuscation is currently easy against AI, but obfuscation that fools an AI and a human could still detect (Purposeful perturbations). This can be pre-trained against by incorporating such perturbations programmatically during training.

Link: https://arxiv.org/pdf/1705.06640.pdf

[rhaps0dy]

>There's a whole spate of "lets trick the AI" in image processing

Check out this "let's trick the human" in image processing:

https://en.wikipedia.org/wiki/Optical_illusion

[lern_too_spel]

"Let's trick the human" is already a large part of law.

[zechs]

In a way, the way contracts and legal documents are written now are already an antagonistic optimization against humans natural language ability. A typical contract is pretty difficult for a human to parse without experience. Imagine how much more tricky and complex legalese can get if lawyers start battling machines!