4 ms·
Sometimes I wonder if it wouldn't help to try to set legal precedents that extend todays extreme intellectual property obligations in software to legal document
by BenoitEssiambre 14y ago
Sometimes I wonder if it wouldn't help to try to set legal precedents that extend todays extreme intellectual property obligations in software to legal documents.
When you come to think of it. A contract is very similar to a piece of code. The main difference is that it is to be executed by humans instead of machines and enforced by a legal system instead of a machine's design.
Contracts have been deemed non patentable and a high bar has been set for copyrightability compared to other text because of the requirement that contract language be well defined with particular sentences having an agreed upon interpretation based on court precedents.
This means lawyers writing contracts don't have to worry, for each sentence that they write, that it may be infringing on someone else having written a similar one. They also don't have to worry that a combination of clauses have been patented and they can't use it in their contract.
Programmers on the other hand worry about it for every new piece of code they write. The bar is so low for patentability and copyrightability that we pretty much know someone somewhere has patented a large part of the code we write. We just hope nobody with the patent is interested in going after us because we aren't worth the trouble. But there is always that feeling that if ever we become commercially successful, the trolls and wolves will be waiting to try to get a slice of our profits. This is quite an annoying disincentive to write new code or 'innovate' which in this case means just plain doing our job.
- BenoitEssiambre 14y agoI was pondering this further and it would actually be conceivable with a bit of natural language processing and machine learning to build a 'verdict machine', maybe a Watson type of thing. You would input legal documents, contracts and laws to 'program' the machine, It would read these legal documents and compile them into legal tests that depend on case facts. Then when you needed to judge a case, you would enter the facts or fact probabilities about the case in the machine. It would then generate a verdict and tell you which legal tests passed or failed to arrive at the verdict. Even it it was just used as support to help courts untangle cases or for law student as an educational tool, it could be a useful machine. Here is the kicker: All contracts and laws are actually software code for this machine.
- jeremyjh 14y agoYou've certainly got lots of data to train it with. And you can test it by simply testing its predictive power. But instead of selling it to the court system I'd sell it as a service to law firms and/or use it in prediction markets. This would ensure that it would be funded and refined until everyone simply accepted its outcomes as a foregone conclusion.