The EU Digital Omnibus - a potential failure in the making - A Citizens' initiative to correct our path - and your help is needed https://t.co/4j5Vv5lQdN via @LinkedIn
@jackwshepherd Agreed. Even though the language is English, it is still about instructing software to do certain things. There should however be predefined levers / check boxes / other configuration options which which have been properly tested and working in a consistent manner.
Generative AI for structured document automation? I have spent A Lot of time on bespoke document automation tools as well as Microsoft's Power platform. When you have a well defined setup/rule base/configuration you get consistent output. So what about us…https://t.co/Uy31yvCG1n
@AlexHamiltonRad I´m all for use of tech but on this specific topic I think raw data directly from executed contracts is to be avoided. However, absolutely to be used for the clause library curation. I would have loved the catchphrase though "when grasping at straws... haystack is there for you"
@jkubicki I've had similar experience where ChatGPT suddenly gets temporary amnesia on topics i just provided content on. More so with 4 than 3.5, maybe something to do with perormance threshholds on how fast to provide response.
@RyanWFinn To add of, the number of if/then is in this use case much lower than in a "normal" template and the queries are rather just getting all the items for the conditions at hand (can be 1 or 10 items). Categorization of topics is key for that setup.
@RyanWFinn The library is in an external database, using connectors. Updating the T/F values in the tables will affect the content without having to change the coding (separating code from data). But ofc you have to maintain a rule set (regardless if going through a doc as "container").
@Clevy_Law @legalytical Fool's gold: "Yay, the solution found me the end date". But then it missed the other document that amended the end date and the third one that terminated the contract... as mentioned in the thread, you need to understand your data layers.