Hi, i am drawing a variable semantic query from a notion table, sorting by jurisdiction and relative professional compliance standards, historical cases, chronologies and .txt data,which are drawn from Dropbox, aggregating texts as the volume is high, passing them into AI Chat GPT 4.0 and then creating reports as to whether compliance infractions are detected so as to generate ultimately a snapshot of a highly complex position. The intention is that once all standards are loaded, predictive scenarios can be interrogated with predictive outcomes. Chat GPT is driving me nuts over 5 weeks and I am burning tokens like there is no tomorrow. Is this something I can get assistance on? Notion/Dropbox/Tools/aggregators/PDF.co/AI chat GPT/ etc
If you can help, you can directly reply to the post with helpful advice, instead of asking for permission to help first…
Do note that forum members are not allowed to solicit or advertise their services in other categories like here (even if it’s free/unpaid). Asking users to post in the Hire a Pro category will allow you to assist others outside of this forum, so you should direct them to post in the Hire a Pro category.
Hope this helps! Let me know if there are any further questions or issues. P.S.: investing some effort into the tutorials in the Make Academy will save you lots of time and frustration using Make!
— @samliew
Hi, can you say how this works? I have in the couse of design, created a scenario construct that enables the interrogator to type a semantic question into (currently) Notion. The data is presently court cases in .txt form that are cloud stored in Dropbox. There are also within Jurisdiction folders (presently Jersey, UK and Isle of Man) laws eg Jersey Trust Law, Jersey Company Law, STEP rules, Jersey Financial Services Rules for Trustees, IEAEW rules of compliance. The intention is that the semantic query eg “did the Trustees need beneficiaries permission to launch litigation costing the Trust funds” or “What are a Settlors powers” or “Should a Trustee inform the JFSC if they are involved in Litigation” etc. the AI considers the selected questions in a non hallucinating manner and acts as a robust advisor by searching through the texts (which can be voluminous) within the jurisdiction, and then advises the interrogator in OSCOLA format whether there is a breach. And searches BAILII for authorities and cites them. Eg on Q3 above the answer is in the JFSC rules for Trustees under Article 6, Q2 answer is in the Trust deed and Trust Law. And Q1 is more complex due to a KC advising such, coupled with Trust law and authorities. This model could apply to any professional including eg an Architect lying to a client, the interrogator would search ARIBA and other compliance rules and give quick answers with cases cited etc. I think if the AI was adequately robust that this model is scalable and could be marketed. Can you say how you would be able to assist in A) the architecture which I have currently designed and B) the robustness of the scenario? I have had it working but changes made by Chat GPT threw me off. Ie answers have returned ‘Breach identified’ cited rules and laws etc. the purpose of this is to take it to the predictive stage for compliance.