From time to time, we publish here some of our thoughts on IT and law, some short, some longer, as well as links to our publications and events available on other websites.
This is a short transcript of my presentation in Warsaw at the Modern Bar Association (2023) New Legaltech challenges conference. You can find the presentation slides here. These are the notes of my presentation with slide numbers in square brackets [ ].
In this blogpost, I would like to highlight the importance of crowdsourcing some datasets by lawyers that are specific to jurisdictions and legal areas. I will also illustrate the difficulties with my own, tiny experiment in the field of legal question answering in Hungarian while using the large language models, both with and without the legal texts provided.
I’ve spent a little bit of time with a front-end for using Whisper, the free transcription model from OpenAI and trying to combine it with OpenAI chat completion (GPT-4) API to use it for transcribing audio files to readable text files.
It’s very promising and it’s multilingual.
See an example result in the previous blogpost. The only bad experience I had is with the usability of GPT-4 API in terms of receiving too many “out of time” requests for it to be usable, at least from this particular point of Earth (probably not the most important place for OpenAI), for any production purposes at this moment.
But this is not a theoretical limitation of course, and I hope it will be sorted in out by some AIs in the coming months…
See a more technical description in the GitHub directory README of the application itself here.
