CustomGPT for Discussions on Proposed New Rules of Evidence on Deepfakes.
For complete background see Ralph Losey’s blogs on the complex and challenging topic of proposed changes to the Rules of Evidence pertaining to AI and Deepfakes. AI-Generated Evidence and Deepfakes: Is it Time for New Rules of Evidence? Part One and Part Two.
To help anyone to better understand this challenging subject Losey created this Custom GPT — Artificial Intelligence and Evidence — with Retrieval Augmented Generation (RAG) capabilities. It incorporates fourteen recent documents, including my two-part article and pertinent Evidence Rules and articles cited therein. This includes an essay by Professor Daniel Capra, Deepfakes Reach the Advisory Committee on Evidence Rules, 92 Fordham L.R. 2491 (2024), which Professor Capra describes as follows:
This short Essay provides (1) a brief introduction to deepfakes; (2) a short description of how Rule 901 operates; (3) a description of the Committee’s review of social media and digital communication—the previous technological developments that challenged the evidence rules on authentication; and (4) a description of the Grimm-Grossman proposal to add a new provision to Rule 901 that will provide a procedure for assessing deepfakes, as well as of two other suggestions for change made in recent law review articles.
Other important RAG documents in my custom GPT updated information bank include the pertinent evidence rules and:
- Professor Delfino, Pay-To-Play: Access To Justice In The Era Of Ai And Deepfakes, 55 Seton Hall L.Rev., Book 3, __ (forthcoming 2025);
- Judge Grimm and Professor Grossman, et al:
- The GPTJudge: Justice in a Generative AI World, 23 Duke Law & Technology Review 1-34 (2023);
- Deepfakes in Court Artificial Intelligence as Evidence, 19 NW. J. TECH. & INTELL. PROP. 9 (2021);
- Deepfakes in Court: How Judges Can Proactively Manage Alleged AI-Generated Material in National Security Cases (Northwestern Public Law Research Paper No. 24-26, (9/10/24); and
- REVISED Proposed Modification of Current Fed. R. Evid. 901(b)(9) for AI Evidence and Proposed New Fed. R. Evid. 901(c) for Alleged “Deepfake” Evidence (9/10/24).
Anyone with a ChatGPT account can use this free GPT Artificial Intelligence and Evidence from the OpenAI store to ask questions or dig deeper into any related topic. Beware this topic is complex and difficult. That is one reason Ralph Losey created the GPT, to help him to figure this out. Still, Ralph has already seen the GPT make mistakes. So as always, trust but verify. Errors and sometimes even hallucinations, where the AI actually makes things up, are still a fact of life these early days of generative AI. Hint – if you see a picture with a person with six fingers, it is fake. AI still has trouble getting hands generated correctly, but it is improving everyday. Also, if you see a speaking video like the one below, look very closely at the mouth and teeth while the “person” is talking. For help detecting AI generated documents, “delve into” my article Stochastic Parrots: How to tell if something was written by an AI or a human? 4/5/24) (delve is an obvious “tell word”).
Here is a video I made for the Stochastic Parrots article. Since it is so obviously an AI generated pirate speaking, perhaps it should be called a shallowfake.
For the full list of AI “tell words” see the article: Stochastic Parrots: How to tell if something was written by an AI or a human?
A word about ChatGPT privacy: there is complete privacy protection from me of your all of your interactions with my custom GPT, Artificial Intelligence and Evidence. I cannot see your questions or answers, or even know if you have ever used it. OpenAI people do not see them either, but their vast array of Nvidia computers may ingest your questions and other input for AI training purposes. They only use the chat conversation data for training if you do not pay the $20 per month for the full privacy protected version, ChatGPT Plus. All legal professionals should use the paid subscription with privacy, instead of the free model where your entries may be saved and used by OpenAI for training.
Ralph Losey Copyright 2024. All Rights Reserved.