Тёмный

law.MIT.edu IdeaFlow: Legal Prompt Engineering Examples and Tips 

law.MIT.edu
Подписаться 1,3 тыс.
Просмотров 5 тыс.
50% 1

* Host: Dazza Greenwood, Executive Director, law.MIT.edu
* Guest: Damien Riehl, VP, Litigation Workflow and Analytics Content, Fastcase
* Date: Tuesday, February 21st, 2023, at 12:00 PM Pacific / 3:00 PM Eastern
On this episode of law.MIT.edu's IdeaFlow, guest Damien Riehl joins host Dazza Greenwood to explore the concept of "Legal Prompt Engineering" including an example Damien recently posted showing how a legal brief can be developed in response to a complaint.
* The current version of this episode: law.mit.edu/pub/ideaflow15
* For more episodes of IdeaFlow, see: law.mit.edu/media
* To signup for IdeaFlow episodes: forms.gle/aZbxfsEHkM7egeyf8

Наука

Опубликовано:

 

20 фев 2023

Поделиться:

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист
Посмотреть позже
Комментарии : 8   
@JG27Korny
@JG27Korny Год назад
As a lawyer the challenges with the advent of AI are numerous. On one hand a lot of the real legal problems include some computational process as the world is getting more and more complicated, on another level the data science also has real legal processes impact. So an AI can really help close the knowledge and data processing gap. For a lawyer, the learning curve is indeed steep because just basic prompt engineering knowledge is barely enough without general culture about how those AI work, and the inherent risks of hallucinations. For that reason I find that the MIT 6.100A Introduction to Computer Science Programming in Python and 6.100B Introduction to Computational Thinking and Data Science has to be a basic course today not just for engineering departments but also for all legal professionals. To tell the truth for a legal professional the learning curve is very hard. In Europe the continental legal system makes the lawyers to study just law for 5 whole years in the university. And that makes them to drift far away from anything STEM related.
@DazzaGreenwood
@DazzaGreenwood Год назад
Thanks for sharing your experiences and insights. And MIT CS courses, or some version of them, really should be part of legal education, as you say.
@JG27Korny
@JG27Korny Год назад
@@DazzaGreenwood AI is not asking less from lawyers. Quite the opposite to be able to use to the fullest the new tools it is necessary to have a lot of knowledge e.g. prompt using game theory components can give results that give astonishing insights. While creating a set of useful prompts is a great endevour, one of the best results come from using the classic Socratic dialogue asking to emulate different persona, e.g. top lawyer, financial analyst, ballistic expert etc. Last but not least advanced jail breaking techniques could be necessary in criminal law ;)
@daslolo
@daslolo 11 месяцев назад
@@JG27Korny can you give examples of "using game theory"?
@JG27Korny
@JG27Korny 11 месяцев назад
@@daslolo There are many in the book Game theory and the law
@JG27Korny
@JG27Korny 11 месяцев назад
For example a prompt describing the context of a tactical situation. You can ask do you see a possible dominant strategy of the opposing team. Or are there possibilities of bayesian equilibrium. @@daslolo
@cyberpunklawyer
@cyberpunklawyer Год назад
Unfortunately, the value of this approach is directly proportional to the verbal reasoning and argumentative writing skills of the lawyer(s) who drafted the original pleading, and in particular, the rhetorical alacrity of the headings and subheadings they formulated for each argument, which Microsoft Word then automatically pulls in to generate the Table of Contents that is the starting point for the subsequent legal prompt engineering process. In coder's terms: garbage in, garbage out. Or not.
@lawMITedu
@lawMITedu Год назад
Yes, I see what you mea about this approach being heavily dependent on the quality of the outline of the document one is responding to, And I see how that may come across as "unfortunate" depending on what one was expecting to start with. This is certainly a limit of the approach. I sincerely doubt, however, that there is a single approach for all potential situations. Meanwhile, this approach, as Damien so aptly demonstrated, may be a very good fit for situations when the headings of the document one is responding to are well crafted. To me, this is a significant advance beyond what was widely and inexpensively available only a few months ago. So from my own point of view, though it is of limited application, it is very fortunate to have such a capability at one's fingertips. Always use the right tool for the job, as they say.
Далее
НОВАЯ ПАСХАЛКА В ЯНДЕКСЕ
00:20
Просмотров 505 тыс.
What Creates Consciousness?
45:45
Просмотров 105 тыс.
ChatGPT Prompt Engineering for Legal - Tips and Traps
57:34
What are AI Agents?
12:29
Просмотров 59 тыс.
Tarpit Ideas: The Sequel
13:50
Просмотров 31 тыс.
2024 MIT Computational Law Workshop
3:04:24
Просмотров 3,8 тыс.
ОБСЛУЖИЛИ САМЫЙ ГРЯЗНЫЙ ПК
1:00