The most prestigious law school admissions discussion board in the world.
|
 |
 
|
LJL at how easy 1L must be these days with ChatGPT
| Let\'s dispel with this fiction | 09/28/24 |
| ..,.,,,...,..,,.,...,.,,.,.. | 09/28/24 |
| Let\'s dispel with this fiction | 09/28/24 |
| ..,.,,,...,..,,.,...,.,,.,.. | 09/28/24 |
| haram hustle | 09/28/24 |
Poast new message in this thread
Date: September 28th, 2024 7:15 PM Author: Let\'s dispel with this fiction 1. Get outlines from 3 previous students for each class from last year.
2. Tell chatgpt to create a master outline.
3. Read each outline once per week. Ask chatgpt to help make analogies for legal issues with shit you like (e.g. South Park, Star Wars, etc.) and mnemonic devices to help you memorize every outline. As it gets closer to exams, tell chatgpt to create a "leansheets" style outline for each class.
4. Quiz yourself on it with chatgpt voice.
All this can be done for the bar exam too lmao. Throw in a bunch of old exam questions and ask chatgpt to create new questions in the same style.
|
Date: September 28th, 2024 7:21 PM Author: ..,.,,,...,..,,.,...,.,,.,.. I was testing out o1 preview edition with LSAT type questions and it got every one right. I understand there is a difference between the LSAT and being a lawyer, but people starting out now should not feel comfortable with the idea they will have a long career. Language models were babbling like idiots a few years ago. (http://www.autoadmit.com/thread.php?thread_id=5602898&forum_id=2...id.#48141371) |
 |
Date: September 28th, 2024 7:27 PM Author: Let\'s dispel with this fiction Exactly. Will be interesting to track law school admission numbers over the next few years. I have a feeling it will plummet significantly.
o1 preview is just today. What will Chatgpt 6 or whatever look like 3 years from now. The law student starting today is royally fucked. |
|
Date: September 28th, 2024 8:10 PM Author: ..,.,,,...,..,,.,...,.,,.,.. They are using o1 to create data for training the next generation of models. So now they can generate as much high quality of reasoning data as they want and then use it to train GPT-5 or 6, which can then be used for more synthetic data creation. Billions and billions of reasoning traces to learn from. The models by 2030 will be completely nuts. (http://www.autoadmit.com/thread.php?thread_id=5602898&forum_id=2...id.#48141492) |
|
Date: September 28th, 2024 7:33 PM Author: haram hustle (http://www.autoadmit.com/thread.php?thread_id=5602898&forum_id=2...id.#48141387) |