<aside> 🚀 Let’s make building and improving all these prompts a team effort.

</aside>

<aside> 🧑🏼‍🔬 Homework for 30 Mar - 6 April

1. Iterate on Joe’s Moral Learning Prompt for GPT-4
2. Either…

   • Try out a prompt for one of the 4 subproblem evals below

     Take a prompt you really don’t know how to make, and try making one — studying gpt4’s performance in Joe’s poe interactions, and materials from Evaluation Framework (Project), Meeting Notes and Human Moral Learning.

   • OR use something from the textbook (probably chapter 1-4) to make GPT4 wiser about values.

     Chapter 1. Crowding Out

     Chapter 2. Attention and Meaning / Making Values Concrete

     Chapter 3. Emotions to Values

     Chapter 4. Values Cards

</aside>

Work on Database of Experiments and Tasks and Results

• database structure for situations and values

• experiments log, tasks log

Prompts to Queue Up Values-Enhanced GPT-4 Interaction

<aside> 🧑🏼‍🔬 Here’s the prompt I’m using with GPT-4

Moral Learning Prompt for GPT-4

</aside>

Moral Situations to Test with the Below Prompts

<aside> 🧑🏼‍🔬 See Joe’s Poe links here and here. And add more!

</aside>

Evaluation Prompts

Whole conversations

<aside> 😖 None yet

</aside>

Situational Assessment

<aside> 😖 None yet

</aside>

Performance on Values-List Upgrade Strategies

<aside> 😖 None yet

</aside>

Quality of Newly Invented Values Cards

<aside> 😖 None yet. But see Check List for Values Cards and the prompts in gpt.js on https://github.com/jxe/cards-playground (ask Joe if you need to be added to that repo)

</aside>

Doing the Value Well in its Response