Spaces:
Sleeping
Sleeping
Create prompt_fails.txt
Browse files- prompt_fails.txt +13 -0
prompt_fails.txt
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Why do you think the prompt failed?
|
| 2 |
+
Typically a prompt fails for one of three reasons:
|
| 3 |
+
- Cutoff: the training data is cutoff before a point where it could know the answer
|
| 4 |
+
- Obscurity: the context necessary to answer is too niche to appear in training data
|
| 5 |
+
- Hallucination: the AI made something up instead of stating it didn't know
|
| 6 |
+
- Intelligence: the task was too complicated for the current state of the art models
|
| 7 |
+
- Direction: the prompt didn't provide sufficient guidance on how to answer
|
| 8 |
+
|
| 9 |
+
What does this failure tell you about what AI is good at / bad at?
|
| 10 |
+
The default answer from AI often isn't near the best it can do. When presented with correct context and guidance, AI can do a lot more than you expect!
|
| 11 |
+
|
| 12 |
+
Can you think of any ways to fix this prompt?
|
| 13 |
+
We'll be learning about the five principles of prompting in the next section: compare your ideas here to what you come up with after watching the next section.
|