Are we training the language models, or are the language models training us?
AI
-
Language Model Performance: Emotional Intelligence vs Reasoning Strategies
By
–
This battle is quite revealing. Karina's prompting strategy was getting the language model to say it understands (emotional intelligence). My strategy was to get the model to reason more (IQ). Our strategies each worked on the respective language models that we are used to using.
-
Failed Claude Prompt Works Successfully on GPT-4
By
–
Then i copy-paste in my prompt that failed on Claude. Much to my relief, it actually works on GPT-4.
-
Confirming GPT-4 Limitation with Simple Prompting Approach
By
–
First, I confirm that it doesn't work on GPT-4 with simple prompting, to prove the prompt matters, and it's not just because GPT-4 is smarter than Claude.
-
Surprise at an Unexpectedly Effective Strategy Without Prior Knowledge
By
–
I am bewildered that this works. Obviously she hasn't read the chain-of-thought paper, since that is the natural strategy to try.
-
Precognition Prompt: Forcing the Model to Understand the Task
By
–
She shows me her prompt. It does what she calls "precognition", which teacher-forces the language model to understand the task. No intermediate reasoning is needed, the language model could already do the task!
-
Last quip about home court advantage with GPT-4
By
–
Embarrassed, i make a last quip. It must be because i gave her the home court advantage! I try my prompt on GPT-4.
-
Conceding Defeat and Asking Karina for Her Prompt
By
–
At this point my time is up and I concede defeat. I ask Karina what her prompt is.
-
Claude AI Model Performance Issues and Limitations
By
–
Unfortunately that still doesn't work. Claude fails miserably.
-
Failed Attempt at Claude Coding Mode Prompt
By
–
I finally started prompting after Karina's 3 minute head start was over. My first try was to have Claude go into coding mode by writing some code and pretending to execute it. Unfortunately it didn't work.