5. Why language models / prompting works. We still don’t have a great understanding of how in-context learning works, or why chain-of-thought/reasoning works. We can learn a great deal by only looking at inputs and outputs (as do we do with humans in psychology).
PROMPT ENGINEERING
-
Prompting Research: Just the Beginning for Language Model Guidance
By
–
1. Prompting research. Maybe hot take, but I think we’ve just reached the tip of the iceberg on the best ways to prompt language models. As language model capabilities increase, the degrees of freedom for guiding a particular generation via a good prompt will increase.
-
PhD Research Directions in the Era of Large Language Models
By
–
I’m hearing chatter of PhD students not knowing what to work on.
My take: as LLMs are deployed IRL, the importance of studying how to use them will increase.
Some good directions IMO (no training):
1. prompting
2. evals
3. LM interfaces
4. safety
5. understanding LMs
6. emergence -
Staying Updated on LLM Jailbreaks and Exploits
By
–
well, now that @gdb qt'd this tweet, I feel I have to share this… keep up w the current state of jailbreaks and LLM exploits by subscribing to my newsletter here: http://
thepromptreport.com -
GPT-4 fails to write without ‘e’ via naive prompting
By
–
Writing without the letter "e" is still beyond GPT-4 with naive prompting, but that's maybe a cheap shot.
-
Conversational Chat Interface for Sensitive Data Editing
By
–
It's full of sensitive data, but enough people asked so here's how what it looked like. (Obviously redacted) Was easy to make edits via conversational chat, add more notes, etc.
-
Jailbreaks Remain Viable With Creative Approaches
By
–
totally agree, jailbreaks are still an evergreen field, it just will take a little more creativity now to write them
-
GPT-4 Jailbreak Discovery Request
By
–
as always, let me know if you devise a GPT-4 breaking jailbreak!
-
GPT-4 Jailbreak Difficulty Scales Exponentially with Output Severity
By
–
There is a sliding scale for jailbreak output that exponentially increases in difficulty to crack It's trivial to get GPT-4 to curse but if you want a set of instructions on making a weapon it's going to take a lot of work
-
The Future of AI Jailbreaks: Increasing Complexity and Sophistication Required
By
–
overall, as I expected, the nature of jailbreaks will need to change jailbreaks will require more complex reasoning and intuition about the model and won't be able to be written in 5 minutes