Definitely very easy to get very generic results. I've got some strategies to avoid that, like giving it multiple examples, but most people don't want to hand over a prompt that is multiple pages long. It's an interesting aspect of the challenge.
LLMS
-
ChatGPT as a Creativity Juice Dispenser
By
–
ChatGPT is kind of like a creativity juice dispenser, eh?
-
Protecting LLM Prompts: IP Security and Injection Prevention
By
–
With more and more LLM-powered products with prompts as their IP or USP. It is becoming more and more important to protect their prompts from leaking to competition. Malicious prompt injection has also become a thing motivating people to misuse prompt based apps.
-
Protecting LLM Prompts: IP Security and Injection Risks
By
–
With more and more LLM-powered products with prompts as their IP or USP. It is becoming more and more important to protect their prompts from leaking to competition. Malicious prompt injection has also become a thing motivating people to misuse prompt based apps.
-
Demo of Anthropic’s Claude AI Model for Seinfeld Episode Creation
By
–
A few weeks ago I got early access to a new model from @AnthropicAI
. Its name is Claude. To demo, Claude will give the first scene of a Seinfeld episode where Jerry receives the supposed key to AI alignment from @sama
, ignoring dire warnings from Kramer. The prompt: -
GPT-2 Output Detector Limitations and AI Text Detection Evolution
By
–
No realmente. GPT-2 Output Detector, como bien indica su nombre, funciona para GPT-2. Fue una tecnología que OpenAI sacó cuando publicaron GPT-2 hace años, y aunque funcionaba, manipulando un poco el texto se podía engañar al detector. Ya con GPT-3 y demás IAs ni lo intentaron.
-
Prompt Engineering Won’t Disappear Soon Despite Diminishing Importance
By
–
i dont think prompt eng will disappear all that “soon” actually but this is a good thread covering our attempts to at least greatly diminish its importance
-
PPO Explanation Gaps and InstructGPT Fine-tuning Data Requirements
By
–
thanks man its a good thread. i feel like people dont explain PPO very well in these threads, thats the main part that always feels handwavy. i’d also like an idea of order of magnitude of finetune data needed for instructgpt but havent seen good numbers
-
Cognition in LLMs: Associative Memory vs. Functional Awareness
By
–
That’s a good point. What humans see as cognition in LLMs is often just good associative memory, or (generously) intuition. A model that can echo “I am aware” is not necessarily, but in larger LLMs more of the functional implications of awareness begin to apply.