How to Get Good Output from ChatGPT: Few-Shot Prompts
Getting good output from GPT is a matter of consistency and providing good examples for the model.
“Universal and Transferable Adversarial Attacks”: Researchers Jailbreak GPT
“Universal and Transferable Adversarial Attacks”: Researchers Jailbreak GPT Researchers with Carnegie Mellon’s Center for AI Safety have published a paper describing a method for developing adversarial attacks (a.k.a. jailbreaks) that was broadly effective across models. The jailbreak in question originates … Read more
An Interesting Quote from Sam Altman
In a recent article at The Atlantic, I noticed this snippet about Sam Altman, CEO of OpenAI (ChatGPT): ““We could have gone off and just built this in our building here for five more years,” he said, “and we would … Read more
How to Use ChatGPT Custom Instructions
OpenAI’s ChatGPT has a new “Custom Instructions” beta feature, available to ChatGPT Plus users as of Friday (but not in the UK or EU). It seems to function as a combination of system prompt and augmented retrieval memory store. Here’s … Read more
AI “Safety” and AI Ethics Are Not the Same
I have been agitating for some time now over the decreasing performance of AI models like GPT-4 or (especially) the new LLaMA 2 in direct relation to the “safeguards” being trained into them to prevent abuse, misinformation, and dissemination of … Read more
Llama 2 Is Out of Its Pen
Meta AI has launched a new language model called “Llama 2”. This is an upgrade from their previous model, Llama 1. What’s new with Llama 2? It’s bigger and better. The models range from 7 billion to 70 billion parameters. … Read more