If you have children, watched children, or watched adults who act like children, you have discovered that trial and error is how they learn. They speed run the mistakes of the past no matter what you tell them in a crazy attempt at a different outcome. To be clear, sometimes

I took a break for a bit from posting about Strategy, LLMs, ChatGPT, Architecture, Data Privacy, Security, and other random bits to migrate all my writing over to my own site. I found LinkedIn to be mercurial at times with its posting methodologies; finding old posts is wonky, it has

Laker AI added another level (level 8) to Gandalf. He came back as Gandalf The White, but I pwned him again.  Previous pwnage: https://www.tinselai.com/i-pwned-gandalf-and-i-liked-it/ I took out Gandalf The Grey last week and Gandalf The White today. Sauron here I come! :-) Thanks for the fun!

I pwned Gandalf from Laker AI. It is a fantastic little prompt hacking challenge they have built based on OpenAI's GPT. In this you attempt to bypass the controls Lakera's Blue Team have built to prevent prompt injection, protection of confidential information, etc. I got through

OpenAI is now putting out some documentation regarding their approach to GDPR. And it hints at some limitations, which I talk about below. https://help.openai.com/en/articles/7842364-how-chatgpt-and-our-language-models-are-developed Given the design of their LLM, I doubt they could comply with data deletion requests under GDPR and other privacy

This is an amazing and scary application of a GPT. TLDR: Researchers used a GPT to help decode human thoughts from MRI images. This technology is still in its infancy and takes significant resources but is capable of greater than 80% accuracy. The amazing applications of this could be for

As an FYI: Now in OpenAI's ChatGPT settings, you can select "Data Controls" which gives you the following options: 💾 Save Chat - It saves new chats to your history like it has been doing for some time. The trade-off now is if you allow for saving

The site has an interesting infographic showing the relative performance of ChatGPT3.5 vs ChatGPT4 on standard tests like the SAT, GRE and competitive programming. How Smart is ChatGPT?We visualize the performance of ChatGPT in various professional and academic exams, highlighting the improvements of GPT-4 over GPT-3.5.Visual

A few days ago, I posted about defining your audience in a prompt, and in the prompt, I used a simple technique to constrain the response to two paragraphs. https://www.tinselai.com/define-your-audience-for-chatgpt-to-improve-your-response/ In the screenshot, I give an example of rewriting a previous post about GDPR and using

Defining the audience for ChatGPT can improve the response. It is a simple trick, but the impact is impressive. You can simply tell ChatGPT the audience, and it will generate or rewrite content for that audience. I also included the "format:" trick in giving you a two-for-one deal.