Notes on the Inspire session on „Client AI at Scale“: Minor quality improvements today, more energy efficient (screenshots #1, #3) Upcoming Intel Meteor Lake to include equivalent to Google TPU, Apple Bionic Engine: „VPU“ (screenshot #2) Some examples of Frameworks given, including DirectML (screenshot #4). Questions about local Llama 2,... Read more 20 Jul 2023 - less than 1 minute read
How the Facebook LLaMA 2 announcement was presented and contextualized on the side of Microsoft at their Inspire event: it started with a lengthy love letter to OpenAI (screenshot #1) and transitioned to elaborating on their enduring engagement in Open Source (screenshot #2): We’re one of the largest contributors to Open Source and when it c... Read more 18 Jul 2023 - 1 minute read
One of the other things that transpired at the Inspire keynote was Bing Enterprise Chat, which addresses data privacy concerns and requirements by businesses. “Not saved” in screenshot #3 means that prompts are not saved, which is different from chat.openai.com. Also, Visual Search in Bing Chat includes the generate-code-from-scribble that ... Read more 18 Jul 2023 - less than 1 minute read
Reports about “AI getting dumber” are making international headlines in daily newspapers and professional outlets alike. I’m convinced that this is not what has happened. Rather, it’s a bugfix gone wrong. I liked to say that no matter how low-effort your inquiry is, GPT-4 will always come back with a good answer (🇩🇪 “Du kannst da unqualifiziert ... Read more 12 Jul 2023 - 1 minute read
Quick reality check on Quivr, a tool to: Dump all your files and chat with it using your Generative AI Second Brain using LLMs (GPT 3.5/4, Private, Anthropic, VertexAI) & Embeddings There has been a seemingly endless stream of such projects over the last several months. The core concept is always the same, but the newest incarnations di... Read more 12 Jul 2023 - 2 minute read