Are you able to write a full essay in one shot? No. So why do we expect a more limited intelligence to be able to do so? When using LLM to solve problems we often think about a one-shot process -I/O prompting, but there are other alternatives 👇 source: arxiv.org/pdf/2305.10601

infinite RPG ♾⚔️ x.com/_philschmid/st…

.Exo Labs transforms your iOS devices into a local AI cluster x.com/exolabs_/statu…

Running Llama 3.1 7B 15k context window is very fast on an M3 chip -see the x1 speed in the video. I use it (for free) to summarize research papers into a single sentence, for this, I need to run multiple inferences per paper on each section because of the context window

Unless we have a paradigm shift in AI, such as symbolic reasoning, no one will ever lock their money in a contract, solely written by an LLM which essentially is an excellent auto completion model. Unless, maybe it's audited by humans. x.com/EMostaque/stat…

⌨️ nice LLMops project Imagine you have a service LLM like GPT, Claude, and want your app to switch to a local small LLM (failure of service, privacy, version control). This tool helps you transition by collecting data and fine-tuning a local LLM 🤌 github.com/deep-diver/lla…

📜 bit-sized paper by Stanford University... ...on a cybersecurity benchmark evaluating LLMs on professional-level CTFs, task decomposition, and exploit execution. We need such one for web3 📷👀 arxiv.org/abs/2408.08926

🤘I am thrilled to share with you that I'm joining Immunefi as their new senior machine learning engineer I'm honored to have the opportunity to contribute meaningfully to the best ecosystem ever alongside incredible talents 🤝