Meta AI has released Code Llama 70B, an advanced code generation model that can write and modify code in multiple programming languages like Python, C++, Java, and PHP. Built on the Llama 2 model, Code Llama 70B has been trained on 500 billion tokens of code, making it highly capable of handling complex coding tasks. It features a large context window of 100,000 tokens for processing and generating lengthy and intricate code.
The model includes a special variant, CodeLlama-70B-Instruct, which excels in understanding natural language instructions and scored 67.8 on HumanEval, surpassing many other models and demonstrating high functional correctness in coding. Another variant, CodeLlama-70B-Python, is optimized for Python and trained on an extra 100 billion tokens of Python code.
Code Llama 70B is available for free under a permissive license, accessible through platforms like Hugging Face, PyTorch, TensorFlow, and Jupyter Notebook. Mark Zuckerberg highlighted the model’s significance in coding and AI development, emphasizing its role in future AI models like Llama 3. Code Llama 70B aims to make software development more efficient and accessible, offering a tool for code creation, improvement, and learning.
Prophetic, a neurotechnology startup has developed an AI model named Morpheus-1, designed to induce and stabilize lucid dreams. Lucid dreaming, a state where the dreamer is aware they are dreaming and can control the dream, was popularized by the film “Inception.”
Morpheus-1 utilizes ultrasound holograms, generated based on the brain’s current state, to achieve this. The technology will be incorporated into a headband product called The Halo, set for a beta release in the spring. This method of brain stimulation aims to offer a safe and controlled way to experience lucid dreams, which have been suggested to have various psychological benefits, including treating nightmares and PTSD.
China has approved over 40 AI models for public use in the first six months since starting the approval process, aiming to compete with the U.S. in AI development. Last week, 14 large language models (LLMs) received approval, including those from companies like Xiaomi Corp, 4Paradigm, and 01-AI. This is part of China’s effort to regulate and control AI technology development. The approval process began in August, with companies such as Baidu, Alibaba, and ByteDance among the first to receive approval.
Although the government has not made the full list of approved companies public, the Securities Times reported the approval of more than 40 AI models. This move comes as Chinese companies intensify their AI development efforts, following the global impact of OpenAI’s ChatGPT. China had 130 LLMs, representing 40% of the global total, just behind the U.S. Baidu’s Ernie Bot, a ChatGPT-like chatbot, has already attracted over 100 million users.
ChatGPT Tips and Tricks – Part 1: Adjusting Conversation Width, Disabling Auto-Scroll, and more…
Hope you enjoyed today’s newsletter
⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.