1. Code Llama Unveiled: A Breakthrough in AI-Powered Coding Assistance
Code Llama, released on January 29, 2024, is a state-of-the-art large language model (LLM) specialized in coding, available in three versions - foundational, Python-specialized, and instruction-focused - all free for research and commercial use. Built on Llama 2, Code Llama enhances coding capabilities, supports multiple programming languages, and outperforms other publicly available LLMs in code tasks, aiming to improve developer workflows and lower entry barriers for coding learners.
Source: Meta
2. Mobile ALOHA Enhances Robotic Dexterity in Bimanual Tasks with Whole-Body Teleoperation System
The Mobile ALOHA project, led by Zipeng Fu, Tony Z. Zhao, and advised by Chelsea Finn, develops a low-cost, whole-body teleoperation system for imitating mobile, bimanual manipulation tasks. By augmenting the ALOHA system with a mobile base and using data from human demonstrations, the system achieves up to 90% success in complex tasks like cooking and cabinet organization through supervised behavior cloning and co-training with static datasets.
Source:GitHub
3. OpenAI Unveils GPT Store Featuring Custom ChatGPT Versions and New Revenue Program for Builders
OpenAI has launched the GPT Store for ChatGPT Plus, Team, and Enterprise users, providing access to over 3 million custom versions of ChatGPT, including those for various applications such as DALL·E, education, and programming. The store also introduces a revenue program for builders based on user engagement, along with new features for managing GPTs within Team and Enterprise plans.
Source: OpenAI
4. Google Lumiere: Revolutionizing Video Synthesis with Space-Time U-Net Architecture
Lumiere is a text-to-video diffusion model that generates realistic, diverse, and coherent motion videos through a novel Space-Time U-Net architecture, capable of synthesizing the entire temporal duration of a video in a single pass. This approach, which contrasts with traditional methods that struggle with global temporal consistency, leverages spatial and temporal down- and up-sampling, integrating a pre-trained text-to-image model to achieve state-of-the-art text-to-video generation and support various content creation and video editing applications.
Source: Lumiere Video
5. Adobe Premiere Pro Boosts Productivity with New AI Audio Features and Future Multimodal Capabilities
Adobe Premiere Pro introduces new AI features to enhance audio clip quality and workflow efficiency, including automatic audio clip categorization and the Enhanced Speech feature for background noise reduction and speech improvement. Future updates may include more multimodal AI capabilities, like text-to-speech, voice cloning, and generative video creation.
Source: Adobe
6. FDA Clears First AI Medical Device for Skin Cancer Detection
The FDA has approved DermaSensor, the first AI-powered medical device capable of detecting the three most common types of skin cancer: melanoma, basal cell carcinoma, and squamous cell carcinoma. This revolutionary device uses non-invasive technology to provide real-time cancer risk assessments, aiming to improve early detection and enhance diagnostic accuracy across all skin types.
Source: DermaSensor
7. Google AI Shows Promise in Medical Diagnoses and Empathy Over Human Doctors
An AI system developed by Google, using a LLM for medical interviews, has shown the ability to match or surpass human doctors in diagnosing respiratory and cardiovascular conditions among others, based on simulated patients’ medical histories, with a noted higher empathy rating. This experimental chatbot, named Articulate Medical Intelligence Explorer (AMIE), emphasizes potential in democratizing healthcare, though it remains untested on real health conditions and is advised to complement rather than replace physician interactions.
Source: Nature
8. OpenAI Policy Update Opens Door to Military Applications
OpenAI has revised its policy to potentially allow military applications of its technologies, removing previous restrictions against “military and warfare” uses, as highlighted by The Intercept. This change aims to provide clarity and enable discussions on national security use cases that align with OpenAI’s mission, including collaboration with DARPA on cybersecurity tools for securing critical infrastructure.
Source: TechCrunch
9. Robot Surpasses Human Speed in Braille Reading with AI Sensor
Researchers at the University of Cambridge have developed a robotic sensor, using artificial intelligence and machine learning, that can read braille at speeds of 315 words per minute, roughly twice as fast as the average human reader, with close to 90% accuracy. This innovation, which is not intended as assistive technology but rather as a step towards creating robotic hands or prosthetics with human-like sensitivity, represents a significant advance in robotic tactile sensing and has potential applications beyond braille reading.
Source: EurekaLert
10. Introducing TinyLlama A Groundbreaking 1.1B Language Model Launched This Month
TinyLlama is a 1.1B parameter language model pre-trained on 1 trillion tokens for about 3 epochs, based on Llama 2’s architecture and tokenizer, enhanced by open-source contributions like FlashAttention for improved efficiency. It outperforms other open-source models of similar size and has made its checkpoints and code publicly available on GitHub.
Source: Arxiv