More Nail-Biting Drama at OpenAI??
++ Google Veteran Joins OpenAI, Apple's Eye Tracking technology, Zoho's Bold Move, LangChain & Hugging Face Partner & more
🚀 AI Breakthroughs
Apple Introduces Enhanced Accessibility Features Including Eye Tracking and Music Haptics
Apple unveils new accessibility features for iPad and iPhone, including innovative Eye Tracking technology for users with physical disabilities.
Music Haptics introduced, enabling deaf or hard of hearing users to experience music through tactile sensations via iPhone's Taptic Engine.
Vocal Shortcuts feature announced, allowing users to perform complex tasks on their devices through custom voice commands.
ElevenLabs Utilizes Google's Veo for Innovative AI Music Video Creation
ElevenLabs uses Google's Veo engine to create a new AI-generated music video.
The video features a dark, synthy audio vibe from a neon-drenched AI video, innovatively translated into music.
Despite some artistic license with the visuals, the project showcases the impressive integration of multiple AI technologies to produce music.
Unitree Robotics Launches G1 Humanoid, Redefining Affordability in Robotics
Unitree Robotics launched the new G1 humanoid at ICRA 2024, priced at an entry-level $16,000.
The G1 humanoid, designed to be the size of an average eight-year-old, is optimized for cost-efficiency and ease of maintenance.
Targeted primarily for R&D, the G1 humanoid features advanced sensors like 3D LiDAR and fast walking speeds, ideal for university lab environments.
OpenAI Enhances ChatGPT: Direct Uploads from Google Drive and Customizable Data Visualizations Announced
ChatGPT now enables direct uploads from Google Drive and OneDrive, streamlining data analysis without the need to download and re-upload files.
Enhancements in ChatGPT include improved natural language understanding for running Python analytics, and the ability to interact with and customize charts and tables.
OpenAI ensures user data privacy by confirming that data uploaded by Enterprise and Teams on ChatGPT will not be used to train AI models, with opt-out options available for Plus subscribers.
Zoho Invests $700 Million in Chipmaking Initiative, Requests Government Support
Indian software company Zoho announces a $700 million investment to enter the chipmaking sector.
Zoho, competing globally against firms like Microsoft, aims to diversify into manufacturing compound semiconductors.
The proposal is under review by India's IT ministry, which seeks more details on potential clients and market strategy.
Introducing Langchain_Huggingface: A New Partnership Enhancing LangChain's Capabilities
Announcing the launch of langchain_huggingface, a new partner package jointly maintained by Hugging Face and LangChain.
Designed for the community, by the community, this package addresses the challenge of deprecated classes due to the lack of insider perspectives.
The partnership promises seamless integration and continual improvements, enhancing the use of Hugging Face models within the LangChain ecosystem.
OpenAI Recruits Google Veteran to Helm Development of New Search Engine Alternative
OpenAI hires Shivakumar Venkataraman, a 21-year Google veteran, as vice president to lead the development of its own search engine.
With expertise gained from leading Google's search ads business and blockchain division, Venkataraman's knowledge will be crucial for OpenAI's new search platform.
Venkataraman's extensive background includes a PhD in Computer Science from the University of Wisconsin-Madison and roles at Hewlett-Packard Labs and IBM.
Reddit Partners with OpenAI to Integrate ChatGPT with Platform Content
Reddit and OpenAI's new partnership integrates Reddit content with ChatGPT, enhancing OpenAI's service offerings.
The collaboration also includes OpenAI acting as an advertising partner on Reddit, potentially diversifying Reddit's income streams.
This partnership follows Reddit's recent agreement with Alphabet for content usage in AI model training, valued at around $60 million annually.
⚖️ AI Ethics:
Voice Actors Sue AI Firm for Cloning Their Voices Without Consent
While driving to a doctor's appointment, voice actors Paul Skye Lehrman and Linnea Sage discover a podcast using A.I. to mimic Lehrman’s voice without consent.
Shocked by the unauthorized use of their voices, Linnea Sage and Paul Skye Lehrman are now suing the A.I. company responsible for voice cloning.
The company involved denies any wrongdoing in the replication of the couple’s voices, leading to a legal confrontation.
OpenAI Researcher Resigns, Cites Safety Concerns Over Product Focus
OpenAI researcher Jan Leike resigns, criticizing the company for prioritizing "shiny products" over safety.
Following his departure, Jan Leike reveals OpenAI disbanded the Superalignment team, raising concerns about AI safety.
Jan Leike's exit underscores growing internal conflicts at OpenAI as it pushes development of potentially dangerous AI technologies.
🎓AI Academia:
UniRAG: Enhancing Multi-Modal Large Language Models Through Universal Retrieval
UniRAG is introduced as a model-agnostic technique enhancing Multi-Modal Large Language Models (MM-LLMs) by utilizing retrieved information during inference.
Evaluation on the MSCOCO dataset reveals significant improvements in generation quality for both proprietary (like GPT4, Gemini-Pro) and open-source (like Llava, LaVIT, Emu2) models when using relevant retrieved information.
Despite common beliefs, retrieval augmentation benefits MM-LLMs not only in handling uncommon entities but also significantly enhances performance with common entities.
Evaluating Large Language Models in Natural Language Generation Tasks
This study is the first to systematically evaluate the performance of large language models (LLMs) specifically in natural language generation (NLG) tasks.
Researchers Xuanfan Ni and Piji Li from Nanjing University of Aeronautics and Astronautics focus on prominent models like ChatGPT, T5, and LLaMA for testing across both English and Chinese datasets in Dialogue Generation and Text Summarization.
The evaluation introduces a common framework using input templates and specific post-processing strategies to ensure consistency and reliability in comparing model outputs.
Assessing the Impact of Retrieval-Augmented Large Language Models on Biomedical NLP Tasks
Researchers systematically examined retrieval-augmented large language models (RALs) across five critical biomedical NLP tasks.
The study highlighted challenges with RALs in handling unlabeled and counterfactual information and maintaining negative awareness.
Despite these challenges, RALs showed enhanced performance and some level of counterfactual robustness on most evaluated biomedical datasets.
P.S.: We curate this AI newsletter daily for free. Your support keeps us motivated. If you find it valuable, please share it with your friends using the button below!