?? KZEE's AI Weekly - May 27 2024 Update ??
1. Cohere's Latest LLMs and Collaborations
?? Aya 23 — Cohere introduced Aya 23, a family of state-of-the-art, multilingual, generative large language research models covering 101 different languages.
?? Command R+ — This is the most powerful model in the Command R family, designed to enhance enterprise workflows and applications with improved performance and response accuracy.
?? Open-Source — Aya 23 is open-source, allowing for collaboration and experimentation by researchers and developers globally.
Collaborations & Integrations
?? Microsoft Azure — Command R+ is integrated into Microsoft Azure's AI model catalogue, enhancing its accessibility to Azure customers.
?? Oracle Cloud — While initially available on Azure, Command R+ will soon be accessible on Oracle Cloud Infrastructure and other cloud platforms.
?? Article - https://cohere.com/blog/aya23
2. Elon Musk's xAI recent funding round
?? Funding — xAI raised $6 billion in series B funding.
?? Valuation — The company is now valued at $24 billion post-money.
?? Investors — Backers include Andreessen Horowitz and Sequoia Capital.
?? Series — The funding round was part of series B funding.
?? Pre-Money Valuation — xAI's pre-money valuation was $18 billion.
?? Article News - https://tinyurl.com/2cne5gnr
3. Elon Musk is planning to build the largest supercomputer for his xAI startup
??? Scale — The planned supercomputer for xAI is set to be at least four times the size of the biggest GPU clusters existing today, utilizing 100,000 Nvidia chips.
?? Purpose — It is intended to support the development of xAI's artificial intelligence projects, particularly the enhancement of its chatbot named Grok.
?? Technology — The supercomputer will string together 100,000 Nvidia chips, specifically Nvidia's flagship H100 graphics processing units, to achieve its computational power.
?? News Article - https://tinyurl.com/28ha6xrk
4. Understanding Meta's Chameleon AI Model
?? What is Chameleon? — Chameleon is a new family of AI models developed by Meta's AI research lab, designed to understand and generate both text and images in a unified token space.
?? Early Fusion Approach — Unlike traditional AI models that use a 'late fusion' approach, Chameleon employs an 'early-fusion token-based mixed-modal' architecture, integrating text, images, and code from the beginning of the processing pipeline.
?? Unique Capabilities — Chameleon can seamlessly reason over and generate interleaved image and text sequences, making it highly versatile for tasks that require a deep understanding of both visual and textual information.
领英推荐
?? Training and Scale — The model is trained in two stages on a dataset containing 4.4 trillion tokens of text, image-text pairs, and sequences of text and images interleaved, using a significant amount of computational resources.
?? Article Read - https://venturebeat.com/ai/meta-introduces-chameleon-a-state-of-the-art-multimodal-model/
5. Google's AI Overview feature faced significant backlash for providing misleading and bizarre answers
?? Nature of Errors — Google's AI Overview feature generated misleading and sometimes bizarre responses, such as recommending people to eat rocks for nutrients and suggesting putting glue on pizza.
Funny yet bizarre ??
??? Recommendation to Eat Rocks — The AI Overview feature advised people to eat 'at least one small rock per day' for nutrients, a suggestion that is obviously unsafe and scientifically unsound.
??Glue on Pizza — Another concerning recommendation was to put glue on pizza to help cheese stick better, which poses a serious health risk if followed.
?? Public and Expert Responses — The errors sparked a significant backlash online, with screenshots of the erroneous answers widely shared. Experts expressed concerns about the feature perpetuating bias and misinformation, as well as potentially endangering people seeking help in emergencies.
?? Google's Action — Google acknowledged the issue and is reportedly working to manually disable AI Overviews for specific searches as the problematic responses circulated on social media.
6. OpenAI's Content Licensing Deal with News Corp
?? Deal Overview — OpenAI and News Corp have entered a multi-year agreement allowing OpenAI to access and display content from News Corp's major news publications.
?? Purpose — The partnership aims to enhance AI products with high-quality journalism and uphold journalism standards.
?? Deal Value — The deal is reported to be worth over $250 million in the next five years, including cash and credits for the use of OpenAI technology.
??? Content Access — OpenAI will receive access to current and archived content from News Corp's publications, including The Wall Street Journal, The New York Post, and others.
?? Expertise Sharing — News Corp will provide journalistic expertise to help maintain high journalism standards in OpenAI's offerings.
?? Link - https://openai.com/index/news-corp-and-openai-sign-landmark-multi-year-global-partnership/
7. BrainBridge's AI-Driven Robotic System for Head Transplants
?? What — BrainBridge, a neuroscience and biomedical engineering startup, has introduced a concept for the world's first head transplant system.
?? How — The system utilizes cutting-edge robotics and artificial intelligence to perform head and face transplantation procedures.
?? Impact — BrainBridge's technology aims to achieve successful head transplants with improved outcomes and faster recoveries, potentially benefiting patients with untreatable conditions.
?? Timeline — The company plans to conduct the first procedure within eight years, highlighting a significant advancement in medical science
How many years ago was a concept of putting a chip in human's mind a sci-fi concept, yet Neuralink is here!
That's a wrap for the AI weekly for now, thank you for reading. See you next week! ??