- The Bright Journey with AI
- Posts
- Apple's On-Device LLMs | Sanctuary Seventh Gen Robot | Nvidia Acquires Run:ai Orchestration Provider
Apple's On-Device LLMs | Sanctuary Seventh Gen Robot | Nvidia Acquires Run:ai Orchestration Provider
The Bright Journey with AI - April 26th 2024
📣 Announcement 📣
I’m making a few changes to the content strategy at Bright Journey AI.
There is no end to the torrent of news and updates coming out of AI and it shows no signs of stopping. That said, three editions a week has become more and more challenging and is taking time away from me developing more original content. For this reason I’m going to change the release schedule to two editions per week, every Tuesday & Friday starting next week.
The Bright Journey with AI will continue, with more original content demonstrating practical uses of AI. Please continue to share with your friends and co-workers so we can grow this community of ours.
Thanks - Mark
📰 News 📰
Apple's OpenELM: New On-Device AI Models
Apple has recently launched OpenELM, a new series of open-source large language models (LLMs) tailored for on-device operations, such as text generation. Available on Hugging Face, OpenELM consists of eight different models, ranging from 270 million to 3 billion parameters. These models are designed for both pre-training and instruction tuning to enhance their responsiveness to specific user prompts. Notably, these models can operate independently on devices without needing cloud connectivity. Apple’s release under a sample code license allows broad utilization and modification, encouraging adoption and innovation within the open-source community. This initiative marks a significant step by Apple towards matching the on-device AI capabilities of competitors like Microsoft's Phi-3 Mini.
Read More At: VentureBeat
Image Source [SantuaryAI]
Sanctuary AI Unveils Advanced Humanoid Robot
Sanctuary AI has introduced the seventh-generation Phoenix humanoid robot, boasting faster learning and lower costs. Unlike its predecessors, this model emphasizes upper body functionality, efficiently learning tasks such as sorting products within 24 hours. These enhancements make the Phoenix significantly more analogous to human workers in terms of task performance, marking a significant step towards general-purpose AI robotics. This latest model also features improvements in motion range, weight, and material cost.
Read More At: TechCrunch
Nvidia Acquires Run:ai to Enhance AI Kubernetes Orchestration
Nvidia has announced the acquisition of Run:ai, an AI-centric Kubernetes orchestration provider, aiming to enhance GPU cluster efficiencies. Run:ai offers a central user interface similar to solutions like RedHat's OpenShift, providing management tools for Kubernetes environments. This acquisition, valued at around $700 million, integrates Run:ai's technology into Nvidia's platforms, supporting a range of its computing solutions including DGX systems. Run:ai's capabilities in workload scheduling and GPU acceleration will continue to be available, with no immediate changes for existing customers.
Read More At: The Register
🧠 Research 🧠
How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites
The paper introduces InternVL 1.5, an open-source multimodal large language model (MLLM), aiming to bridge the capability divide between open-source and commercial models in multimodal understanding. Three major improvements are highlighted: (1) Strong Vision Encoder, enhancing visual capabilities through continuous learning strategies, (2) Dynamic High-Resolution, adjusting input images up to 4K resolution, and (3) High-Quality Bilingual Dataset, enabling better performance in OCR and Chinese-related tasks. InternVL 1.5 demonstrates competitive results, outperforming several proprietary models in 8 out of 18 benchmarks. Key limitations include reliance on open-source datasets and potential resolution limitations affecting performance on some tasks. Future research could focus on expanding multilingual capabilities and integrating additional sensory modalities to further enhance the model's robustness and applicability.
Read the paper on arXiv
🔨 AI Powered Tools 🔨
A run down of the latest SaaS products and services which leverage AI to help you take back some time.
Adobe Firefly - Leverages AI to assist creators in generating images and text, enhancing their ability to produce original and unique content efficiently.
SiteGPT - AI-powered customer support tool tailored for websites across various industries. It enhances customer support by offering a personalized chatbot trained on a website's specific content.
Rask.AI - Task management and automation tool that incorporates AI to streamline operations for businesses. It offers AI-driven features to automate repetitive tasks, schedule resources, and optimize workflows, making it ideal for teams looking to enhance productivity.
🗞️ Other News Rollup 🗞️
School AI Scandal - Athletic director arrested for framing principal with AI voice synthesis, sparking community outrage. Voice-cloning technology involved.
AI Technological Leap - Nvidia CEO delivers advanced DGX H200 to OpenAI, enhancing research capabilities and catalysing AI industry advancements.
Data Monetization Innovation - Carv raises $10M Series A to empower users in gaming and AI sectors. Valued at $40M.
AI Practical Applications - AI history, defining intelligence, language impact, practical ML applications in networking, focusing on usefulness over hype.
AI-Powered Math Guarantee - Edia's AI ensures better math outcomes in school districts within a year or money back. Revolutionary approach.
AI App Development - Cohere's toolkit streamlines AI app building, enhancing scalability and reducing development time.
AI Data Development Upgrade - Snorkel Flow integrates with Gemini AI and Llama 3, enhancing data labelling efficiency and AI model accuracy.
AI-Driven Self-Healing Security - Amplifier Security uses AI for self-healing security, personalized engagements, and real-time insights to revolutionize enterprise security.
LLMs' Learning Advancements - DeepMind study shows many-shot in-context learning boosts large language models' performance without fine-tuning, excelling in tasks.
Apple's Ethical AI - Apple leads in ethical AI training, licensing major news publications' works, prioritizing privacy and legality standards.
Samsung's AI Memory Breakthrough - Samsung unveils 10.7Gbps LPDDR5X DRAM, enhancing AI performance with increased speed, capacity, and power efficiency.
AI Video Enhancement Breakthrough - Adobe's VideoGigaGAN AI upscales video by 8x, enhancing sharpness without artifacts, revolutionizing video quality.
🎶 Prompts 🎶
Prompts used to generate some of this issues images. Unless otherwise stated all images are using Dall-E and ChatGPT Plus.
|
|
|
Reply