• AI Breakfast
  • Posts
  • AI Device Reads Your Sweat to Tell You About Your Heart's Health

AI Device Reads Your Sweat to Tell You About Your Heart's Health

Author

AI Breakfast
July 03, 2023

Good morning. It’s Monday, July 3rd.

Did you know: AI voice cloning is reviving singers from the past. Listen here to an incredibly realistic Frank Sinatra singing Coolio’s Gangster’s Paradise.

In today’s email:

  • YOPI's heart-monitoring AI device

  • EU firms challenge AI rules

  • AI creates rapid CPU

  • Humane's AI wearable revealed

  • MIT's AI for chart interpretation

  • Red Ventures' AI content plan

  • US-China AI clash at UNESCO

  • AI aids fruit fly, human genetics study

  • Audi, Volkswagen use AI for quality control

  • G/O Media staff unhappy with AI content

  • Japan leans towards softer AI regulations

  • 5 New AI Tools

  • Latest AI Research Papers

You read. We listen. Let us know what you think of this edition by replying to this email, or DM us on Twitter.

Today’s edition is brought to you by:

Guidde AI is a GPT-powered tool that creates video documentation 11x faster.

Simply click capture on our browser extension and we will automatically generate step-by-step video guides complete with visuals, voiceover and call to actions.

Share or embed your Guidde anywhere with our smart copy and turn your boring documentation into stunning visual guides.

Today’s trending AI news stories

AI-Powered Wearable by YOPI Technologies Monitors Heart Health Through Sweat Analysis: Israeli tech firm YOPI Technologies has launched YOPI, an AI-enhanced wearable that analyzes sweat to identify early signs of heart issues. The device, worn on the arm, features a unique sensor for real-time monitoring of workout intensity and personalized physiological modifications. View YOPI’s website here.

European firms criticize upcoming EU AI regulations in an open letter: More than 150 top executives from European businesses such as Renault, Heineken, Airbus, and Siemens, have openly criticized the EU AI Act in an open letter. They argue that the proposed regulations risk undermining European competitiveness and technological self-reliance, highlighting concerns about overly stringent rules, particularly for generative AI systems, and the potential for disproportionate compliance costs and liability risks.

AI Designs Self-Optimizing CPU in Record-Breaking Time: Chinese scientists have leveraged AI to create a functional RISC-V CPU, with a performance comparable to an Intel i486SX, in less than five hours. This remarkable achievement signifies a major step forward in the development of self-evolving machines, showcasing the AI's ability to accelerate the design process and discover detailed architecture optimizations. The researchers published a paper on the discovery.

Humane Reveals AI-Powered Wearable Device, AI Pin: Humane, a discreet hardware startup formed by former Apple employees, has disclosed information about its product, the Humane AiPin. This wearable device boasts a projected display and AI functionalities, capable of tasks like email summarization, language translation, object recognition, and interface projection. Join the waitlist here.

MIT Presents AI Tool to Improve Chart Interpretation and Accessibility: Researchers at MIT have created VisText, an AI-powered tool that enriches chart interpretation and accessibility by generating detailed, adaptable captions. The tool, trained on a dataset of over 12,000 charts and corresponding captions, has shown superior performance over other auto-captioning systems, addressing the varied needs of users with different abilities.

Red Ventures Aims to Power Content Creation with AI: Red Ventures, the parent organization of CNET, plans to significantly incorporate AI into their operations. The company plans to transition from merely using AI to becoming an AI-centered organization, with a focus on boosting productivity and tailoring content.

US-China Rivalry Intensifies Over AI in UNESCO Policymaking: The US's re-entry into UNESCO, with China opposing it, signifies a new chapter in their rivalry within the realm of AI. The US voices concerns about China's attempts to establish global standards and restrict its AI capabilities, underlining the escalating technological competition between the two nations.

New AI System Decodes Fruit Fly Behaviors, Holds Promise for Human Genetics Research: Researchers at Tulane University have developed an AI system, MAFDA, that tracks and interprets the behaviors of fruit flies. The system's potential to accelerate human genetics research is significant given the large genetic overlap between humans and fruit flies.

Audi and Volkswagen Turn to AI for Quality Control: Audi and Volkswagen are implementing AI for quality control within their manufacturing plants. The companies will use AI to automate the examination of spot welds, the AI system enhances efficiency and focuses employees on irregularities. After successful trials, the AI system will be deployed across other facilities.

G/O Media Staff Reacts to AI Content Announcement: G/O Media, the parent company of Gizmodo and Kotaku, is facing backlash from its staff following the announcement of a trial for AI-generated content. The move has raised questions about the quality and reliability of AI-produced articles and sparked fears about the future of journalism at the company. Earlier this year, Buzzfeed’s stock jumped 150% after announcing AI will write a significant amount of online content for the company.

Japan Prefers Less Restrictive AI Rules Compared to EU: Japan is leaning towards a more relaxed AI regulatory approach than the EU, in an effort to stimulate economic growth and gain a competitive edge in advanced chip technologies. This stance may undermine the EU's endeavor to position its regulations as the global standard. The chair of the Japan Deep Learning Association, Prof. Yutaka Matsuo, criticizes the EU's regulatory focus on corporate accountability at the expense of fostering innovation.

🎧 Did you know AI Breakfast has a podcast read by a human? Join AI Breakfast team member Luke (an actual AI researcher!) as he breaks down the week’s AI news, tools, and research: Listen here

5 new AI-powered tools from around the web

Featured Tool:

Cognitiev PRO is the ultimate AI sidekick offering various features such as 20+ chat modes, secure and private functionality, coding improvement, book/movie explainers, art/code analysis, and more. Interact with AI through WhatsApp or Telegram.

Magicast.ai is an AI-driven platform that produces engaging audio content across various topics. It prioritizes accessibility by providing visually impaired users with audio versions of online text. It serves as a hub for high-quality, on-demand podcasts.

Podium is an AI copywriter for podcasters, providing tools to streamline content creation. It offers show notes, chapters, transcripts, highlight clips, and social media posts, saving time and money while reaching broader audiences.

Checkmyidea-IA is a service powered by AI that aids entrepreneurs in validating their side business ideas. It evaluates the idea’s viability by considering market demand, competition, and current trends. The service offers guidance on launching the idea and acquiring users, along with special discounts for the grand production launch.

RaqiAI is a fast, design-centric, and data-driven creative agency that empowers businesses to succeed. With services in design, marketing, and coding, RaqiAI delivers high-quality work, speedy results, and personalized solutions.

arXiv is a free online library where scientists share their research papers before they are published. Here are the top AI papers for today.

The paper introduces PACGen, a personalized and controllable text-to-image generation model. It addresses the challenges of controlling object identity, size, and location in generative models. PACGen combines data augmentation techniques with pre-trained adapter layers from GLIGEN to achieve high fidelity and localization control. The proposed model enables users to generate personalized objects in any scene with fine-grained control. The paper compares PACGen with existing models and demonstrates its effectiveness in generating realistic and personalized images. The approach has potential applications in art, entertainment, advertising, design, and more.

Magic 123 is a two-stage pipeline for high-quality 3D object generation from a single unposed image. It combines both 2D and 3D priors to optimize a neural radiance field in the coarse stage and a memory-efficient differentiable mesh representation in the fine stage. By balancing the trade-off between exploration and exploitation of geometry, Magic 123 achieves realistic and detailed 3D reconstructions. The pipeline leverages reference view supervision, novel views guided by diffusion priors, textual inversion, and monocular depth regularization. Extensive experiments demonstrate its superior performance over previous image-to-3D techniques. The code, models, and generated 3D assets are available for use.

The paper explores the use of OpenAI’s ChatGPT for robotics applications. It presents a strategy that combines design principles for prompt engineering and a high-level function library to enable ChatGPT to adapt to different robotic tasks. The effectiveness of different prompt engineering techniques and dialog strategies is evaluated for various robotic tasks, ranging from basic reasoning to complex domains. The study shows that ChatGPT can effectively solve these tasks and interact with users primarily through natural language instructions. Additionally, the paper introduces PromptCraft, an open-source research tool, and a robotics simulator with ChatGPT integration to facilitate the use of ChatGPT for robotics.

Proposed Ranking Optimization (PRO) is proposed as an alternative to reinforcement learning from human feedback. (RLHF) for aligning LLMs with human values. PRO extends the pairwise Bradley-Terry comparison to accommodate preference rankings of any length. By iteratively contrasting the likelihood of generating responses, PRO instructs the LLM to prioritize the best response while progressively ranking the remaining responses. Experimental results show that PRO outperforms existing alignment algorithms and achieves comparable results to ChatGPT and human responses. PRO offers advantages such as utilizing longer ranking sequences, self-bootstrapping, and differentiable optimization, addressing the limitations of RLHF.

ArrayBot is a system that uses reinforcement learning to manipulate objects. It has a 16x16 array of pillars with tactile sensors. By reshaping the action space and focusing on local actions and low frequencies, ArrayBot’s RL agents learn to manipulate objects using only tactile work with unseen objects in simulations and can be transferred to the physical robot without needing to randomize the environment. ArrayBot showcases the potential of reinforcement learning for distributed manipulation.

Thank you for reading today’s edition.

Your feedback is valuable.


Respond to this email and tell us how you think we could add more value to this newsletter.

Attending Ai4 this year? We will be!