maadaa Al News & Datasets: Smart Glasses, Personalized Chatbots, OpenAI and Time magazine
(AI News Weekly: June 25 ~ July 1)
1. Solos Unveils AirGo Vision: World’s First Smart Glasses with GPT-4o Integration
News:
Brilliant Labs has introduced Frame, the world’s first smart glasses with integrated GPT-4 capabilities. These AI-powered glasses feature a camera, microphone, and display, allowing users to interact with an AI assistant through voice commands or by capturing images.
Key points:
- Frame glasses use GPT-4 for real-time information processing and responses
- They include a camera, microphone, and display for user interaction
- The glasses can translate languages, identify objects, and provide contextual information
- Frame is priced at $349 and is available for pre-order
- The device aims to offer a seamless AI assistant experience in daily life
Why it matters:
Implementing Frame smart glasses with GPT-4 integration significantly enhances training datasets. The system can collect diverse and contextual data as users interact with the AI assistant in various real-world scenarios. This data can be used to improve AI models’ understanding of human queries, visual recognition, and language processing in different environments. The continuous stream of real-world interactions will likely contribute to more robust and adaptable AI systems, potentially accelerating the development of more advanced and context-aware AI assistants.
2. Meta Unleashes User-Created AI Chatbots on Instagram: The Dawn of Personalized Digital Companions
News:
Meta has begun testing user-created AI chatbots on Instagram, allowing users to create and interact with personalized AI characters. This new feature aims to enhance user engagement and provide a more interactive experience on the platform.
Key Points:
- Users can create custom AI chatbots with unique personalities
- The feature is currently in testing phase
- Meta is leveraging its large language models for this functionality
- The chatbots can be shared with other users on Instagram
Why It Matters:
By allowing users to create and interact with custom chatbots, Meta gains access to a vast array of diverse conversations and personalities. This user-generated content provides rich, real-world data that can be used to improve language models, making them more natural, context-aware, and capable of understanding various communication styles
3. OpenAI Taps Century-Old Archive to Supercharge ChatGPT’s Knowledge
News:
OpenAI and Time magazine have entered into a multiyear content partnership, granting OpenAI access to Time’s extensive archive of news content spanning over 100 years. This collaboration aims to enhance OpenAI’s AI tools, particularly ChatGPT, by incorporating Time’s trusted journalism into its knowledge base.
Key Points:
-OpenAI gains access to Time’s current and past content from its 101-year-old archive.
- ChatGPT will be able to cite and link back to original Time.com sources when answering user queries.
- Time will have access to OpenAI’s technology to develop new products for its readers.
- The partnership aims to expand global access to accurate and trusted information.
- Time will provide feedback and share practical applications to improve how journalism is delivered via ChatGPT and other OpenAI services.
Why It Matters:
The deal is significant because it indicates a growing trend of collaborations between AI companies and established media outlets. It enhances OpenAI’s training dataset by providing access to high-quality, curated content and addresses concerns about attribution and copyright infringement.
4. Additional News
1. OpenAI introduced CriticGPT, a GPT-4 based AI to find errors in ChatGPT code, enhancing AI output accuracy and reliability.
2. Researchers at the University of Tokyo and Harvard University developed a humanoid robot face with self-healing lab-grown human skin, capable of smiling and making realistic facial expressions.
3. YouTube is negotiating with major record labels to license their music for training AI tools that generate music. They aim to offer lump-sum payments instead of ongoing royalties in exchange for access to music from various artists.
4. Google launched Gemma 2, the latest in its open AI model series, and upgraded its Gemini 1.5 Pro model with a 2M token context window and better coding abilities.
5. SoftBank Group has teamed up with Tempus AI to transform healthcare using AI. Their partnership aims to analyze personal medical data and offer AI-informed treatment recommendations, advancing the integration of AI in healthcare
6. The IBM study on 18,000 international sports fans reveals a growing use of AI and technology for enhanced sports experiences, especially among 18–29 year-olds who prefer personalized, AI-generated content and real-time updates.
5. maadaa.ai’s Shared Open & Commercial Datasets
Open Dataset 1: BDD100K (Berkeley Deep Drive)
Description: BDD100K is one of the largest and most diverse driving datasets available. It contains 100,000 videos that have been split into training, validation, and test sets, along with annotations for various road objects, lane markings, and weather conditions. This dataset is particularly useful for developing and testing autonomous driving technologies.
URL: http://bdd-data.berkeley.edu/
Open Dataset 2: Visual Genome
Description: Visual Genome is an extensive dataset that includes over 108,000 images with 3.8 million object instances. This dataset is designed for object detection and scene description tasks, featuring dense annotations of objects, attributes, and relationships between objects, making it suitable for advancing research in detailed visual understanding and interaction.
URL: https://homes.cs.washington.edu/~ranjay/visualgenome/index.html
Open Dataset 3: Fashion & E-commerce Open Dataset
We’re excited to launch our “Fashion & e-Commerce Open Datasets” for AI innovation in fashion and e-commerce. The dataset offers 24 real-world scenarios and 33 sub-datasets with detailed annotations, supporting applications like object detection, pose estimation, personalized recommendations, virtual fittings, and product recognition. This versatile dataset aims to inspire innovative solutions.
URL: https://maadaa.ai/datasets/OpenDatasetDetail/Fashion-&-E-commerce-Open-Dataset
Commercial Dataset 1: Multi-modal Generative AI Large Datasets — Licensed
Over 300 Million Image-Text pairs: covers an extensive range of high-resolution professional shooting images, including humans, animals, scenes, photography, and vector images.
More than 6 Million Video-Text pairs: provides rich text descriptions of characters, scenes, relationships, actions, etc.
More than 2 million e-books and 15,000 journals: enriching the dataset with literary and academic depth.
Genuine Media Reporting Data: Incorporating text data from major domestic media outlets ensures the inclusion of current and relevant content.
