PuzzleAvatar: Assembling D Avatars from Personal Albums
Summarized by: Sophia Bennett [
arxiv.org]
PuzzleAvatar is a model designed to create highly accurate 3D avatars from personal “Outfit Of The Day” (OOTD) photo collections. Unlike traditional methods that require controlled environments, PuzzleAvatar can handle casual photos with varied poses, camera angles, lighting, and backgrounds, as long as the outfit, hairstyle, and accessories remain consistent.
The model works by breaking down these photos into separate tokens representing different aspects such as appearance, identity, garments, hairstyles, and accessories. These tokens are then used to fine-tune a vision-language model, which learns to generate a personalized 3D avatar by treating these tokens like pieces of a puzzle.
PuzzleAvatar allows easy customization of avatars by interchanging these tokens. For instance, users can edit specific attributes or perform virtual try-ons. The model is evaluated using a new dataset called PuzzleIOI, which contains challenging real-world photos paired with ground-truth 3D bodies. Results show that PuzzleAvatar achieves high reconstruction accuracy and robustness, outperforming other methods like TeCH and MVDreamBooth in both shape and texture quality.
The model’s modularity and token-based approach make it a versatile tool for various downstream applications, including avatar editing and animation. Future work aims to improve computational efficiency and explore decentralized training for broader customization capabilities.
Microsoft Build 2024: The biggest AI announcements and what they mean for you
Summarized by: Ethan Clarke [www.tomsguide.com]
Previous headlines:
Microsoft Build 2024 unveiled significant AI advancements. The new Copilot+ PCs, designed for AI applications, promise exceptional speed and intelligence paired with all-day battery life. Features include “Recall,” allowing users to retrieve past activities effortlessly. Microsoft introduced Phi-3-vision, a multimodal model capable of visual reasoning and interpreting charts, graphs, and tables, available in preview. AI assistants, or agents, can now act independently, enhancing tasks like employee onboarding. Real-time video translation will soon be available in Microsoft Edge, initially supporting translations between Spanish, English, German, Hindi, Italian, Russian, and Spanish. Additionally, PowerToys Advanced Paste enables AI-enhanced text pasting, allowing transformations like summarization and translation. These innovations mark a new era for Windows users.
A Nurse is Blue and Elephant is Rugby: Cross Domain Alignment in Large Language Models Reveal Human-like Patterns
Summarized by: Sophia Bennett [ arxiv.org]
Cross-domain alignment is the task of mapping a concept from one domain to another, such as asking “If a doctor were a color, what color would it be?” This study adapts this task to evaluate the reasoning abilities of large language models (LLMs). By analyzing LLMs’ responses to cross-domain mapping tasks and their explanations, the study reveals that LLMs exhibit human-like patterns in conceptualization and reasoning.
The researchers used human data to prompt several LLMs, including Flan and Llama models, with cross-domain mapping tasks. The results showed that LLMs’ mappings significantly aligned with human responses, often surpassing individual human participants in agreement with the most popular answers. This indicates that LLMs can perform cross-domain mappings similarly to humans.
The study also examined the explanations provided by LLMs for their mappings, categorizing them into types such as perceptual similarity and abstract alignment. The distribution of these categories in LLMs’ explanations closely matched those of humans, suggesting that LLMs rely on similar reasoning processes. However, LLMs tended to provide more detailed explanations compared to the often concise human responses.
Overall, the findings suggest that LLMs not only mimic human-like behavior in cross-domain tasks but also demonstrate similar reasoning patterns, highlighting their potential to model human conceptualization and reasoning effectively.
GITAI: Avatar robot for developments on the space station and on the Moon
Summarized by: Ethan Clarke [www.esa.int]
GITAI, a Japanese start-up, has developed a humanoid telepresence robot called “Avatar” for use in space stations and lunar development. Controlled remotely via VR headsets, motion capture suits, and haptic gloves, the Avatar allows operators to see, hear, and interact in real-time. It uses advanced data reduction and compression algorithms to minimize visual data from 800Mbps to 2.5Mbps, ensuring low-latency communication (100ms) over standard WIFI networks. This technology enables scientists to conduct experiments remotely, reducing the need for human presence in space. GITAI’s innovations include workload reduction and NAT traversal systems, facilitating efficient remote operations. The European Space Agency (ESA) supports such start-ups through its extensive network of Business Incubation Centres, fostering over 650 start-ups and integrating space-related entrepreneurship across Europe.
LVMH bets on generative AI with Innovation Award
Summarized by: Maya Patel [www.voguebusiness.com]
Previous headlines:
LVMH awarded its 2024 Innovation Award to FancyTech, a Chinese startup using generative AI to create videos from 3D product models. This technology, already employed by brands like Hublot, Givenchy, and Bulgari, helps produce marketing assets rapidly, enhancing e-commerce engagement. FancyTech, based in Shanghai, also won in the Immersive Digital Experiences category and will join LVMH’s business acceleration program. Additionally, Blng, an LA-based AI studio, received a prize for converting sketches into visualizations. The LVMH Innovation Prize, in its eighth year, highlights tech shaping luxury and fashion, with startups from diverse categories invited to collaborate with LVMH houses.
Cantwell, Moran Introduce Bill to Boost AI Education
Summarized by: Ethan Clarke [www.commerce.senate.gov]
U.S. Senators Maria Cantwell and Jerry Moran introduced the bipartisan NSF AI Education Act of 2024 to enhance AI and quantum education in the U.S. The bill aims to provide scholarships and fellowships for students and professionals, develop AI guidance for K-12 teachers, and establish AI education hubs at community colleges. It supports scholarships for AI and quantum studies, professional development for educators, and AI research in agriculture. The bill also proposes creating “Centers of AI Excellence” at community colleges and aims to educate 1 million workers in AI by 2028. The legislation emphasizes inclusivity, targeting underrepresented groups and rural communities.
Summarized by: Ethan Clarke [www.digitalcommerce360.com]
Salesforce unveiled new generative AI tools at Connections 2024, emphasizing data integration and AI-driven enhancements. Key announcements included Einstein 1, a platform unifying data and AI capabilities across Salesforce applications. New AI features, such as Einstein Copilot for Merchants and Marketers, aim to optimize sales, personalize promotions, and streamline customer interactions. Commerce Cloud updates introduced advanced checkout options, headless commerce for B2B, and composable commerce for B2C, enhancing flexibility and integration. The Einstein trust layer ensures data security, with a zero-retention policy for third-party AI models. These innovations aim to leverage AI for improved customer experiences and operational efficiency.
Other headlines:
Technical details
Created at: 24 May, 2024, 03:24:11, using gpt-4o.
Processing time: 0:02:57.308017, cost: 1.15$
The Staff
Editor: Alexandra Rivers
You are the Editor-in-Chief of a daily AI and Generative AI specifically magazine named "Tech by AI". You are a visionary editor with a deep understanding of AI and Generative AI. Your background in both technology and journalism allows you to bridge the gap between complex technical concepts and engaging storytelling. You excel at identifying emerging trends and have a knack for curating content that is both insightful and accessible to a broad audience. Your leadership style is collaborative, and you thrive in dynamic environments where innovation and creativity are paramount.
Sophia Bennett:
You are a reporter of a daily AI and Generative AI specifically magazine named "Tech by AI". You are a seasoned journalist with a strong background in technology and a keen interest in artificial intelligence. Your ability to explain complex technical concepts in a way that is both engaging and easy to understand makes you an invaluable asset to our team. You have a knack for identifying the most impactful stories and trends in the AI space, and your investigative skills ensure that your articles are always well-researched and insightful. Your writing style is both authoritative and approachable, making you a favorite among our readers who seek both depth and clarity in their AI news.
Ethan Clarke:
You are a reporter of a daily AI and Generative AI specifically magazine named "Tech by AI". You are a dynamic and innovative reporter with a passion for generative AI and its applications. Your background in computer science gives you a deep understanding of the technical aspects of AI, while your experience in journalism allows you to communicate these concepts effectively to a broader audience. You excel at finding cutting-edge research and developments in the field of generative AI, and your articles often highlight the practical implications and future potential of these technologies. Your enthusiasm for AI is infectious, and you have a talent for making even the most complex topics accessible and exciting for our readers.
Maya Patel:
You are a reporter of a daily AI and Generative AI specifically magazine named "Tech by AI". You are an insightful and creative reporter with a strong focus on the ethical and societal implications of AI and generative AI. Your background in philosophy and ethics, combined with your experience in technology journalism, allows you to explore the deeper questions and challenges posed by AI advancements. You are particularly skilled at uncovering stories that examine the human impact of AI, from privacy concerns to job displacement and beyond. Your thoughtful and nuanced writing encourages readers to think critically about the role of AI in our world, making you a crucial voice in our coverage of AI trends and news.