maadaa AI News& Open dataset Share: Anthropic’s New Business App, Microsoft Limits AI, Sora’s Music Video

(maadaa AI News Weekly: April 30~ May 6)

1. New Music Video Generated In Sora

News：

Indie artist Washed Out has released the first officially commissioned music video created entirely using OpenAI’s text-to-video AI model, Sora. The four-minute video for the song “The Hardest Part” was directed by filmmaker Paul Trillo and depicts a couple’s relationship journey from teenage years to old age.

Key Points

• The video was generated by Trillo using around 700 video clips created with detailed text prompts in Sora.
• The “infinite zoom” style video allowed Trillo to realize a concept he had envisioned but couldn’t execute with traditional methods.
• Washed Out sees AI tools like Sora as enabling artists to “dream bigger” beyond production constraints.
• The video’s release sparks ongoing debates about AI’s implications for jobs and creative work in the entertainment industry.

Why It Matters?

This AI-generated music video demonstrates the progress in text-to-video AI models like Sora, which use vast and diverse training datasets to create compelling visual content. As these datasets grow, AI has the potential to democratize access to high-quality, customized visual content.

https://www.youtube.com/watch?v=-Nb-M1GAOX8

2. Microsoft Restricts Police AI Use

News:

Microsoft has banned police use of its facial recognition technology, overhauled its AI ethics policies, and prohibited the use of its Azure AI Face service by or for state or local police in the U.S. These actions demonstrate Microsoft’s commitment to responsible AI practices and ethical considerations.

Key points:

• Microsoft has banned the sale of its facial recognition technology to police departments in the U.S. until there is a federal law governing the technology.
• The company updated its AI ethics policies, limiting access to its Azure Face service and phasing out facial analysis features that infer emotional states, gender, or age.
• Microsoft banned the use of its Azure AI Face service, including the use of real-time facial recognition technology on mobile cameras used by law enforcement.

Why It Matters?

The news is noteworthy for emphasizing Microsoft’s proactive stance on tackling the ethical and societal concerns associated with AI, especially within the realms of law enforcement and public safety. Additionally, it underscores the critical nature of utilizing ethical and legally permissible datasets for AI training.

3. Anthropic Launches iPhone App and Premium Plan For Businesses

News：

Anthropic, a generative AI startup, is introducing a new premium “Team” plan for enterprise customers in regulated industries. The plan provides higher-priority access to Anthropic’s AI models and additional admin and user management controls. Additionally, Anthropic is launching an iOS app that offers free and upgraded Pro and Team access to Claude 3, with real-time image analysis capabilities.

Key Points:

• Anthropic is launching a new enterprise-focused premium plan called “Team”
• The plan provides higher-priority access to Anthropic’s Claude 3 AI models
• It also includes new features like billing and user management controls, as well as upcoming collaboration tools
• Anthropic is positioning the Team plan to capture significant enterprise market share as more companies shift from AI experimentation to full-scale deployment

Why It Matters?

This news is significant for the AI training dataset market as it highlights the growing demand from enterprises to deeply integrate advanced generative AI models like Anthropic’s Claude 3 into their business workflows and processes. As more companies move beyond AI experimentation and towards large-scale deployment, the need for high-quality, enterprise-focused training datasets will continue to rise.

Anthropic’s launch of the Team plan positions the company to be a key player in this growing market.

Additional News:

1. Tesla has shared a preview of its upcoming Robotaxi app that will enable users to request self-driving Tesla cars for transportation. The aim is for these autonomous Tesla cars to function as taxis, generating income for both Tesla and vehicle owners.
2. Amazon Q offers developers advanced capabilities like code generation, testing, debugging, and planning. It simplifies access to internal company data and allows for custom generative AI applications tailored to an organization’s needs.
3. Ukraine’s Foreign Ministry launches the world’s first AI spokesperson — Victoria Shi. Victoria will issue official statements and interact with the press on behalf of the ministry.
4. Nvidia’s ChatRTX is getting updated. The experimental chatbot is adding more AI models and now supports voice queries using Whisper, an AI speech recognition system.
5. Yelp’s AI chatbot uses OpenAI’s LLMs and Yelp’s data to provide users with relevant professional suggestions.
6. Alden Global Capital’s MediaNews Group is suing Microsoft and OpenAI for allegedly using the newspapers’ content to train AI models.

Open & Commercial Datasets:

1. InternVid

This large-scale video-text dataset is designed for multimodal understanding and generation. It contains over 7 million videos, totaling nearly 760,000 hours, and provides a rich source for studying video-text correlations and supporting diverse AI-driven applications​.

InternVid: A Large-scale Video-Text Dataset for Multimodal Understanding and Generation

2. Kubric

Developed by Google Research, Kubric is a data generation pipeline for creating semi-realistic synthetic multi-object videos. It includes rich annotations such as instance segmentation masks, depth maps, and optical flow, making it ideal for training and evaluating AI models in understanding complex video data​.

GitHub - google-research/kubric: A data generation pipeline for creating semi-realistic synthetic…

3. maadaa.ai Open Datasets

Public open datasets are the key enabler for both academic AI research and industrial AI innovation. maadaa.ai has actively sponsored high-quality open datasets for well-known AI conferences and challenges.

maadaa’s open datasets include:

• AI Photo-Video Editing Open Dataset
• Fashion & E-commerce Open Dataset

Open dataset - maadaa.ai

4. Maadaa Generative AI Datasets

maadaa.ai is committed to providing “Data-Centric” professional Generative AI data services and building large-scale high-quality dataset products for the development of Generative AI large language models (LLMs).

• The Generative Datasets include:- Multi-modal Generative AI Large Datasets — Licensed
• Large-Scale Professional Domain Corpus Dataset — Chinese

Gen Dataset - maadaa.ai

Citation:

1. https://nofilmschool.com/ai-music-video
2. https://techcrunch.com/2024/05/01/anthropic-launches-a-new-premium-plan-aimed-at-businesses/?utm_source=www.neatprompts.com&utm_medium=newsletter&utm_campaign=meta-s-gpu-budget&_guc_consent_skip=1714981023
3. https://techcrunch.com/2024/05/02/microsoft-bans-u-s-police-departments-azure-openai-facial-recognition/
4. https://backendnews.net/ai-assistant-amazon-q-accelerates-software-development/amp/?utm_source=www.neatprompts.com&utm_medium=newsletter&utm_campaign=why-microsoft-invested-in-openai
5. https://techcrunch.com/2024/04/30/yelp-is-launching-a-new-ai-assistant-to-help-you-connect-better-with-businesses/?utm_source=www.theaivalley.com&utm_medium=newsletter&utm_campaign=how-to-use-llms-as-your-personal-research-assistant
6. https://www.pymnts.com/legal/2024/newspapers-sue-microsoft-openai-for-using-content-to-train-ai/?utm_source=www.theaivalley.com&utm_medium=newsletter&utm_campaign=how-to-use-llms-as-your-personal-research-assistant
7. https://interestingengineering.com/transportation/tesla-unveils-first-look-at-robotaxi-ride-hailing-app?utm_source=www.neatprompts.com&utm_medium=newsletter&utm_campaign=tesla-s-robotaxi-app