(maadaa AI News Weekly: June 4~ June 10)

1. Apple Intelligence: Siri is now powered by GPT-4o

News:

Apple has announced a partnership with OpenAI to integrate ChatGPT into its products and services, including Siri, the company’s virtual assistant. The collaboration aims to enhance Siri’s conversational abilities and make it more natural and useful for users by leveraging OpenAI’s GPT model.

Key Points:

- Apple will integrate OpenAI’s ChatGPT technology into its operating systems, including iOS, iPadOS, and macOS.

- With user permission, Siri will tap into ChatGPT’s intelligence to provide more accurate and helpful responses.

- ChatGPT will also be available in Apple’s systemwide Writing Tools, allowing users to generate content and images using the AI model.

- The integration will be powered by OpenAI’s GPT-4o model and will be available for free later this year, with premium features for paid subscribers.

Why It Matters?

This partnership is significant because it will significantly improve the training dataset for Siri and other Apple AI services. By integrating OpenAI’s advanced language model, ChatGPT, Apple will have access to a vast corpus of conversational data and natural language processing capabilities. This enhanced training dataset will enable Siri to provide more accurate, natural, and contextually relevant responses, setting a new standard for virtual assistants and improving the overall user experience.

2. Ex-Google CEO Eric Schmidt’s Project Eagle: AI-Powered Military Drones in the Making

News:

Eric Schmidt, ex-CEO of Google, has been developing AI-powered military drones via Project Eagle, previously known as White Stork. The project, drawing top talent from Apple, SpaceX, Google, government agencies, and his networks, is conducting tests in Silicon Valley and Ukraine. The goal is to produce drones that can precisely identify and engage targets with AI technology.

Key Points:

- Project Eagle has recruited around a dozen employees from Apple, SpaceX, Google, and the federal government.

- The AI drones are being tested at Schmidt’s family office in Silicon Valley and in Ukraine, where he has frequently visited.

- Schmidt believes AI can revolutionize modern warfare and has advocated for increased military aid to Ukraine.

- Experts have raised concerns about the legal and ethical implications of rapidly adopting AI in combat.

Why It Matters?

It emphasizes the increasing role of AI in military uses, especially in target identification and strikes. Project Eagle’s development of AI drones is set to produce extensive data from both tests and real-life use, which is crucial for training AI to identify and track targets accurately, thus improving future AI-driven military tech.

3. NewsBreak’s Fictitious Stories Highlight the Need for Better AI Training

News:

NewsBreak, a popular free news app in the US with over 50 million monthly users, has been accused of using AI to generate and publish fictitious stories since 2021. The app licenses content from major outlets but also rewrites information using AI tools, leading to numerous inaccurate and fabricated articles affecting local communities.

Key Points:

- NewsBreak published a completely fabricated story about a shooting in Bridgeton, New Jersey, on Christmas Eve 2023, which the local police debunked.

- Since 2021, there have been at least 40 instances where NewsBreak’s AI-generated content contained errors or false information.

- The app has faced criticism for copying articles from websites without permission and settling copyright infringement cases.

Why It Matters?

The NewsBreak case shows the need for AI training datasets to:

- Differentiate real from fake news and spot inaccuracies.

- Follow journalistic ethics, like copyright respect and avoiding plagiarism.

- Improve language processing for better summarization and context preservation.

- Address data privacy and misuse concerns.

Incorporating real-world examples like NewsBreak helps AI create trustworthy content responsibly.

4. Apple Pivots to Home Robotics After Scrapping Car Project

News:

Apple is reportedly exploring the development of home robots as a potential new product line, following the cancellation of its self-driving car project earlier this year. Engineers are working on a mobile robot that can follow users around their homes and a tabletop device with a rotating display designed to mimic human head movements during video calls.

Key Points:

Apple is in the early stages of investigating home robots, likely seeking its “next big thing” after abandoning the car project.

The mobile robot would accompany users throughout their homes, while the tabletop device would use robotics to rotate a display during video calls.

The tabletop robot project is more advanced than the mobile robot, but Apple executives disagree on whether to continue.

Why It Matters?

Apple is venturing into home robotics, which could improve AI models in areas such as object detection, natural language processing, computer vision, and path planning. The data gathered from these robots could enhance home automation systems.

5. Additional News:

Musk diverts Tesla’s Nvidia chips to X and xAI, delaying its AI projects and concerning investors.
AI models are revolutionizing weather forecasting with faster, more accurate predictions.
The World AI Creator Awards and platform FanVue are hosting the first ‘Miss AI’ contest with over 1,500 AI-generated models competing.
Kling, an AI model for hyper-realistic video generation from a Chinese company. It can make videos up to two minutes long at 1080p resolution and 30 frames per second.
Google’s new AI tool aids marine biologists in analyzing and conserving coral reef ecosystems.
Siwei Lyu from the University of Buffalo developed the DeepFake-o-meter to detect manipulated media and offered tips for identifying altered photos, videos, and audio.

Recommended Open & Commercial Datasets

Open Dataset #1. PolyAI-LDN Conversational Datasets

This dataset from GitHub provides a collection of conversational datasets designed for training machine learning models in natural language understanding and response generation. It includes structured conversations that can be utilized to develop chatbots that need to handle a sequence of conversational contexts and responses.

https://github.com/PolyAI-LDN/conversational-datasets

Open Dataset #2. Customer Service Corpus

Part of the Pchatbot dataset, this large-scale resource includes 435,005 dialogues based on customer service interactions from JD.com. It is designed to assist in building task-oriented dialogue systems.

https://paperswithcode.com/paper/the-jddc-2-0-corpus-a-large-scale-multimodal/review/

Commercial Dataset #1. Large-Scale Professional Domain Corpus Dataset — Chinese

Key Features:

- 120M Electronic Documents

- 2PB fine structured data

- Most of popular e-book formats

- Hundred of professional domains

- Comprehensive Format Support: most of the popular e-book formats such as PDF, EPUB, mobi, azw (3), and DjVu.

- Advanced OCR engine for Formulas: Equations and multiline formulas in PDFs are easily transformed into Latex text.

- Precise Layout Reproduction: Ensures the original formatting of PDFs is preserved, including text arrangement, headings, and diagrams.

https://maadaa.ai/datasets/GenDatasetDetail/Large-Scale-Professional-Domain-Corpus-Dataset---Chinese

Commercial Dataset #2. Chinese Bills Dataset

A diverse bill in different scenarios. Data collection equipment includes phones, cameras, and tablets. It covers over 10 types of commercial receipts and invoices used in mainland China, including flight tickets, train tickets, hotel receipts, general tickets, taxi receipts, quota invoices, value-added tax invoices, toll receipts, coach tickets, and others.

Over 20 labeling categories, including types, provinces, quality, codes, invoice date, company/certificate numbers, fax numbers, phone numbers, car licenses, IDs, boarding time, drop-off time, price, mileage, wait time, surcharge, service charge, and receipts, etc.

Application Scenarios: Tourism, Retail, Financial, etc.

https://maadaa.ai/datasets/DatasetsDetail/Chinese-Bills-Dataset

1. Apple Intelligence: Siri is now powered by GPT-4o

News:

Key Points:

Why It Matters?

2. Ex-Google CEO Eric Schmidt’s Project Eagle: AI-Powered Military Drones in the Making

News:

Key Points:

Why It Matters?

3. NewsBreak’s Fictitious Stories Highlight the Need for Better AI Training

News:

Key Points:

Why It Matters?

4. Apple Pivots to Home Robotics After Scrapping Car Project

News:

Key Points:

Why It Matters?

5. Additional News:

Recommended Open & Commercial Datasets

Open Dataset #1. PolyAI-LDN Conversational Datasets

Open Dataset #2. Customer Service Corpus

Commercial Dataset #1. Large-Scale Professional Domain Corpus Dataset — Chinese

Commercial Dataset #2. Chinese Bills Dataset

Source: