28 Jan 2025
5 menit waktu baca
745view
5 menit waktu baca

Introducing GJW Ethical AI - Starting A New Trend for Non-addictive Online Viewing

At Gan Jing World, we believe that technology should serve people—not control people. In a digital world where communities and regulators are increasingly concerned about online addictions, children’s mental health, and the effects of harmful content, many online platforms design their algorithms to glue users. Gan Jing World’s mission is to promote kindness, respect, and trust among people.

Leveraging technology for good, Gan Jing World’s innovative Ethical AI technologies contain a suite of advanced deep learning models that measure inappropriate content in video, image, text, and audio. These intelligence modules work together to effectively filter harmful content, while prioritizing content that promotes kindness and mental wellbeing, helping prevent cyberbullying.

The Intelligence Modules:

1. Video Quality Rating (VQR)

The VQR intelligence module evaluates image quality and aesthetics of videos and/or thumbnails. Content is then rated on a scale of 0-5, gauging a level of appropriateness.
 

AI-generated rating for video thumbnails, on a scale of 0-5. All images under CC licenses 1, 2, 3, 4, 5, 6, 7, 8, 9.

2. Inappropriate Content Detection (ICD) with >90% Accuracy

The ICD module identifies inappropriate content in video thumbnails and frames, achieving more than 90% accuracy in detections. Inappropriate content covers four key areas: “Sexual Indication Content”, “Harmful Drugs”, “Violence”, and “Horror Content”. 

A sample output from the ICD module is shown in this figure, with the inappropriate content for kids marked in red. All images under CC licenses 1, 2, 3, 4, 5, 6, 7, 8.


3. Textual Insights

Textual Insights identify the tone, sentiment, and emotions in unstructured text, while also recognizing real-world entities. This intelligence module supports multiple languages. Results are used to label bad language as inappropriate for underage groups.
 

-    Multilingual Detection

-    Sentiment and emotion detection

This figure shows emotion and sentiment recognition in text. Positive (green), mixed (peach), and negative (red) sentiments are also captured.

Gan Jing World, an Online Platform That's Safe for All

Creating a safe digital environment, especially for the next generation, has been at the heart of Gan Jing World. We use proprietary technologies and AI models that are designed and implemented with a mission: to create a truly positive, vibrant digital community filled with wholesome content, free from gratuitous violence, crime, and objectively harmful behavior.