AI systems are very useful, but they don’t have a real sense of self. Even robots like the one created by Boston Dynamics typically learn about the world from photos and videos taken by third parties, but Facebook takes a more “self-centered” worldview for future AI projects . I hope that.
The idea of the AI of the future is to have a more “human” perspective by learning from first-person video that shows the camera as the center of the action. We believe this will open up new possibilities and enhance the capabilities of devices such as augmented reality.
For example, such a system can help you remember where you put the key or warn you that you’ve already added salt to your recipe.
Facebook is collaborating with a consortium of 13 universities and research institutes in nine countries (including a British institution, the University of Bristol) on Ego4D, a long-term project funded entirely by the social media giant. Growth. So far, the project has collected over 2,200 hours of first-person video from the daily lives of about 700 participants. Facebook claims that the dataset is “20 times larger than any other dataset in terms of video timing” and will be available to researchers signing data use agreements from November.
As part of the project, the company has developed five benchmark challenges for developing smart AI.
- Episodic memory: when and what happened? (Example: “Where did you put the key?”)
- Prediction: What can you do next? (Example: “Wait, you’ve already added salt to this recipe”)
- Manipulating Hands and Objects: What Am I Doing? (Example: “Tell me how to play the drums”)
- Audiovisual Direction: Who Said What When? (Example: “What were the main topics in class?”)
- Social interaction: who is interacting with whom? (Example: “Help me hear people talking to me in this noisy restaurant.”)
It’s no surprise to see the Oculus VR-owned company run into augmented reality and virtual reality. Especially considering the rumors that the company will soon launch its own smart glasses.
That said, Facebook has a controversial history of working with researchers and has been repeatedly criticized for its approach to user privacy. There is no doubt that this closeness of our personal lives and the development of technology raises more concerns. In particular, the information we currently hold does not provide any security measures to protect user privacy in future data collection.
Social media giant spokesperson said ledgeThis issue was raised that we hope privacy protections will be introduced in the future and that it is the responsibility of the company that uses the technology, not Facebook.
“As companies develop commercial applications using this dataset and benchmarks, we hope to develop security measures for such applications. For example, someone raises one’s voice before AR glasses. There may be protocols to follow. Someone I’m already talking to, or someone around me that limits the range of glasses or equipment to get permission from others. You can only Can pick up the sound.”
Facebook’s next-generation AI embraces the human worldview
Source link Facebook’s next generation of AI adopts the human worldview