AI seems to be advancing very fast. Facebook announced that their algorithm was able to identify unsorted, unlabeled images with nearly 85% accuracy. The self-supervising algorithm managed to decide which pixel belongs to which concept without being disturbed by the different parts of each picture, such as the background images or pets in one’s arms. They claim it is much more difficult than processing language because individual words are easier to identify. This technology is expected to be able to sort or categorize images without human intervention or bias. It then may be able to identify the images of someone who altered the appearances or whose appearances have changed from a huge database.
Will this make the world fairer for everyone? In any case, it’ll make or help the visual identification process more accurate and reliable.
Enjoy reading the article and think which is more reliable to identify an individual, a human, or an AI algorithm.
Dear MEL Topic Readers,
Instagram photos help Facebook AI ‘teach itself’
AI seems to be advancing very fast. Facebook announced that their algorithm was able to identify unsorted, unlabeled images with nearly 85% accuracy. The self-supervising algorithm managed to decide which pixel belongs to which concept without being disturbed by the different parts of each picture, such as the background images or pets in one’s arms. They claim it is much more difficult than processing language because individual words are easier to identify. This technology is expected to be able to sort or categorize images without human intervention or bias. It then may be able to identify the images of someone who altered the appearances or whose appearances have changed from a huge database.
Will this make the world fairer for everyone? In any case, it’ll make or help the visual identification process more accurate and reliable.
Enjoy reading the article and think which is more reliable to identify an individual, a human, or an AI algorithm.
https://www.bbc.com/news/technology-56321296
電話:0422-27-5366
住所 〒181-0013 東京都三鷹市下連雀3-33-13 三鷹第二ビル101
電話受付9時~18時(土日祝祭日を除く)