Artificial intelligence (AI) created videos and pictures have become much popular and that can create some serious problems as well, because you can create fake videos, and manipulated images of any type to put anyone in trouble. Deepfakes use deep learning models to create fictitious photos, videos, and events. These days, deepfakes look so realistic that it becomes very difficult to identify the real picture from the fake one with a normal human eye, therefore, Facebook's AI team has created a model in collaboration with a group of Michigan State University that has the ability to identify not only the fabricated picture or videos, but it can even trace the origin.
The latest technology of Facebook checks the resemblances from a compilation of deepfakes datasets to find out if they have a common basis, looking for a distinctive model such as small specks of noise or minor quirks in the color range of a photo. By spotting the small finger impressions in the photo, the new AI model is capable to distinguish particulars of how the impartial network that produced the photo was invented, such as how large the prototype is and how it was prepared.
The experts experimented with the AI technology on the Facebook platform by working on data of about 100,000 fake pictures created by 100 diverse creators making a thousand snaps each. The aim was to use few pictures to make the AI technology competent enough while the rest of the images were detained and then it was shown to the technology as the picture with unidentified inventors and from where they have created. The experts working on this experiment repudiated to show how precise the Artificial intelligence’s evaluation was during the test, but they have assured that they are trying their best to make the technology even better, which can assist moderators of the platform to detect the corresponding bogus content.
The author of deepfakes wonders how effective the technology will be beyond the environment of the lab, confronting fake pictures on the internet wild. The author further said that fake images that were identified were based on the abstract database and then it was organized in the lab. There is still a chance that creators may make many realistic-looking videos and pictures that can bypass the system. The experts had no other research data so that they can compare their results with them, but they know that they have made this system work in a much better way than before.
Read next: Facebook Is Launching New Features In Order To Helps Admins With Group Moderation, Even Utilizing The Power Of AI To Do So
The latest technology of Facebook checks the resemblances from a compilation of deepfakes datasets to find out if they have a common basis, looking for a distinctive model such as small specks of noise or minor quirks in the color range of a photo. By spotting the small finger impressions in the photo, the new AI model is capable to distinguish particulars of how the impartial network that produced the photo was invented, such as how large the prototype is and how it was prepared.
The experts experimented with the AI technology on the Facebook platform by working on data of about 100,000 fake pictures created by 100 diverse creators making a thousand snaps each. The aim was to use few pictures to make the AI technology competent enough while the rest of the images were detained and then it was shown to the technology as the picture with unidentified inventors and from where they have created. The experts working on this experiment repudiated to show how precise the Artificial intelligence’s evaluation was during the test, but they have assured that they are trying their best to make the technology even better, which can assist moderators of the platform to detect the corresponding bogus content.
The author of deepfakes wonders how effective the technology will be beyond the environment of the lab, confronting fake pictures on the internet wild. The author further said that fake images that were identified were based on the abstract database and then it was organized in the lab. There is still a chance that creators may make many realistic-looking videos and pictures that can bypass the system. The experts had no other research data so that they can compare their results with them, but they know that they have made this system work in a much better way than before.
Read next: Facebook Is Launching New Features In Order To Helps Admins With Group Moderation, Even Utilizing The Power Of AI To Do So
