Ramah Al-Dalqamouni - Al Jazeera Net
A decade ago, social networking sites spread a strange challenge known as the "ten-year challenge" where users publish two adjacent images, the first being modern and the second going back to ten years. With the spread of this "challenge," charges are being raised that Facebook is behind it to take advantage of images The performance of its face recognition technology, despite its explicit denial of this.
In an effort to prevent misinformation, Al Jazeera Net continued with Mr. Ahmed Al Shalabi, co-founder of Cloud AI, a specialist in artificial intelligence technology, and asked him a number of questions about the truth of the issue.
In this regard, Al-Shalabi said that Facebook can benefit from any image on its platform, but if you want to use images published at a time interval of ten years, this is different.
He explained that the development of artificial intelligence techniques needs laboratory controlled samples, in the sense that the sample used in computer training must be reliable and standard, and this is not available in the images spread on the marking of the challenge of the ten years.
"You simply find many of these images comic or purposeful, and there is no proof that the interval is 10, 15 or 20 years," he said.
Face recognition is a stable technique and is only developed in certain directions not previously used. According to Al Shalabi, these techniques have now been used in dozens of airports around the world to identify the dangerous ones once they enter them using dozens of cameras.
He stressed that competition between companies today in this technology is no longer in the accuracy, but the speed of recognition, pointing out that some of them reached a speed of up to a few seconds.
And whether Facebook can benefit from these images in any way, explained Shalabi that to benefit the company from the images, they must be disciplined and reliable, and the standards required change the change of purpose, if the purpose, for example, to identify the presence of people or not in the picture we need images containing People from different angles and in varying degrees of lighting.
If the object is to identify a particular person, a number of images must be provided to him in different forms in different situations and at different angles, and his time can be identified in fixed or televised images.
Electronic fingerprint
But this does not mean that the user's dissemination of personal image on the Internet is free from any danger, such images may find someone who benefits in one way or another, and in this regard Shalaby said that "advice recommended by all users of the Internet: strive to reduce your electronic mark as much as you can "He said.
For example, Deep Faking requires a few dozen audio and video clips and pictures to make a complete video of the same person doing the work that the owner did not do or talk with. Gets.
He added that the most famous example of this is the video made by actor Gordon Bailey to former US President Barack Obama, where the artist speaks like Obama, and the computer issue audio and video like Obama, noting that the falsity of such videos can be detected only with specialized programs.
Chalabi pointed to previous attempts by some technology companies - such as Microsoft - to take advantage of the images of users on the Internet, but not in the identification of the identity of the face, but in the area of anticipation of the age of the person of his image.
According to the expert, the results of the trial conducted by Microsoft five years ago was a "laugh", and the company came out that it made the technology to humor no more, but others followed the experience, including some open source programs like this.
"The privacy issue has been decided in favor of companies. The last decision about your privacy is to use your smartphone and networking sites. Everything becomes a result," he concluded.