The dark side of AI: what does artificial intelligence do while we sleep?

In recent years, and even more so with the rapid development of artificial intelligence tools, the technology has moved from research laboratories into almost every area of our lives. But alongside the ability to write, create images, analyze information, and streamline work, its dangerous side is also being exposed: there are already AI tools capable of creating particularly convincing voice and image fakes, companies are examining how

Now14Author: Efrat Briner
Source
The dark side of AI: what does artificial intelligence do while we sleep?
Photo: Now14 / אילוסטרציה | נערך באמצעות בינה מלאכותית

In recent years, and even more so with the rapid development of artificial intelligence tools, the technology has moved from research laboratories into almost every area of our lives. But alongside the ability to write, create images, analyze information, and streamline work, its dangerous side is also being exposed: there are already AI tools capable of creating particularly convincing voice and image fakes, companies are examining how the technology will change their workforce, and security systems are grappling with the question of how much authority can be delegated to an algorithm. Research and data from recent years show that the dangers are not all futuristic scenarios – some have already become reality.

One of the most troubling claims surrounding AI concerns what happens on personal devices when their owners are not using them. Modern phones do perform actions in the background, and technology companies also use local processing – that is, actions that are performed within the device and not necessarily on an external server.

One of the methods is "Federated Learning," through which models can be improved without transferring all of the user's raw data to a central location. The idea is that the device performs part of the learning process itself and shares results or updates necessary to improve the system.

However, here one must distinguish between the real technology and claims for which sufficient support has not been found. I have not found a reliable basis for the figure that an average device sends "20–40 megabytes of psychological and commercial profile" every night, nor that phones comprehensively analyze private messages and photos at night to build a personality profile for training large AI models. Therefore, it is incorrect to present such claims as fact.

Also, regarding claims about "active listening" for advertising purposes, one should be wary of a sweeping determination that the phone routinely listens to the bedroom, analyzes coughing or conversations, and prepares advertisements as a result. This is a dramatic claim that requires stronger evidence.

Related News