AI lacks human experience:How engineers use ‘shadow thinking’ to confuse AI with junk code

AI can easily memorize rules and documented data, but it cannot learn the human skill that comes from years of experience, intuition, and insight. For example, an experienced doctor can predict a disease by observing a patient’s breathing even before looking at reports, or a coder can find complex bugs without any specific rules. Humans cannot articulate this skill in words. To steal this skill, companies are monitoring every click and action on employees’ computers, and AI is being trained using head cameras, which has angered employees, causing them to hide information. Learn about these tactics used by employees… Writing logic on rough paper Software engineers know that their keystrokes and coding patterns are being tracked by AI models. To avoid this, they now engage in ‘shadow thinking’. In this, developers write the logic and algorithms of the code on rough paper or in a diary instead of thinking directly on the computer. They only come to the system to type the final code. To confuse the AI, they deliberately insert junk code and later remove it. With this strategy, AI cannot grasp the precise chronology of their thought process. Distance from chat boxes In call centers and consulting firms, AI tracks chat and voice. To avoid this, employees have come up with the ‘between the lines’ (the unspoken meaning) method. When a complex case arises, instead of using AI-monitored official chat boxes, they use codewords or personal phones to communicate among themselves. They decide the solution ‘off-the-record’ among themselves and write a formal answer on the screen, so AI cannot learn what human ingenuity was used to appease a difficult customer. Use of mouse jitterers Tech giants like Meta track mouse clicks and even cursor speed for AI training. To circumvent this, designers have started using ‘mouse jitterers’ or hardware tools that randomly move the cursor across the screen in the background. While the designer is actually contemplating design intricacies, this tool feeds fake data to the AI. This prevents the AI algorithm from distinguishing between genuine creative choices and fake movements, leading to confusion. AI poisoning Artists are now using anti-AI tools like ‘Nightshade’ and ‘Glaze’ in the background before submitting final work or processing it on screen. These tools, created by computer scientists at the University of Chicago, make changes to the pixels of digital paintings that are not normally visible, but when AI steals this data, it confuses the AI. For example, the artist might be creating a ‘dog’, but the AI tracker will feed it as a ‘cat’, which ruins the entire AI training.