 Imagine a law enforcement agent – an FBI agent, or a detective in a large police department – who wants to track people passing out leaflets. Current technology might use facial recognition to search for specific people who are known activists, prone to such activity. Or the agent could try not to fall asleep while watching hours of surveillance video to pick out leaflet-passers. Or, with enough time and money, the agent could task an AI system to analyze endless hours of crowds and human behavior and to eventually train it to recognize the act of leaflet passing, probably with mixed results. A new technology, Vision Language Models (VLMs), are a game-changer for AI surveillance, as a modern fighter jet is to a biplane. In our thought experiment, all the agent would have to do is simply instruct a VLM system, “target people passing out leaflets.” And she could go get a cup of coffee while it compiled the results. Jay Stanley, ACLU Senior Policy Analyst, in a must-read piece, says that a VLM – even if it had never been trained to spot a zebra – could leverage its “world knowledge (that a zebra is like a horse with stripes.)” As this technology becomes cheaper and commercialized, Stanley writes, you could simply tell it to look out for kids stepping on your lawn, or to “text me if the dog jumps on the couch.” “VLMs are able to recognize an enormous variety of objects, events, and contexts without being specifically trained on each of them,” Stanley writes. “VLMs also appear to be much better at contextual and holistic understandings of scenes.” They are not perfect. Like facial recognition technology, VLMs can produce false results. Does anyone doubt, however, that this new technology will only become more accurate and precise with time? The technical flaw in Orwell’s 1984 is that each of those surveillance cameras watching a target human required another human to watch that person eat, floss, sleep – and try not to fall asleep themselves. But VLMs make those ever-watching cameras watch for the right things. In 1984, George Orwell’s Winston Smith ruminated that: “It was terribly dangerous to let your thoughts wander when you were in a public place or within range of a telescreen. The smallest thing could give you away. A nervous tic, an unconscious look of anxiety, a habit of muttering to yourself – anything that carried with it the suggestion of abnormality, of having something to hide." Thanks to AI – and now to VLMs – the day is coming when a government official can instruct a system, “show me anyone who is doing anything suspicious.” Coming soon, to a surveillance state near you … Comments are closed.
|
Categories
All
|
RSS Feed