Big data, AI and Charlottesville 2017
In two decades, tech will make confrontation harder to happen
AS I WATCH THE NEWS while focused on artificial intelligence for a book I’m writing, I can’t help but realize that in 20 years, the world will change so much that the white-supremacist march at UVA and the demonstration in Charlottesville in August 2017 won’t transpire in the same way. Consider how data and algorithms may change things.
First, “predictive policing” software is already used by many cities to forecast where crimes are likely to happen. It’s basically stats with past crime data. Yet as the tech improves, police agencies will know in advance that there’s a higher chance of violence and by whom. They will tap phone movements of people, matched with gun-ownership, unemployment and poverty data. Weather data will note the temperature is high which correlates with clashes, etc. This will inform the police to take more precaution. Yet protestors themselves may exercise restraint, knowing that they are in effect tracked at all times.
Moreover the web platforms that demonstrators and counter-protesters use to get information and coordinate their activities may detect that there’s a large amount of interest in the event, and generate warnings to avoid the area due to the potential for violence. Today Facebook and Google can algorithmically detect when a person is considering suicide and they subtly intervene. The same may happen with potentially violent protests.
At the demonstration, big data and AI will have a field day. Consider the UVA protest. Already it is possible to run every image into a facial-recognition system to identify the demonstrators. It will be easier in the future — and in realtime. In fact, it may not be the police who do this but the public, ushering in era of hyper-cyber-vigilantes. (Are we sure we want that?)
Next, there’s the terrible incident of Deandre Harris being beaten by white supremacist in a parking garage, photographed at the top. Currently the police can get a general sense of who may have been involved via cellphone location records. But the system is designed to connect calls, not stop crime. The technology is so blunt that it would identify several hundred people in the crowd. (Though one could match it with video recordings of the march to narrow down who was where, when.) In a decade or two, geo-location targeting will be far better. The cops will know exactly who was in the garage at exactly what time — though again, the security video and facial recognition system would pick that up too.
In 20 years, it is likely that a robotic security guard, in the form of an airborne drone, would intervene and use a Tazer against those acting violently. It would identify the skirmish, issue a warning and intervene to protect a human from other humans (as well as alert the police).
Yet consider the second-order effects. Knowing that one’s “digital footprints” incriminates oneself in realtime, and that robotic guards are ever-present to swoop down, would thugs really still charge? It’s not clear they would. Perhaps the surveillance society means snowflakes are protected from violent fruitcakes.
The good news in all this? For one thing, law-enforcement will have more tools with which to do their work: protect the public and arrest criminals. If one believes that the police have integrity, this should be welcomed. If one is wary of the abuse of police power, one should be very concerned.
The drawback? The idea of liberty will certainly be winnowed as our “data shadow” is used in ways that couldn’t happen before. We all lose anonymity and a bit of freedom in the name of preventing crime.
As for James Alex Fields Jr., the neo-Nazi driver who slammed into a score of people and killed Heather Heyer? Of course self-driving cars would never do that. But people would. And I don’t think AI can predict everything. People will still have free will and can exercise moral choice even if some odious ones decide to commit evil. The AI era means many things will change. But the demons of human nature will remain.
