Tons of pixels have been spilled debating the pros and cons of AI. It’s not all that reliable now but I can easily imagine it being the spark that sets off the fabled singularity. Right now, though, I am very skeptical about anything that AI proclaims as true.
For me, its major problem is its abuse by the suits. In our own industry, we see this most starkly in the laying off of technical staff because, “AI can do all that much more cheaply and without the need for those annoying nerds”. Of course, when the network starts falling down, as we’ve seen recently, and there’s no one with the knowledge and experience to fix things, that calculation doesn’t seem so wise.
But it’s not just the tech industry that’s learning this lesson. Even the criminal justice system is discovering the perils of relying on AI. Over at techdirt, Tim Cushing has a report about a Kentucky detective using facial recognition to identify and charge a person for passing a bad check. The problem was that the person identified lived in a different state and could prove he wasn’t even in Kentucky when the bad check was passed.
Now, of course, the victim is suing the officer involved. The District Court ruled that AI can not be used by itself to establish probably cause under the 4th amendment. It’s another example of those in authority thinking that AI is a low effort alternative to doing the actual work that they’re supposed to do. Instead of doing the investigation required by the case, the detective just ran a facial recognition, got a hit, and made a charge. This despite that fact that the accused looked nothing like the perpetrator except that they were both black men with facial hair.
The lesson should be clear. Use AI if you must but use it wisely and with scepticism. In the meantime, it’s heartening to know that the 4th amendment is withstanding the onslaught.