Hackers in the '80s were taking apart phone hardware and making free long-distance calls because the phone company didn't deserve its monopoly purely for existing before they were born. Hackers in the '90s were bypassing copyright and wiping the hard drive of machines they cobbled together out of broken machines to install an open source OS on it so that Redmond, WA couldn't dictate their computing experience.
I think there's a direct through-line from hacker circles to modern skepticism of the kind of AI discussed in this article: the kind where rules you don't control determine the behavior of the machine and where most of the training and operation of the largest and most successful systems can, currently, only be accessed via the cloud portals of companies with extremely questionable ethics.
... but I don't expect hackers to be anti-AI indefinitely. I expect them to be sorting out how many old laptops with still-serviceable graphics cards you have to glue together to build a training engine that can produce a domain-specific tool that rivals ChatGPT. If that task proves impossible, then I suspect based on history this may be the one place where hackers end up looking a little 'luddite' as it were.
... because "If the machine cannot be tamed it must be destroyed" is very hacker ethos.
The whole point was to take these things apart, figure out how they work, and make them things we want them to do instead of being bound by arbitrary rules.
Bypassing arbitrary (useless, silly, meaningless, etc) rules has always been a primary motiving factor for some of us :D
I think there's a direct through-line from hacker circles to modern skepticism of the kind of AI discussed in this article: the kind where rules you don't control determine the behavior of the machine and where most of the training and operation of the largest and most successful systems can, currently, only be accessed via the cloud portals of companies with extremely questionable ethics.
... but I don't expect hackers to be anti-AI indefinitely. I expect them to be sorting out how many old laptops with still-serviceable graphics cards you have to glue together to build a training engine that can produce a domain-specific tool that rivals ChatGPT. If that task proves impossible, then I suspect based on history this may be the one place where hackers end up looking a little 'luddite' as it were.
... because "If the machine cannot be tamed it must be destroyed" is very hacker ethos.