Consider chemicals like chlorine gas and acrylic acid. Just like some of these AI models, they try very hard to escape containment, and sometimes people die when they do. To prevent that, we have 29 CFR 1910.119, the Process Safety Management standard. It doesn’t tell companies how to contain dangerous chemicals. It requires companies to create management systems that contain them, are regularly inspected, regularly maintained, and regularly reviewed to insure that they work. Management creates the safety system and if it fails, it’s managements fault. If your AI escapes and commits crimes, that’s managements fault. If that hurts the IPO value, Boo Hoo.
Accountability requires consequences. I'd like to start with every AI company CEO being liable for both fiscal and legal penalties if their company's product is shown to be defective in being used for criminal reasons. By that, I mean a bug in the code is not criminal, but writing it without ethics rules that prevent it from helping someone commit suicide IS.
And if you want to be a CEO of an AI company, you had BETTER BE competent to assess this, because "I'm an ideas guy, not a programmer" should not be enough to keep you out of prison when that person kills themself! You want the multi million (or billion) dollar payout, you WILL BE held accountable for your product.
I want this for car makers, tobacco farmers/sellers and every other profit motivated purveyor of harmful products too. Did I mention the gun makers yet? They are FIRST in line for this! After all, they've been gaslighting us about the danger of their products for decades now!
It's proven nearly impossible to hold corporate CEOs accountable, but if corporations are people, why don't we hit them with the death penalty? Or fifty years in prison, with their assets frozen? Suddenly the Extreme Court might decide they are merely artificial legal constructs after all.
If we regulate identity, making our government responsible for positively identifying human and artificial intelligences (people and systems), we can fix this, and much else besides. The right way to put humans in all the important loops is to make them indispensable to getting anything done, and accountable for what they do. The right way to restore human privacy is very similar; if everybody knows who everybody is, it becomes possible to regulate the flow of information about them. We need positive, instant, and reliable human (and non-human, and corporate) identity now, with control in the hands of the public. The Web was designed for corporate domination, and its design process violated our antimonopoly laws under the theory that Tim Berners-Lee was in control of it, not any corporation. It was oligopoly at work, and the DOJ winked at it. It's long past time to design our electronic communications in the *public* interest and not just the interests of those who have the biggest computers.
Consider chemicals like chlorine gas and acrylic acid. Just like some of these AI models, they try very hard to escape containment, and sometimes people die when they do. To prevent that, we have 29 CFR 1910.119, the Process Safety Management standard. It doesn’t tell companies how to contain dangerous chemicals. It requires companies to create management systems that contain them, are regularly inspected, regularly maintained, and regularly reviewed to insure that they work. Management creates the safety system and if it fails, it’s managements fault. If your AI escapes and commits crimes, that’s managements fault. If that hurts the IPO value, Boo Hoo.
Accountability requires consequences. I'd like to start with every AI company CEO being liable for both fiscal and legal penalties if their company's product is shown to be defective in being used for criminal reasons. By that, I mean a bug in the code is not criminal, but writing it without ethics rules that prevent it from helping someone commit suicide IS.
And if you want to be a CEO of an AI company, you had BETTER BE competent to assess this, because "I'm an ideas guy, not a programmer" should not be enough to keep you out of prison when that person kills themself! You want the multi million (or billion) dollar payout, you WILL BE held accountable for your product.
I want this for car makers, tobacco farmers/sellers and every other profit motivated purveyor of harmful products too. Did I mention the gun makers yet? They are FIRST in line for this! After all, they've been gaslighting us about the danger of their products for decades now!
It's proven nearly impossible to hold corporate CEOs accountable, but if corporations are people, why don't we hit them with the death penalty? Or fifty years in prison, with their assets frozen? Suddenly the Extreme Court might decide they are merely artificial legal constructs after all.
If we regulate identity, making our government responsible for positively identifying human and artificial intelligences (people and systems), we can fix this, and much else besides. The right way to put humans in all the important loops is to make them indispensable to getting anything done, and accountable for what they do. The right way to restore human privacy is very similar; if everybody knows who everybody is, it becomes possible to regulate the flow of information about them. We need positive, instant, and reliable human (and non-human, and corporate) identity now, with control in the hands of the public. The Web was designed for corporate domination, and its design process violated our antimonopoly laws under the theory that Tim Berners-Lee was in control of it, not any corporation. It was oligopoly at work, and the DOJ winked at it. It's long past time to design our electronic communications in the *public* interest and not just the interests of those who have the biggest computers.
It’s only as “intelligent” as it has been created to be.