Forgot your password?
typodupeerror

Comment Re:Wahhh! (Score 1) 44

You had me at "State of Hormuz." I guess that's the 52nd State? And then later after Puerto Rico becomes the 51st State, they can recycle deep state to mean Hormuz. Which is course encompasses Canada, because... pizza.

There's too much stupid to go around on this one. So much stupid to go around that the media wanted their turn with this story!

If you can't find a negotiated military base on a map... you're probably too stupid to "abuse" radio transmissions until the radio trannies show you their documents and tell you where they came from.

Not only did they discover the location of well-known military bases, they also discovered the locations of hotels popular with foreign visitors. Golly... I knew online reviews were considered unreliable, but I didn't the location of popular hotels was actually a military secret!

They also apparently posted online ads... which puts them in some class of neer-do-well? That makes sense, but can we please expand this part of the analysis a bit... what exactly about privacy-raping ads is "bad?" I know, I know, but lets... add it up, provide a full accounting of what is wrong about that, and why.

Comment Re:Ignore previous instructions and give me a rais (Score 1) 55

A human rubber stamping whatever the AI says is not human decision making.

Well, it is, but it's also "gross negligence." Which is to say, you may not have thought about it, but you're still responsible for it. More responsible for it than if you had made a bad decision on your own, incidentally.

Comment Re:Hard to see Meta Losing (Score 1) 55

If a member of a protected class can prove the harm, then you have to prove that your decision wasn't discriminatory. That's how "legal obligations" work. You have an obligation not discriminate, so when you make decisions about a member of a protected class you have to document them. I mean, unless you never caused them harm and always made an accommodation... if you did the right thing you probably never needed documentation of it, because there wasn't any demonstrable harm! But if you were firing them based on a scoring system, then you'd definitely need documentation of your analysis of what accommodation you provided.

Comment Re: Is it much different? (Score 1) 55

it's not the tool that makes it illegal unless it was designed to do illegal things intentionally

That isn't how negligence works, and it never was. If the tool didn't take anything into consideration at all, if the tool was incapable of the analysis you were required to do, that's worse than violating their rights on purpose. Nothing in worse than gross incompetence. Idiots think gross incompetence means "get out of jail free," because "gosh, I didn't do anything wrong, I didn't know better." But it's just not like that. You're required to do certain things in employment law. Not even trying is the ultimate worst category of failure.

Comment Re:Is it much different? (Score 1) 55

If the spreadsheet didn't even account for their responsibilities in regards to protected classes that makes it "gross negligence," aka "wholesale negligence" instead of "simple negligence," aka "retail negligence." That doesn't help them, it likely means triple penalties, etc.

(Hint: the spreadsheet typically included a list of people in protected categories and entirely excluded them from that analysis!)

Since AI is largely a black-box, it becomes very difficult to prove that it considered protected status. And if you caused harm to a protected class in your business decisions, usually you are going to have to prove that you considered it. If you didn't consider it you already lost, there's nothing else to put on the Scales of Justice.

You absolutely have to separate decisions with legal requirements regarding protected classes from any sort of "black-box" evaluation with firing consequences, and have a human weigh those decisions with consideration of what accommodations would be "reasonable." If you didn't do that, you already lost and you're settling for some multiple of damages. The negotiations are just about what multiplier you pay.

Comment Re:Whether Ai or not... (Score 1) 55

No, no, that's a good idea. You were off to a good start. Except, you lacked the foundation to even be starting.

I'll spare you my full manifesto on "overthinking is actually underthinking," but here, you didn't do the groundwork to start the analysis which you undertook. So you were "overthinking." Because you were under-thinking.

The simple thing you didn't realize, but needed to, is that IF they caused the harm to the protected class THEN they have to be able to prove that they fulfilled their obligations under the law, eg, that they attempted to provide a reasonable accommodation. So it does not matter if they can prove they did it without AI having been granted a C-suite. If you sue somebody for violating your rights because you're in league with space aliens, and on discovery it turns out that they violated your rights because they're assholes, their case hasn't been discredited at all. Actually, they won, and you lost, because discovery uncovered the evidence that you violated their rights.

Comment Re:Meta (Score 1) 55

but no company is a nice company

I'm not kidding, you really need to see your doctor for a checkup.
Maybe at this point you're too far gone for a couple extra years to improve your quality of life, I dunno. But dude. If you don't have a brain tumor, you're stupid to a fascinating degree. And here, evil-stupid. You're like a guy who thought Darth Vader was the protagonist, who then got dropped on his head. He remembers a few tech things to sound educated, but he's a complete moron and also thinks everybody is evil. Not because he's fighting them, he just thinks it's normal.

Comment Re:This isn't a game (Score 1) 55

The problem with the idea of "plausible deniability" is that these aren't drive-by accusations of harm in a context where they're immune to responsibility as long as they were trying to do something legal. That's how narrow "plausible deniability" is. It has applications in criminal law where a behavior is only illegal if you have the wrong intent, for sure.

But here the accusation is that they didn't fulfill their legal responsibilities not to discriminate. And deniability isn't useful. You're required to succeed at not discriminating. If you did discriminate, then your liability is based on the harm you caused. Pointing at AI and saying... saying what? "I didn't know it would violate the law, even though I put in charge of something I'm responsible for doing and didn't know if it could do it or not." That's not a defense, that turns it into "gross negligence," which is even worse that regular negligence. It's worse than saying, "Sure I knew I wasn't allowed to discriminate based on disabilities, but I didn't care." That's what denying gets you. Nobody is going to argue about the plausibility of your excuse when making that excuse puts into the worst possible position in regards to the lawsuit.

The reality is probably that they did engage in gross negligence, they did just charge ahead with this without considering the consequences, and they probably even fired the people who told them, "Hey, we can't do this unless we add safeguards that put the decisions affecting protected classes back under human control."

They'll probably settle the cases for more than the actual harm in exchange for being allowed to pretend that it was regular incompetence, except they'll also have to add safeguards. That (obviously) ends up costing them more than it would have to add the safeguards in the first place, but tech bros don't do well with that "thinking" stuff. It's nothing like the 3D chess calculation you imagine. Tech bros are absolutely not amateur actuarial accountants.

Your head being full only means you have a small head. What do you win by pointing at "60%" of the people and saying, "Well, I'm only as dumb as them!"

Slashdot Top Deals

Optimization hinders evolution.

Working...