I'm sure it was pretty easy to figure out which accounts he created at the end.

Well, Microsoft still hasn't figured out a way to detect the people creating thousands of hotmail and outlook accounts and using them for porn site spamming.

He is the One. He can stop bullets. He can fly! He can do ANYTH... (hold on, are we making sequels?)

Stopping bullets and flying, that's it.

In UK, sports gambling is the natural companion of liquor stores, payday lenders and similar businesses preying on / doping down poor people.

Elsewhere in Europe, it's typically regulated.

The most impressive thing is that the usual suspects still haven't shown up in the Slashdot thread to tell us how unimpressed they are and how this is not "real AI". I had anticipated progress, but that they would shut up so soon defied even my most hopeful predictions.

The second stage of any scam is to package it up and sell it to others, once it's too much work and risk for too little profit. Ready-made scam kits are about on the same level as "guaranteed" lottery schemes.

OSINT is solidly groomed, funded and even set up by the intelligence agencies, out of fear of it arising for real and being independent.

They remember Wikileaks, even if they're trying to make us forget. An "intelligence agency for the people" is their greatest fear.

It started with the blog Brown Moses. A Something Awful goon who had taken as hobby watching videos out of Syria since one of his forum buddies had decided to drop everything and become a millitant. He was pretty good, he noticed and documented a lot of things the intelligence agencies hadn't. They were scared by that, but very relieved he'd focused on that conflict. So they flattered and groomed him, and paid him, and got him to start a small media outfit (bellingcat), and got their own people into it.

Nobody actually cares about the Navajo nation, you know. These organised minority groups play much the same role in US politics as Nauru does in the UN. They vote however the highest bidder asks them to vote, and the bids aren't high. It's very apparent in US presidential election years like this one: organized minority interest groups (not to be confused with the actual minorities they supposedly speak for) can be directed to have concerns where the money points.

So if the White House really had a meeting about this, it's not out of concern for Navajo beliefs. It's about "hey, are they just trying to raise their profile for the upcoming election season, or did someone we care about pay them to say this?"

They are not selling it. It's also not a dataset of images, it's a dataset of links to images. Precisely because they wanted to limit liability in cases like this (or copyright).

Yes, the famously strong candidate Hillary Clinton.

The incumbent wasn't term limited last time around either.

Google did change its algorithm in 2016 to fight disinformation, trying to favor sites with "experience, expertise, authoritativeness, and trustworthiness."

And I'm sure that worked out fine and there was no blowback whatsoever.

I was about to comment that it was not Nokia, but a bunch of corporate lawyers wearing Nokia's flayed skin as a suit.

But your comment, although less in the Halloween spirit, was more informative.

There is already plenty enough competition in driving people for money that it's hanging right there on the It's-not-worth-doing-this line.

Competition drives down costs, but only asymptotically to the cost of production. Double the competition, you get maybe a few cents lower prices, if it's already close to the cost of production.

But you know what doesn't move asymptotically? Incentive to cheat. As this news story illustrates: When you live on the margin, when everyone is barely getting by, those who cheat even a little will have a tremendous advantage. It's the same reason there's so much doping in Tour de France: all the legitimate ways of getting even a tiny edge on your competition have already been exhausted. And the cost of policing, of keeping people honest, quickly eclipses the tiny savings from getting ever so slightly closer to the asymptote.

So your auction scheme would make things worse, I guarantee it. I don't know exactly what way they would find to cheat, but they'd find one. If not they wouldn't be driving for long.

If glibertarians got their way on such things, you'd certainly get cheap taxis, but the only way taxi drivers could make a living was by occasionally selling one of their passengers to the Mafia's kidney harvesting operation.

You can't define "understand" in a non-circular way. Should not a Slashdot poster understand that we, too, can be seen as executing algorithms - and that whatever hard constrains about what algorithms can and cannot do, also apply 100% to us?

Repeating it won't make it true.

AI models need more data because they're more general learners than humans. They have less built-in priors about what they should be doing, they could potentially learn all sorts of patterns that humans can't, so naturally it takes more data to constrain them to one particular pattern.

We make them more powerful by making them less general. Making them worse at fitting all the patterns we won't need, is a fine price to pay for making them better at the patterns we do need (In fact, it's basically the only way we get better models - that's the NFL theorem for statistical learning).

The ability of a few poisoned pixels ability to throw them off, is also due to their overeagerness to learn complicated exceptions to the rules. As models get less general and more powerful, today's deceptive tricks will get less useful - but also, probably new ones will appear.