nickodell

Assuming you discover child exploitation on Facebook, why would you report it to Facebook, rather than the police? The police can actually arrest people.

That's true, but Google has made no commitment to keep it that way.

>Sure, but I bet Andrej Karpathy's team is using SOTA and mitigates most of the issues like the ones you mentioned.

Can you give an example of an image classification technique where classification failures are independent, or independent enough?

>In perception you hit false positives/negatives all the time but appearance of the same issue on two-three consecutive frames goes to 0 pretty quickly.

If the frames are all very similar, wouldn't the probability of failure be correlated? In other words, if I feed a picture of a panda to a NN, and it incorrectly classifies it, why would it be correct the second time around?

Here's an interesting sentence from the complaint filed by Rothschild:

>On information and belief, Defendant sells, offers to sell, and/or uses network storage systems and methods including, without limitation, the Gnome Shotwell platform, and any similar products (“Product”), which infringe at least Claim 4 of the ‘086 Patent.

Does Gnome Foundation sell Shotwell?

I don't agree. I just looked in my password manager, and I have roughly ~220 accounts across the web. If I want to go through that list and see which website rank well and which rank poorly, and I want to do that in under two hours, that gives about 30 seconds per service.

In other words, giving a single score plus a two-sentence highlight is probably about the right amount of information.

>You just have to change the onboarding process to include a fee, and offer an incentive/rebate of the subscription price to match the cost to basically make it free. That way if they decide to change their mind and not give their data, you're still in the clear by removing the incentive and either charge their payment method or disable their service until they do.

What's the practical difference between these two scenarios?

1) I offer a free service if you let me collect your data, but charge you $3/mo if you refuse to let me collect it.

2) I offer a $3/mo service, but give you $3/mo back if you let me collect your data.

Option #1 is explicitly illegal under the new California law. If option #2 is legal, then what is the law supposed to ban?

If it's a joke, or sarcastic, why is it in a document labeled "Plain English Terms of Service?" For that matter, why follow it up with "You've been warned?"

>Has anyone ever seen them do anything like this?

No, I've never seen them do this. But at the same time, I'm not going to give them the opportunity.

EasyDNS's Plain English Terms of Service make them seem like a really unprofessional company:

>We are NOT a DDoS Mitigation Service. [...] If you come on this system knowingly bringing a DDoS on your heels we shut down service (we may also wildcard your DNS to localhost and set the TTL on your zone out to a year. You’ve been warned).

or

>Guilt-by-Association: not only do we terminate any domains or websites which violate our policies, we ferret out every other domain you have on the system under different names, accounts, etc and we terminate those too (don’t worry, we can tell). There is no appeal.

I'm not looking for a DNS provider, because I'm perfectly happy with my current one, but sheesh.

>China Telecom is a global transit provider. As a sibling commenter mentioned, the document doesn't show that traffic actually went to China, just that China Telecom was an available path -- even assuming traffic transited their network, it's unreasonable to assume it actually transited through China without a traceroute at least;

The paper is based upon data from the DIMES project, which uses many volunteer-contributed nodes to run traceroutes and pings. See https://arxiv.org/pdf/cs/0506099.pdf

>"the person performs work that is outside the usual course of the hiring entity’s business" is the tricky one, but I think it more has to do of what you explain your business to be. If you are running a SaaS platform and enter into a contract with someone who provides software development expertise the relationship between your business and the contractor is different.

This argument seems too general. One way of looking at Uber is that it's a taxi service. Therefore, Uber and an Uber drivers are engaged in the same line of business. Another way of looking at Uber is that it's a platform to connect drivers and passengers. Therefore, Uber and Uber drivers are engaged in different kinds of business.

If you need to explain what your business is in a particular way, then you're on thin ice.

>There should be laws to prevent companies from refusing to do business with "high maintenance" customers without proper documentation and regulation.

Would requiring documentation significantly change what happened here? Suppose that in addition to the letter about termination of service, the Sprint customers also got a spreadsheet showing the times that they called customer service and how long those calls took. Would that help customers? I am somewhat skeptical of the value of this regulatory scheme.

>In 2016, biased search results generated by Google’s search algorithm likely impacted undecided voters in a way that gave at least 2.6 million votes to Hillary Clinton.

This is the most important claim that he makes, and as far as I can tell, he gives zero backing for it.

He has two citations. The first citation is a paper showing that re-ordering search results to show links to articles that favor candidate A will cause readers to favor candidate A. The second citation does not give any explanation for the 2.6 million figure.

----

Let's take a look at the second citation: https://www.theepochtimes.com/10-ways-big-tech-can-shift-mil...

>I found pro-Clinton bias in all 10 search positions on the first page of Google’s search results. [...] Because, as I noted earlier, Google’s search algorithm is not constrained by equal-time rules, it almost certainly ends up favoring one candidate over another in most political races, and that shifts opinions and votes.

This is literally true but realistically meaningless. Suppose that there are two candidates for office, A and B. Days before the election, it is revealed that candidate A fucks dogs. The news publishes many articles about candidate A fucking dogs. Google places those articles highly in search results, and many voters see them.

Has Google shown pro-candidate B bias by showing those articles? Obviously not. Unless Epstein can show that a viewpoint-neutral ranking would have produced a different result, what he has is meaningless.

>I hate to be a downer, but nobody even knew about this space until it was mentioned here.

Speak for yourself. :)

>The rules of the modern Internet are "encrypt everything and trust nothing." The laws of ham radio are literally "you must not encrypt anything". It wasn't meant to be, unless you want to change the laws in every ITU member state.

More to the point, the regulations around third-party communications [0] mean that this could never be routable from the public Internet.

And if 44/8 will never be routable, then what's the point of using public addresses? It could use private network addresses instead.

All the same, I find ARDC's response to be high-handed. They refuse to acknowledge that they are selling off what is essentially a public resource. They refuse to believe that others might not trust them to be a responsible steward of ~$50 million.

[0]: https://www.law.cornell.edu/cfr/text/47/97.115

It's more than just zero. Every number can be represented in multiple ways in DEC64. For example, you could represent 3 as

3 x 10^0

or

30 x 10^-1

If you have two rules like this:

    H -> O
    H -> 2O

then the interpreter will pick one at random. It is possible to write entirely deterministic programs, but some programs are non-deterministic.

I don't think your solution works.

Suppose N=20. Imagine the server and 19 of the voters are compromised, and are trying to de-anonymize your vote. It "randomly" creates a group composed of you and 19 voters that are collaborating with the server to de-anonymize your vote.

Assuming that the server has M collaborators, it can discover the votes of M / (N - 1) of the citizens.

>This is a crucial part of their process and any delays will impact their bottom line.

If the government servers go down, JetBlue can fall back to checking paper boarding passes.

Am I supposed to see something here?

https://i.imgur.com/WNBbIep.png

Yeah, it's pretty annoying. The Chrome extension "eKill" is a godsend. It makes websites like Medium readable.

https://github.com/rhardih/ekill

Here's a second writeup, which also contains a response from pipdig: https://www.wordfence.com/blog/2019/03/peculiar-php-present-...

>As computing the NullPointerException message proposed here is a considerable overhead, this would slow down throwing NullPointerExceptions.

You could lower the overhead by not producing these messages if the exception is thrown frequently. It would be similar to the existing OmitStackTraceInFastThrow mechanism [0].

[0]: https://www.oracle.com/technetwork/java/javase/relnotes-1391...

Seems interesting, but unfortunately the leaf guide is empty.

https://uucp.dataforge.tk/~uucp/wiki/index.php?n=Main.LeafNo...

>Volk One devices can connect to each other, from several miles away.

Assuming that a voice call requires 8kbit/s of bandwidth, and assuming that you have about a watt of power, how feasible is it to connect through a phone that is a mile away?

Serious question - I don't know how to do the math here.

Amazon and Google do operate in several of the same markets. Amazon sells Kindle/Fire devices. Google sells Android phones. Amazon does product search. Google has a product search engine.

>For a snake less than 4ft long, like this one, hunters receive $50. Every additional foot in length earns them another $25. On top of that, they get the minimum wage –$8.25 an hour in Florida – for the time they spend on district land.

See Wikipedia:

>In 24 U.S. states, employees who are working in a unionized shop may be required to contribute towards the cost of representation (such as at disciplinary hearings) if their fellow employees have negotiated a union security clause in their contract with management. Dues are generally 1-2% of pay. However, union members and other workers covered by collective agreements get, on average, a 5-10% wage markup over their nonunionized (or uncovered) counterparts.

>Some states, especially in the south-central and south-eastern regions of the U.S., have outlawed union security clauses; this can cause controversy, as it allows some net beneficiaries of the union contract to avoid paying their portion of the costs of contract negotiation.

>Regardless of state, the Supreme Court has held that the Act prevents a person's union dues from being used without consent to fund political causes that may be opposed to the individual's personal politics. Instead, in states where union security clauses are permitted, such dissenters may elect to pay only the proportion of dues which go directly toward representation of workers.

https://en.wikipedia.org/wiki/Collective_bargaining#United_S...

>involves building a whole new communication system.

Did this require building a new communication system? Comcast has a similar injection system for reminding you to pay your bill, or to notify you that you've been accused of piracy.

Maybe this was done using hardware and software they had lying around.

I get the impression that they did connect with yolo root access.

>Although the ACIS application required access to only three databases within the Equifax environment to perform its business function, the ACIS application was not segmented off from other, unrelated databases. As a result, the attackers used the application credentials to gain access to 48 unrelated databases outside of the ACIS environment.

Everything above and including #define C1 returns a single value, and therefore could be done with constexpr.

More to the point, constexpr would be a heck of a lot more efficient - changing M4 to M5 increases the preprocesed size from 80 MB to 240 MB. This is an exponential increase, despite that a Mandlebrot pixel can be computed in linear time. (Linear to the maximum number of steps.)

I suspect that this is because if you write a macro and use your macro argument twice, this doubles the length of whatever your macro argument was.