halflings

Clearly this is not a cryptographic hash, and hence it's known hashes are not uniformly distributed.

Apple explained in their technical summary [0] that they'll only consider this an offence if a certain number of hashes match. They estimated the likelihood of false positives there (they don't explain which dataset was used, but it was non-CSAM naturally) is 1 out of a trillion [1]

In the very unlikely event where that 1 in a trillion occurrence happens, they have manual operators to check each of these photos. They also have a private model (unavailable to the public) to double-check these perceptual hashes which also used before alerting authorities.

[0] https://www.apple.com/child-safety/pdf/CSAM_Detection_Techni... [1] https://www.zdnet.com/article/apple-to-tune-csam-system-to-k...

Apple's scheme includes operators manually verifying a low-res version of each image matching CSAM databases before any intervention. Of course, grey noise will never pass for CSAM and will fail that step.

The fact that you can randomly manipulate random noise until it matches the hash of an arbitrary image is not surprising. The real challenge is generating a real image that could be mistaken for CSAM at low res + is actually benign (or else just send CSAM directly) + matches the hash of real CSAM.

This is why SHAttered [1] was such a big deal, but daily random SHA collisions aren't.

[1] https://shattered.io/

I wouldn't have any regrets re:seeing something "similar" sold for millions of $.

Unless "heading in the right direction" means that you had a growing user base and/or solid revenue, in all likelihood the product you were building just wasn't succeeding. The idea in itself is not worth much: think Uber, Coursera, Airbnb, ... these are not novel ideas, many many people had the same ideas but didn't manage to make them a success (bad execution, or just bad luck/timing).

You seem to still be very very young, most successful companies are founded by people in their mid 30s, so don't despair just because a couple of your attempts were not successful. The best is still ahead of you.

This comment seems unrelated to the original story or the comment you're responding you, mostly reads like an ad.

From another one of your comments:

> I'm arguing that you can probably have both -- by using human moderators.

I don't think human moderators can review 720K hours of video per day. (on YouTube) You would need to hire >100,000 moderators to continuously watch videos all day and have some redundancy in reviews, then 1,000s of supervisors to make sure moderators are doing their job correctly. (almost impossible when you review videos all day with no break) Let's not even get into comment moderation...

> You can see it in Stackoverflow as well. The gamification encourages people to prove they are "worthy" of even deigning to ask a question by showing they've done "the work". Those who haven't are subject to humiliation and dismissal, regardless of need or authenticity.

I don't see how stackexchange's policy is related to the article, or "Protestant/Calvinist work-ethic". They are banning low-effort questions (e.g "File couldn't be opened" when the person just has a typo in a file path) because they devalue the platform, waste time that could have instead been spent on useful questions.

> I suspect that's because it's an export which they've accepted as their own because it happens to jive with the strict authoritarian hierarchies which are common in Asian countries.

Or they independently developed a glorification of hard-work from other sources. It's nothing new. All the Abrahamic religions do, and the same is true for secular ideologies such as communism.

If you use tensorflow, tfhub [1] is the go-to model repository.

[1] https://tfhub.dev/

Chrome is entirely removing third-party cookies by 2022.

https://www.theverge.com/2020/1/14/21064698/google-third-par...

That had nothing to do with "separation of church and state". It was banning private individuals to use a specific form of architecture. Pretty much any country in the world allows building mosques with minarets... did they threaten anyone's liberty / change their culture?

Going back to GP's point, the deeper issue is that people thought this was worth writing in the constitution. Other countries apply such laws in different ways, so I don't think the Swiss direct democracy is to blame here.

Thanks, I tried to clarify this in my original comment, but it can no longer be edited.

I'm not sure that changes much. To me, "BIPOC" is anything but inclusive. It's an acronym that apparently was used for about a month, by people of a very specific political ideology, and seems to be criticized by the very people this term is supposed to represent. [1]

And what does totalitarianism have to do with anything? I am all for putting a bit more power in employees' hands (esp. for issues where HR will throw an employee under the bus), but can we please keep the divisive political rhetoric out of it? Not shoehorn "BIPOC" and politically-charged language everywhere?

I am a "BIPOC", but I feel this rhetoric will be more effective with (mostly white) progressives/liberals, not the people it purports to represent.

[1] https://www.nytimes.com/article/what-is-bipoc.html

"Google employs more than 130,000 contractors and temp workers, a shadow work force that outnumbers its 123,000 full-time" [1]

130K+123K = 253K (article from May, likely closer to 260K now)

One of the main points this group is making is that they want to equally represent fulltime employees and contractors.

"All Alphabet workers deserve a voice: full-time employees, temporary employees, contractors, and vendors. We care for and support each other by striving for open and continuous dialogue among union members." [2]

[1] https://www.nytimes.com/2020/05/29/technology/google-rescind... [2] https://alphabetworkersunion.org/principles/mission-statemen...

On this group's homepage [1]: "We are BIPOC workers who fight against totalitarianism."

I am not sure what this has to do with unionizing Google workers. Using obscure acronyms to arbitrarily focus on race and ethnicity (something frowned upon here in Europe at least) + the hyperbole of "fighting totalitarianism" doesn't give me great confidence that this is representing workers, as much as it is pushing the ideology of a vocal minority.

If they truly believe they represent the majority of Google employees, then they are free to widely share this call internally and see how many people actually agree with them and want to join their group. So far, it's ~0.2K out of ~260K employees.

[1] https://alphabetworkersunion.org/

> By July I realized a couple things (and adjusted my investments accordingly)

You're definitely making valid points, but this sounds like a whole lot of hindsight bias: you're rationalizing something that was rather unpredictable (the full impact of the pandemic on the economy, and whether that would surpass the market's expectations or not).

How did you possibly adjust your investment strategy based on these "learnings"? It sounds like this can only lead to the same forms of losses you've had with your initial risky investment (overly confident that pandemic = stock market crashing).

> - Population cap, in SC2 the population is capped at 200 supply, this is an inheritance of SC1 technical limitation, but it has a strong impact on SC2 late game, not for the best as players tend to invest in defensive structures to compensate, leading to potentially static/ boring games.

I actually find the population cap to be super useful, and serves the opposite purpose: Without a supply cap, the player with the better economy can just turtle up, defend their bases with strong tanky units, while they build a gigantic army. The supply cap forces them to move out and attack, trading units.

Any service running way over quota could break all other services. (at the end of the day, resources are physically limited to what you have in your datacenters)

Quotas are one way to isolate this impact to that service in particular.

Of course, when it's a critical service like authentication, it hardly isolates anything... but I can't think of a better alternative.

How would YouTube take a cut of this ad revenue? It's not disclosed anywhere on the YouTube platform.

> does not make any reference to costs or benefits to other users of the web platform

> arguments for are being presented publicly in that thread while those against are being discussed elsewhere (perhaps at Google?)

It's right there at the end of the highlighted comment:

"we would love to [...] replace them with something that generates less security bugs. Increasing the capabilities of XML in the browser runs counter to that goal."

So security bugs in XML seems to be the main issue?

YouTube Premium covers both ad-free YouTube viewing and YouTube Music. (so that number is likely for both)

> I’ve stopped using YouTube because of the crazy ads.

Why not pay for YouTube Premium? At the end of the day, you need to pay in some way to fund the creators making new videos for you to enjoy (+ YouTube for hosting one of the biggest services on the internet).

With a sample size of 7, do you really think the issue is that it's not double-blind?

This looks like the first step towards larger studies.

Also remember that given this was quite experimental and could've easily had adverse health consequences, it had to be on a volunteering basis.

Notable exception: The Wire.

5 seasons, each covering a specific topic (politics, drug dealing, media) beyond telling the story, and when they ran out of topics and told the entire story they had, they just wrapped it up... they could've easily milked it and made 5 more meaningless seasons but luckily didn't.

> what about the hundreds of thousands of other objectionable groups on Telegram

Exactly! If they want to start policing all content generated through apps that can be installed on iOS... they'll start getting millions of takedown requests from governments around the world.

Very weird decision strategically, and ofc very questionable morally.

> The main problem is not the idea of God (imho god = love)

Well there goes your problem, you're countering something that these religions don't claim in the first place: None of the Abrahamic religions (Christianity, Judaism, Islam, etc.) claim that God is pure love, butterflies and spreads happiness on earth.

Their scriptures explicitly talk about how life on earth is (also) filled with pain, disease and man-created problems.

I'm not sure why this criticism ("What kind of god gives kids cancer") on religion is so common in Western countries... it's a giant strawman that a simple question to a believer of such religion would dispell.

The biggest issue in proto3 is the missing "has_<field>" functions (e.g no distinction between a field's default value, and a special <missing> state).

There are suggested workarounds for this, but I'd just stick with proto2 for the time being as proto3 is still WIP.

I mentioned that you can do this on top of saving money. Just saving money with no goal in mind (or with the goal of buying a $1M house) is IMHO not always worth it.

Absolutely for saving to be able to survive ~1 year as life is very unpredictable, but we're talking about altruism here... Let's for a second assume we all want to have the largest positive impact on other people, is that $1M really the best way to spend your money to both improve your and other people's life? Could you instead buy a house/work elsewhere, pay $400K, and invest $600K in saving a hole bunch of people from certain death / extreme poverty?

There are many beneficial things people can do but are not doing. For example: investing your money. Many people aren't doing it (instead leaving cash in their bank account for a large part of their life) out of pure laziness/ignorance of its benefits.

Let's assume you're a junior SWE in the Bay Area, making $130K. Annual take home is ~$85,753 according to [1] You pay $3K in rent (nice 1 bedroom apt in South Bay, or just a decent one in SF), $2K for various expenses, that's $60K and you're left with $25K.

This is for a junior engineer, and granted they might want to save $25K vs $15K (arguable; I would consider those savings don't matter much in the long run, the goal being to quickly increase your salary by advancing your career), but an engineer in a large tech company is making closer to $250 - $300K total comp, and you'll see that the $10K donation isn't making much of a dent in your savings at that point.

[1] https://smartasset.com/taxes/california-paycheck-calculator#...

Shouldn't it be a % of (income - essential expenses), rather than % of non-essential spending?

If you're very frugal, you'd end up giving away a very small amount of money, saving all the rest... Saving is super important, but giving money to effective charities is also a great way to use your money.

To take a typical example of someone that might be reading HN, any software engineer working in NY or the Bay Area can easily give away $10K a year while saving enough money to get a very cozy retirement.

This assumes Feeney is more qualified to know where his money will be most effective and that no one else is out there already doing a ton of research validating the effectiveness of different charities. (both questionable)

I think this could be a convenient excuse that keeps some from donating more money (I'll wait until I can spend time creating my own charity / doing some hands-on vetting on who gets my money). There are already plenty of organizations (GiveWell, OpenPhil, etc.) that have very good research on the impact different charities have, often with RCTs on those interventions proving how much each $ buys (in terms of quality-adjusted life years or else).

This is amazing. Thanks for sharing! Didn't know about netron.

If I send you links to 2-3 rare videos and you view them, I can probably uniquely identify you (or restrict the list of potential users to ~5 users then recognize you based on other videos you would have watched). I then have access to your full video watching history.

Even without this, I could just check videos you shared on Twitter, follow the same process, and have a decent chance of identifying you.

Anonymization is very very hard. Maybe federated learning is more promising here, keeping all your data privately stored on your own devices.