SomeStupidPoint

Not that person, but I objected to:

0 -- The outrage caused by misidentified profiles is larger than the outrage or societal damage caused by the existence of fake profiles.

...which was an unstated assumption underpinning why that entire argument would be an objection to implementing account review, and in my estimation is incredibly incorrect. They're being forced into Senate hearings over the false profiles, constantly being harassed by merchants about possible ad fraud over them, and are losing the trust of their communities. By contrast, the storm in a teacup over a few misidentifications is nothing to Facebook. (There's numerous other unstated, highly subjective assumptions as well.)

I also object to:

1 -- This is true, but irrelevant. Facebook cares about the rates of outrage in communities, not the mere existence. Ironically, this person seems to have fallen into the trap they called out, by thinking Facebook cared about instances of outrage (as the poster would). They also care about the relative outrage -- how mad are people compared to another option we can try. This point very poorly frames the problem as being about trying to stop any outrage instead of trying to minimize outrage.

4 -- This is community specific and untrue of the bulk of people. Decreasing error increases the vitriol involved, but it seems to come from a decreasing number of people as the outrage evaporates in the face of better error rates. So I think they're outright wrong here: it's like water cooling as it evaporates. The water is going away, even though it's cooler in the area.

5 -- This one becomes irrelevant (and/or wrong) if you correct the mistakes in 1 and 4.

Overall, this argument is --

a) irrelevant because even if it was correct, it wouldn't matter as it fails to contextualize its point by comparing levels of outrage between sources

b) wrong internally, in that it's largely irrelevant to the problem at hand or just outright wrong.

20 million reviews per year * 1/(2000 hours per personyear * 10 reviews per personhour) = 1000 people

Facebook could afford a thousand fulltime account reviewers, they just don't want to.

Seriously: a box of dice, some graph paper, and a little patience goes a long way in terms of security.

Your threat model almost certainly is one in which writing down a strongly-generated password is much more secure than anything else, if you take efforts to secure the written copy.

I feel like passwords are some kind of magical Wild West of crypto, where everyone has rolled their own and the "best practice" recommendations are often nonsense that ignores threat modeling and I'm just taking crazy pills for thinking we should treat it like other kinds of crypto where users only apply standard methods with verified entropy generation.

My point is that retard Rick could be the "central finite curve", and the other Ricks just hate the fact so much they made up a euphemism for him but don't kill him out of a fear of destabilizing the system of Ricks.

We don't really know what it is at the core, and so these girls could very well be the human "central finite curve", and not "off" it in any sense.

I was suggesting we not project our biases onto things like "central finite curve", eg that it's a curve which embodies "normalcy", until we see what role they actually play in the story.

> I wonder why they are considered two different persons at this point.

Because the system isn't well-defined in this case, which happens less than 1 in a million times, so isn't worth trying to fix explicitly.

The simplest solution is to give them two identities and move on with life having a little oddity -- as a pragmatic solution, it seems like the one that would end in the least litigation, as it confers an advantage not a disadvantage on the person(s) in question and can't possibly disenfranchise anyone.

It's not like you can't already create extra 'selves' (in most senses; eg voting is different) legally at the cost of $200 a pop -- what does it really matter if less than 1 in a million people happens to have two for what is otherwise going to be a very hard and likely short life?

That said -- I agree with you that it's not really well-defined here, because our notion of 'identity' is already highly problematic.

Edit:

This is a strange thought: what is the impact on someone with multiple personalities if you allow each personality to incorporate an LLC, all owned by a parent LLC (obviously, all legally owned by the 'person') -- ie, create an external legal framework that parallels the structure of the internal mental one, that all of the personalities should be able to view?

I'm honestly not sure if that's helpful for providing a lens to see the interaction play out or harmful for legitimizing the split. (And to some degree, all of us possess such multiple personalities -- we're rarely entirely the same person at work as at home or church or... whatever. I'd just argue that 'healthy' adults understand the interaction better.)

As for 3, I too am curious: the article made it seem like they normally do a 3-1, 1-3 split on limbs but can each do 2-2 if they want. From a control theory standpoint, that make sense -- you have one person focused on arms and one focused on legs, instead of an even split. (Probably not so different from being right or left handed.)

Is this accurate?

Also interesting: the one that controls more arms is the outgoing and high-strung one, while the one that controls more legs is the quieter and relaxed one. It seems unlikely that these facts are uncorrelated, though the causal relationship is anything but clear.

Incredibly, in an article full of pictures, they never actually tell you which twin is which and only address them as a unit of 'twins' in descriptions, which seems incredibly rude and depersonalizing to Krista and Tatiana. (Or, alternatively, never mention that the twins prefer to be spoken of as a unit -- which is itself an incredibly interesting fact.)

> It uses underspecified concepts in an attempt to sound smart, but all it does is fail to convey much meaning.

That sounds like it conveyed the meaning perfectly: they wanted to present an idea that struck you as smart, but which you didn't understand -- we're meant to see the show as Morty, not Rick. So we just get a Rick-ism that clearly has some structure to how the Ricks are organized (which the Ricks understand), but which doesn't immediately mean anything more to us. Rick and Morty isn't a science lecture -- we should expect narrative language, not scientific language.

It suggests that the Ricks are organized around some central, finite core of Rick-ness that we don't know about (notice that they have worse lives as they acquire letter variations on their curve numbers, eg TV personalities on the citadel). That's the point -- they don't want to get into a detail physics argument about it, but frame a (later revealed) structure among the Ricks.

I just don't think we should assume that structure is the shell of a normal distribution instead of something Fry-esque (ie, I'm-my-own-grandfather shit) -- we should leave them the room to insert whatever structure fits narrative-wise when we get to the point that makes sense in the story.

Netflix is my goto example of "do your core competency in-house, outsource the rest".

They do one thing themselves in their stack: their distribution network for content, and they do an incredible job of it. Every post I see about Netflix's CDN for video is insightful and a learning experience.

Then they throw a CRUD app in the cloud on top of it and call it a day. Okay, a little simplified -- there's still some neat tech in the DRM, in load-balancing the CDN, and in keeping all of their tech highly available. But conceivably, Netflix could retain much of their value by simply offering all the content to other websites that displayed it to customers -- the hard part of what they do is the CDN (and contracts with content owners), and opening their platform to other interfaces doesn't change that. (Heck, Netflix might be worth more if they opened their content to other interfaces, since they're not actually very good at the front end experience.)

But when I was doing consulting (about cloud stuff), that was my advice: do the core of your business yourself (eg, CDN) then offload as much of the rest as you can.

> far from the central finite curve of human experience

Do you mean anything in particular for this, or are you appropriating a TV show phrase to mean "unusual"?

(Just pointing out that the 'central finite curve' is something that has meaning to Ricks but isn't really a stand in for 'usual' in any meaningful sense in the context of the TV show -- I'd hate for what's potentially a cool sci-fi idea to turn into a euphemism for "the shell of a high dimensional normal distribution".)

Your theory doesn't explain why they don't retrain in welding, plumbing, electrical work, construction, vehicle maintenance, etc which are all more similar to mining (particularly within certain parts of each field) than they are office work.

At least in my state, many of the (re)training programs are for those and consist of a couple weeks of lecture followed by a few months of apprenticeship -- I suspect that they're simply sticking with the familiar while there's hope. People do that.

The empty classrooms is probably referring to the fact that every program begins with a little domain specific knowledge, so the people observing the intake look at that part of the funnel that every program goes through -- and then make comments about it from their perspective.

The article is a little deaf in saying "100 programs from coding to nursing". I'm guessing the fields are actually widely broader than that in the kind of work, since he found a program in coal mining among those 100, and I don't consider that between coding and nursing in any sense. (I think that left you with the impression they only train office or service work, for instance.)

Also, it turns out people in $industry who don't need $thing might know people who do.

If you leave a good impression, there's every chance they'll recommend you to someone who does need $thing, doing much of your selling for you -- you just have to leave them thinking "well, $thing wasn't right for me, but I'd do business with $person and $company".

Why would you want to do that?

I won't pretend I'm great at sales, but this is directly contrary to the advice I get, which is you want to filter out people who are a sure "no" or not a match (eg, on budget) as fast as possible so you don't waste sales resources on them while minimizing the chance of leaving a bad impression from trying pressure tactics on someone who wont buy.

What am I missing? (I guess it could be intending to be unethical, but I prefer to assume I don't understand.)

The problem is that malicious psychology has become so prevalent, that any sort of marketing formula that isn't direct honesty is parsed as manipulation. (And for good reason -- the vast majority of the time, it is.)

I don't care that your product is going to make me the reborn Buddha, because if you lead with that there's an overwhelming likelihood that your product doesn't stand on its merits and there is no underlying value. Blame the other marketers for spoiling the game if you want, but I think marketing is fundamentally a business mistake in that it makes you an adversary of your customers, not a partner. It's based on attacking customers with weaponized psychology. That's the premise of the modern field, and there's no escaping that.

If you had a product that I actually needed and provided value, you could just tell me what it was. It should be straightforward how that addresses my needs -- or it likely doesn't.

When driving, that probably doesn't really matter:

If the map and road sign disagree, just do what everyone around you is doing (ie, go with the flow) until you recover agreement. It's not perfect, but stretches out attacks to having to compromise long stretches of road and multiple vehicle types. If the cars for the next mile somewhere can find pace, the whole group can; similarly, if any kind of car can find pace, they can all use that to set.

Obviously safety checks and such, but human drivers only loosely follow signs and often get behavioral cues from other cars -- why can't automated cars do the same when they get confused?

Nuclear missiles, land mines, biological weapons, torture research, etc has all been carried out by willing and often motivated engineers or doctors.

I'm skeptical something that hasn't worked in other fields is going to fix our woes in software.

How does it get arbitrary requests signed?

The module only supported two interfaces:

1. Network -> Buffer, where it takes a packet with a particular structure and encrypted data and emits a signed plaintext.

2. Buffer -> Network, where it takes a request, result, and proof object and sends them out after signing and encrypting.

We were using it to front solvers that did a lot of work to solve constraints and emitted a proof object, so clients would send us requests (not our problem how they generate them) and then we had to show we did the right thing. The CPU didn't know either key, so it could either:

1. Compute the right thing, have results signed.

2. Compute something that doesn't match the signed request; have its faulty proof signed and returned. (Detected by the consumer when they verify.)

3. Fail to compute.

So this was guarding the case where a CPU was compromised and could possibly emit faulty (or malicious) results.

The point is that HSMs can allow for securing a computation chain if you can securely sign the root, even against compromised CPUs.

We didn't use Yubikeys, but I've used a hardware module to go from encrypted request -> signed plaintext request. A compromised CPU has no way to emit a new signed request, so it can't forge a a request + computation, only fail to compute or emit an invalid proof object.

I don't know about Yubikeys, but if they can sign their emitted plaintext, they could be used to similar effect.

It's more likely something that would make civilian smart tech useless against the top 5-10 militaries.

A military guidance system would overtrain a network using a supercomputer and then cryptographically make it stupider so they could give every missile 5 individually generated networks that vote, forcing attackers to compromise 3 networks per missile. (There may be some back and forth for a while, but here you can create a diversity of networks that makes the ensemble hard to attack.)

So they'll just shoot them down with lasers cause that's less energy intensive than running the compute node to crack missiles in real time.

Those regions are already evaporating, on the brink of collapse because of their policies:

Their kids leave then don't come back and slowly, family by family, whole groups are forced to abandon the region; they're net emigration regions, especially when you count rural versus urban (eg, the pattern becomes clear when you look at Georgia minus Atlanta).

Their migrated children tend to be more liberal, as urban populations are in general. Starvation is a wonderful motivator to change social order, and help mix society.

Now, I'm not even arguing this is a good thing, but I'm pretty powerless to do much when I consistently vote to help those regions develop new industry and they consistently vote to overrule that.

So, I hope they enjoy what they voted for: they won the debate, now they can have their prize.

I tend to think of it as: we go into a recession but not a depression because automatic + manual safety checks were deployed.

It's actually hugely troubling how much has been cut from government safety nets, because it seems liable to send the next recession into a depression and completely wipe out the very base who voted for those laws -- the poor South and the industrialized Midwest. I don't think rural America will survive the next depression with the social safety nets gutted: they've literally cut their own parachute lines, so they're going to crash not just fall.

But hey -- they saved a few years on taxes, right? Oh... wait... that was billionaires. Nevermind.

I've always been curious -- what's the best way for people in industry to help with this problem?

I'm not going to compete for a lottery ticket to take a huge pay cut to go work as a professor, with fewer resources and a worse schedule than even mid-level SDEs have. That's clearly not a workable solution for having the best and brightest train the next generations.

But I am concerned by the prospect of a brain drain in the training pipeline and am interested in helping with that. What are viable strategies to try and share some expertise back from industry?

Setting aside for a moment if my employer would agree or not, what should I even be pushing for as a collaboration/training program?

My (somewhat glib) opinion:

Marketers want quantity; salespeople want quality.

It's because that's what each one is paid for -- and so companies end up paying too much for marketers to undercut their salesteams' effectiveness because they misaligned the incentives, pay for marketing they don't need (or even harms them!), and then sit there wondering why it's not translating into sales (or worse, saturates salespeople with duds and costs sales).

I used work in marketing and my (slightly exaggerated) opinion is you should fire everyone on your marketing team, hire more salespeople, a socialmedia person, and maybe a single dedicated person to track overall sales performance.

But what's the point of marketing? You should be selling from that first contact -- and if you have two teams doing the same job with differing incentive structures, no wonder your organization is inefficient and pulling at different goals.

Intentionally denying 1 in 10 healthy adults their rights is illegal, no matter how much good you're doing by denying some smaller group their rights.

(It's also not clear that even if they have suicidal ideation that they're not entitled to gun rights, but I don't feel like having that debate today.)

My point is stormtroopers:empire::UP:P

You're correct that stormtroopers will never hit, wrong that stormtroopers are representative of the empire, and wrong about the correlation between stormtrooper hits and empire hits, because you're ignoring higher order effects (eg, while stormtroopers always miss, the empire must hit something to be the "bad guy" -- and they do: Luke's parents, Alderaan, Obi-Wan, Hoth, Luke's arm, etc.

But purely from the existence of stormtroopers consistently missing, you can infer there's a Darth Vader that must score some hits, or the plot wouldn't work. I think we might be in a similar case for P ?= NP, and you're all way too blasse predicting from stormtroopers.

Very little research has been done on P - UP, where UP is all polynomials with degree less than 10,000 and coefficients less than 10^100.

In fact, I don't know a single P-UP algorithm that wasn't explicitly constructed to be -- do you?

I just think most of us are subtly talking about UP, not P. (Including the blog post.) And that you've incorrectly inferred a "plot" when all you've observed is "redshirt fire".

How so?

Clearly there's a reason you think that isn't true -- perhaps you could share it instead of hiding behind just calling it "contentious".

It seems the stronger claim to suppose that the lower dimensions are rough and the top are smooth, especially when you think about where that roughness comes from.

Let's imagine a class "UP" which is the useful polynomial algorithms -- things we might actually compute.

The arguments raised here are a heuristic that UP != NP, which it certainly supports. My objection is that not only does that argument for UP != NP not support P != NP, it actually is evidence for P = NP, because UP seems to get asymtotically close to NP, and we know there are portions of P outside of UP -- suggesting we can find a "high" P region that sits atop UP and gets nudged over the line as UP approaches NP.

The suggestion that UP approaches NP but there's no P that crosses bears some supporting. Our disagreement seems to be over the obviousness of that relationship, and how much UP resembles P as a whole.

....Duh?

But it's absolutely insane to take the impossibility of low-constant, low-degree embedding as evidence as a general lack, given the rich structure of "large" numbers.

Which is what it's being used as in this post.

> Your "stress before breakage" argument is literally the gambler's fallacy.

It's not even anything close: the gambler's fallacy requires that the events you're trying to relate are uncorrelated.

The logical structure of P vs NP is anything but uncorrelated along its surface, and a rough surface along the low-dimensionality region is a good indication of a rough surface along the higher-dimensionality regions.

So -- the fallacy requires you be trying to correlate uncorrelated things; I am not. I'm correlating things that are going to be correlated because of how the spaces are constructed.

What's the argument here, though?

When we find a single bridge, all the complexity classes change, so we can only really divide the problems here into "ones we've found polynomial time solutions for" and "ones we haven't".

There may be something interesting about the ones we havent -- or we just didn't think of something obvious/a lot of special cases with high degree polynomials.

It's worth mentioning the near collisions happen in the low-dimensionality region of the border -- which doesn't have the same separating forecast, as it becomes easier to touch in higher dimensions. We can think of a sphere and a cube -- in low dimensions, a sphere lies inside of a cube; in higher dimensions, the surface bulges past the sides. You just don't see it until something like the 17th dimension. (Similarly, knot theory explodes with complexity in higher dimensions.)

Taking a "scientific approach" to asymptotic problems just makes the CS people advocating it seem vaguely hacky, since they're not even advancing a proper density argument -- it's just such a low-dimensional, poorly bounded argument that it's clear they didn't make a real calculation, they tried to dress up personal feelings in science.

You touch on something I had thought, but not said:

His argument seems to be "well, the border doesnt touch using under degree 100 polynomials, so it must never touch!"

I'm disinclined to believe that order 10^374738393874 polynomials can be accurately forecast by order under 100 polynomials, and suspect that portion of the border simply hasn't been examined at all or only in the most basic cases.

My point is that his heuristic has dropped my expected value for betting on P != NP, because his argument is one that suggests we will find a class, not that we won't.

His description of "passing tests" actually sounds like the buildup of stress before breakage, where many near-faults are generated.

I'm less confident of the resolution post reading that -- so it seems extremely weird to present as evidence that P != NP, as it shifted my beliefs the exact opposite way.