HNHacker News
TopNewBestAskShowJobs

SilasX

17,815 karma · joined December 6, 2011

blog: http://blog.tyrannyofthemouse.com

Github: http://github.com/SilasX/

email: sbarta@gmail.com

Location: Austin, TX (left the Bay Area August '18).

Studying infosec to make a new career out of it.

[ my public key: https://keybase.io/silasbarta; my proof: https://keybase.io/silasbarta/sigs/EIsQUp3Ln8z5fOgzWFj9alDI8Ci9Vl4JSgP5OZQcmjU ]

submissionscomments
SilasX··on The Math Is Haunted
I played through a lot of it, and while it was fun, I wouldn't call it gamified in the sense that Zachtronics does it. As an FYI, here are some sticking points I ran into as well:

1) It uses the term "goal" for the the current stage of the proof you're working on, which is counterintuitive. In common speech, the goal is the theorem you're trying to prove, not the current step you're on in proving it. It would be like if I were helping you get to work in the morning, and I said, "Okay, you're current [sub]goal is that you're done brushing your teeth." No. That's not a goal. That's progress toward a goal.

2) It doesn't warn you early on about the left-associativeness of the operator binding, or what to do about it. So you can apply what seem like valid transformations but then parentheses pop up that you don't expect, and which you have no guidance on how to deal with. I had to ask ChatGPT what was going on to solve it.

SilasX··on Amazon's AI Coding Revealed a Dirty Little Secret
This. Statements like the grandparents are in the general category of

- "life isn't fair"

- "people are bigoted against the outgroup",

- "brutal wars of expansion are a thing".

Like, yeah. Obviously. But that's supposed to be the kind of thing you push back against, when you don't like the result, not fatalistically accept as some fundamental invariant of reality. That's how progress happens.

SilasX··on The rise and fall of the Hanseatic League
When I was an exchange student in Germany, the host family had Hanse: Die Expedition[1] on their computer, which they explained to me but I also had to kind of feel my way through, not being completely fluent in German. You'd play as a merchant, sending out salt to other ports to trade for goods. Each turn was a year, and it would go through many generations, the goal being to contribute enough from your profits to build a cathedral.

Other things I remember:

- When you'd die, you'd pick an heir from your kids. They would be described as (sometimes you were supposed to understand as) being skilled at either fighting, exploring, or trading. If you didn't have an old enough one to pick, you'd have to pay a lot for some godparent to take over in the interim.

- You could raise capital by selling shares in your operation, but the more shares outstanding, the more interest (Zinsen) you'd have to pay. (You could also buy them back, but they'd be more expensive when you had a good year.)

- If you couldn't pay your obligations, you'd lose a turn to the Schuldturm (debtor's prison, lit. "debt tower").

- You could send explorers to find more lucrative ports, Novgorod being a good one.

- Depending on the weather, you might lose ships. Sometimes you wouldn't want to send any out, in which case it would warn you, "Keine Koggen auf der See! Zug wirklich beenden?" ("No cogs[2] on the sea! Really end turn?")

- You would get random events each turn, one of them being a soothsayer that would offer to predict your fortune. They would then give you the option to pay for extra advice, which would always be "Don't trust charlatans!"

- As your empire expanded, you'd be promoted in rank, some of the higher ones being Senator or Patrician.

- There would be battles, as nobles would sometimes attack your ports.

[1] https://www.mobygames.com/game/9273/hanse-die-expedition/

[2] A kind of ship used back then (also mentioned in the article as replacing the Knarr): https://en.wikipedia.org/wiki/Cog_(ship)?useskin=vector

SilasX··on Scientists may have found a way to eliminate chromosome linked to Down syndrome
Gideon’s Crossing:

“[A cochlear implant for my kid?] You think hearing people are better than deaf people?”

‘I’m saying it’s easier.’

“Would your life be easier if you were white?”

SilasX··on Scientists may have found a way to eliminate chromosome linked to Down syndrome
Right, there’s eueugenics and dyseugenics.
SilasX··on Air Force unit suspends use of Sig Sauer pistol after shooting death of airman
Not sure that's the best example, since Toyota did have the unintended acceleration issue, though I still don't get whether the consensus among smart people was that it was purely user error.

https://en.wikipedia.org/wiki/2009%E2%80%932011_Toyota_vehic...

SilasX··on Hyatt Hotels are using algorithmic Rest “smoking detectors”
Oh wow I ran into problems with those too. When I brought my cats to a Hilton, they would get the zoomies and run around at random in the middle of the night, which would make the lights turn on. I think I found some way to block the sensor.
SilasX··on Fully homomorphic encryption and the dawn of a private internet
There’s still Druge Report.

https://www.drudgereport.com

SilasX··on Bypassing Google's big anti-adblock update
Yeah, I've had some weird results from Firefox's spellchecker. It didn't recognize bachelorette or Shabbat, and it insisted on replacing "misclassifying" with "misidentifying". (Hm, doesn't seem to do the latter now.)
SilasX··on Final report on Alaska Airlines Flight 1282 in-flight exit door plug separation
If it makes you look good at the expense of being easy to understand, then yes, it's a problem.

Flowery writing that you have to read ten times to appreciate is great for novels -- not so much for effective technical communication.

SilasX··on Final report on Alaska Airlines Flight 1282 in-flight exit door plug separation
The problem of writing to look smart rather than communicate vital information.
SilasX··on Final report on Alaska Airlines Flight 1282 in-flight exit door plug separation
Well that’s one source (of many) where the problem is coming from.
SilasX··on July 5, 1687: When Newton Explained Why You Don't Float Away
It's relative. I operationalize "why is X true?" as "update my worldmodel to the point that X is not surprising". The typical way to do that is to show a more general rule that applies, and which implies X. But yes, you can keep asking the question about the more general rule.

Why do things fall? -> A special case of the general law of gravitation.

Why does reality adhere to the general law of gravitation? A implication of matter distorting the shape of spacetime.

Why is reality such that matter distorts space-time?

SilasX··on Don’t use “click here” as link text (2001)
I already kind of get this with extensions that let you browse from the keyboard (e.g. click a link without using the mouse) like Tridactyl. It lets me know when clickable elements are not detected as clickable, and, sadly, this happens quite a lot.
SilasX··on Don’t use “click here” as link text (2001)
In theory, yes. In practice, the typical specialist designer is going to optimize for things that come at the cost of accessibility.

But yes, in general you're absolutely right, that a good designer will take into account all factors, including accessibility. But the way that "design" has evolved in practice in means that the thing we all think of as "web design[er]" is not optimizing for it.

SilasX··on The Fed says this is a cube of $1M. They're off by half a million
Possibly, but I generally see it as a problem from pretty much everyone, especially specialists, speaking to people outside their domain, to be horribly miscalibrated about what jargon the layperson understands, or to just not give it any thought.

Obligatory xkcd: https://xkcd.com/2501/

So you don't need to explain it by any specific failing of the OP, it's common among the lucid as well.

SilasX··on The Fed says this is a cube of $1M. They're off by half a million
Separate comment so you can separately downvote/flag me:

Why, OP, why? How much self-awareness does it really take to realize JBR is a non-standard acronym people won't recognize? It almost feels like a superpower that I take an extra half-second to think about what jargon the average person needs to have defined.

SilasX··on The Fed says this is a cube of $1M. They're off by half a million
JBR = JonBenet Ramsey

https://en.wikipedia.org/w/index.php?title=Killing_of_JonBen...

SilasX··on How to negotiate your salary package
I wasn't making up excuses: the danger is real. If you don't have enough leverage to withstand the party walking away, you don't have enough leverage to ask for more. I didn't say "don't negotiate", or "it's too scary"; I'm saying, be ready for the consequences, and don't falsely zero out real risks. That shouldn't be controversial.

It's kind of telling that you have to resort to pathologizing me instead of addressing the concern.

SilasX··on How to negotiate your salary package
I didn't have to, that's just a prominent, citeable example that immediately came to mind. People get ghosted all the time for being too difficult, it just doesn't make the news. It can always make sense to end negotiations when you expect the other person to have unreasonable expectations and/or if you have a backlog of difficult alternatives you can pick.

If you're going to reply, please drop the absolutes -- you're asserting a confidence you can't possibly have here. "No incentive"? Of course there is, you just don't think it's big enough to worry about. So say that instead of asserting a model that can't be true.

"No downside"? Come on, reality rarely works like that.

SilasX··on How to negotiate your salary package
That’s just not true. They can withdraw the offer when you try to negotiate it. It’s bad etiquette, but nothing physically stops them from it.

https://www.forbes.com/sites/jmaureenhenderson/2014/03/13/wh...

SilasX··on A thought on JavaScript "proof of work" anti-scraper systems
From my understanding: to pose the problem for miners, you hash the block you're planning to submit (which includes all the proposed transactions). Miners only get the hash. To claim the reward, you need the preimage (i.e. block to be submitted), which miners don't have.

In theory, you could watch the transactions being broadcast, and guess (+confirm) the corresponding block, but that would require you to see all the transactions the pool owner did, and put them in the same order (the possibilities of which scale exponentially with the number of transactions). There may be some other randomness you can insert into a block too -- someone else might know this.

Edit: oops, I forgot: the block also contains the address that the fees should be sent to. So even if you "stole" your solution and broadcast it with the block, the fee is still going to the pool owner. That's a bigger deal.

SilasX··on Death of Michael Ledeen, maker of the phony case for the invasion of Iraq
Alternatively, for $2 trillion, you could probably solve cold fusion and synthesis of liquid fuels, making the Middle East's oil wealth effectively irrelevant.
SilasX··on AI's energy footprint
I’m assuming that calculation is amortizing the cost of running the water system (including wastewater treatment) adding the cost pumping it to the point that gravity can push it through the pipes. It’s never free.
SilasX··on Company Reminder for Everyone to Talk Nicely About the Giant Plagiarism Machine
>Human learning doesn't involve fixing anything,... AI training does involve fixing something into a fixed medium, which makes it disanalogous to human learning.

No, it breaks one part of the analogy, which has never been considered relevant: how long the learning persists. Yes, computers can store weights of a AI model much longer than any human could. But storing this "impact of having viewed a copyrighted work" has never been a factor in considering something infringing, regardless of how long it's stored. Courts don't consider it infringement if you simply use what you have learned from reading previous novels (the updates to your brain's neural weights) in producing new content.

Your argument is saying that the "fixedness of storing model weights" (aggregated high level understanding) can cause it to be infringement. That's without precedent. It would imply that if you're really good at "fixing" (remembering) the style of an author you read 50 years ago, it somehow crosses over into infringement because you "fixed that understanding into a medium" (your brain). That's not how it works at all.

>Among the reasons that Google removed outside access to its cached copies

I wasn't referring to the cached part of a site that Google serves to users, but the undistributed cache that they hold merely to know which sites to point you to, so you're not addressing the analogy. Here Google does store an exact copy (of at least some portions) and even then it's not considered copyright infringement until they start redistributing that content (or at least, too much of it).

My point was that acceptance of this practice even further bolsters the case that AI models aren't infringing, because, even if they did store exact copies, that generally not considered infringement until they start serving close-enough copies of the original copyrighted content.

SilasX··on Company Reminder for Everyone to Talk Nicely About the Giant Plagiarism Machine
>Either they aren't evil in which case they're being demonized, or they're already evil in which case demonization is redundant.

If you were trying to be charitable rather than clever, you would have read "evil corps" as "corps that the critic regards as evil".

>Keeping aside the motives of people, what is clear is that scale effects of AI cannot be ignored.

Okay, so just give some kind of standard -- any clear, articulable standard -- for how and why the scale matters. It's a cop-out to just rest your case on a hand-wavy "it changes at scale".

SilasX··on Company Reminder for Everyone to Talk Nicely About the Giant Plagiarism Machine
>The term "learning" (I presume from "machine learning") shoulders a lot of weight. If we describe the situation more precisely, it involves commercially exploiting literature and other text media to produce a statistical corpus of texts, which is then commercially exploited.

It's "commercially exploiting literature" in the same sense that an author would if they read a bunch of novels and then wrote their own based on what the learned from the pre-existing text. The whole point in dispute is whether that turns into infringement when an AI does it.

By labeling only one of them as "commercially exploiting literature" but not the other, you're failing to distinguish them in any meaningful way, and basically arguing from name-calling.

>It has little to do with "learning" as used in common English — a person naturally retaining some knowledge of what they've consumed. Plain English "learning" doesn't describe the whole of what's happening with LLMs at all! It's a borrowed term, so let's not pretend it isn't.

That's fair, that you can't just call them both "learning" and call it a day. But then the burden's on you to show how machine learning breaks from the time-honored tradition of license-free learning/"updating what you write based on having viewed other works". What's different? What is it about machine learning that makes it infringement in a way that it isn't when humans update their weights from having seen copyrighted works?

>What's happening is closer to buying some music cassettes, ripping parts of songs off them into various mixtapes, and selling them. The fact that the new cassettes "learned" the contents of the old ones, or that the songs are now jumbled up, doesn't change that the mixtape maker never had a license to copy the bits of music for commercial exploitation in the first place.

Okay, but (as above) to make that case, you'd need to identify where "acceptable" learning/"updating what you write based on having viewed other works" crosses over into the infringing mixtape example, and I have yet to see anyone try beyond "they're evil corps, it must be bad somehow".

SilasX··on Company Reminder for Everyone to Talk Nicely About the Giant Plagiarism Machine
>Human learning doesn't involve making a copy (or any other use of an exclusive rights) as defined in copyright law (the human brain not being a fixed medium), AI training does, because digital storage is.

That's just false -- AI models themselves only store parameter weights, which represent a high-level, aggregated understanding across all data that was learned on, i.e. what human brains do. This is clear from all the examples where you have to painstakingly trick them into producing exact text.

And even if they did store something that's "effectively a copy", that's no more copyright infringement than when Google caches a site in order to handle search queries. It's not copyright infringement until they start redistributing [non-fair-use] content from the sites.

SilasX··on Company Reminder for Everyone to Talk Nicely About the Giant Plagiarism Machine
Learning from copyrighted works to create new ones has never been protected by copyright[1], and has never needed separate licensing rights. Until 2022, no one even suggested it, to a rounding error. If anything, people would have been horrified at the idea of being dinged because your novel clearly drew inspiration from another work.

That narrative only got picked up because people needed a reason to demonize evil corps that they already hated for unrelated reasons.

[1] Yes, if you create "new" works from your learning that are basically copies, that has always been infringement. I'm talking about the general case.

SilasX··on Why did U.S. wages stagnate for 20 years?
That feels like sleight-of-hand though.

Previously (example numbers): Median compensation is $55k + health insurance.

After: Median compensation is $55k + health insurance.

"So nothing changed?"

"Wrong! Health insurance costs more for the same benefits, so really, you got a raise!"

Yeah, sorry if I'm not celebrating.

← PreviousPage 6 of 34Next →