HNHacker News
TopNewBestAskShowJobs

bnjmn

477 karma · joined July 6, 2011

[ my public key: https://keybase.io/benjamn; my proof: https://keybase.io/benjamn/sigs/94-u377AG0oTnRb3A88rmP2WOk88eq9DrT9K2RGfsJk ]
submissionscomments
bnjmn··on Unprovability comes to machine learning
Please see my reply to @baddox.

You say you do math and not philosophy, but the whole CH independence proof kinda moves this discussion out of the realm of (ZFC-based) math, doesn't it? If you don't want to consider philosophy here, there's not much more to say.

bnjmn··on Unprovability comes to machine learning
> Now, you might mean that most mathematicians suspect that adding CH to ZFC is more intuitive/natural/useful than adding !CH to ZFC, but that's a quite different statement.

I guess I'm a constructivist about this.

You can't just wave your hands and take !CH as an axiom. Remember what that means: you reject that there is no set with cardinality between the integers and the reals, so you're implying there is such a set. I realize not all mathematicians share this point of view, but if you can't somehow construct the intermediate set you're claiming exists, or show that it can be constructed, then you don't really have any grounds for taking !CH to be true. Without a positive reason for believing such a set exists, I have a hard time pretending that !CH could possibly be a "useful" (your word) axiom to add to ZFC.

By contrast, imagining that CH might be true is easy. You don't have to construct any never-before-seen mathematical objects to make your point. You just presume, without fear of being contradicted by ZFC, that the integers are countably infinite, and the reals are uncountably infinite but only sort of "minimally" so (there are higher orders of uncountable infinity, but none lower). You may never be able to prove CH (certainly not with ZFC, and almost certainly not with any other logical system), but at least the model makes sense: there's no cardinality in between that of the integers and the reals.

Which is not to say that CH is particularly useful! In fact, maybe the most interesting thing about the paper is the idea that the CH could be meaningfully connected to anything in an applied form of math like machine learning. Does that weird connection between CH and ML have any practical consequences for ML in general, or the EMX learning algorithm in particular? Probably not, especially since assuming CH is true just implies EMX should work, so it's fine to keep using it to solve learning problems.

In other words, if you wanted to convince an ML practitioner that their learning algorithms might be unreliable because the Continuum Hypothesis might be false, I think they would be within their rights to request you show them an actual set with cardinality between the integers and the reals. When you couldn't, they would then be within their rights to ignore your objection.

bnjmn··on Unprovability comes to machine learning
I suspect the Halting Problem was already an obvious theoretical limitation to machine learning. Just imagine a network trained to classify programs according to whether or not they halt. Then construct a program that runs forever if the network says the program halts, and halts if the network says the program does not halt. No training strategy would ever produce a network that always gives the correct answer in this case.

In the Generative Adversarial Network (GAN) framework, this would be the One Weird Trick the generator could pull that would always foil the discriminator.

This paper is interesting because it connects learnability to a different kind of undecidability than the Halting Problem: the Continuum Hypothesis. In some ways it's more troubling than the Halting Problem, because at least we can usually figure out whether some actual program halts, using some cleverness, but no one can begin to say whether the Continuum Hypothesis is true or false.

The Continuum Hypothesis is more like P!=NP, in that most mathematicians suspect the CH is true (that is, there's no cardinality in between the integers and the reals). That's good news, because it implies the EMX learning model will work:

> Their results imply that the finite subsets of the unit interval have monotone-compression schemes, and therefore are learnable in EMX, if and only if the continuum hypothesis is true, which is known to be unprovable.

If someone comes along and casts doubt on the Continuum Hypothesis through some mystical but persuasive line of reasoning (news flash: not gonna happen), then there might be some EMX functions that can't be learned, but no one is going to lose sleep over that. There are lots of more immediate reasons why machine learning fails in practice, like not having good training data.

bnjmn··on Introducing Hooks
You're not alone! https://github.com/reactjs/rfcs/pull/68#issuecomment-4331815...
bnjmn··on The Big Hack: Statements From Amazon, Apple, Supermicro, Chinese Government
Ah you’re right!
bnjmn··on The Big Hack: Statements From Amazon, Apple, Supermicro, Chinese Government
This sentence, wow:

> Two of Elemental’s biggest early clients were the Mormon church, which used the technology to beam sermons to congregations around the world, and the adult film industry, which did not.

bnjmn··on Intel discloses new L1 Terminal Fault vulnerability
Another speculative execution side-channel cache timing vulnerability, a la https://meltdownattack.com

Related disclosure from DigitalOcean: https://blog.digitalocean.com/a-message-about-l1tf/

Intel is evasive about the precise attack scenario, but DigitalOcean reports "an attacker could theoretically use one Droplet to view another Droplet’s memory," implying that the side-channel is invasive (that is, theft of data is possible) rather than cooperative (that is, data can be intentionally leaked between processes, a less harmful sort of side-channel).

bnjmn··on Model Zoo – Pretrained deep learning models
Could this kind of model registry enable efficient model search based on input/output queries, a la this paper about semantic code search? https://kstolee.github.io/papers/JSS2015.pdf

Perhaps the registry would need to be a lot larger before semantic search would be really useful?

bnjmn··on Meteor 1.7 and the evergreen dream
Disclaimer: I'm a core maintainer of Meteor.

One thing that makes this release tricky is that the differential modern/legacy bundling system sounds so much like something that any modern JS application must be doing already, but no! I've yet to come across any framework besides Meteor that makes differential bundling possible, let alone easy.

I'm happy to answer any questions, and I'd love especially to hear about any similar systems that you've used, heard of, or even just imagined. If you're involved with another full-stack web framework, we should chat about how you might implement a system like this that works for your developers.

Everyone in the JavaScript community should be able to use native ECMAScript features directly, without the layers of compilation, and without sacrificing support for older browsers. That's the future we've all been hoping for, so I hope we can share our best ideas across frameworks and between the various JS sub-communities.

bnjmn··on Node.js 8.10 runtime now available in AWS Lambda
Wow, I commented about this on the blog post, and they apparently deleted my comment. Cool.
bnjmn··on Node.js 8.10 runtime now available in AWS Lambda
This update was a very long time coming, so I'm grateful that it finally happened.

Still, it's disappointing that they couldn't change course (ever so slightly!) to accommodate the 8.11.0 security update from a week ago: https://nodejs.org/en/blog/release/v8.11.0/

Really hope they'll be keeping up-to-date with incremental Node 8 releases, now that they've (almost) caught up with the times.

bnjmn··on The Best Books on the Philosophy of Mind
For what it's worth, Daniel Dennett spends considerable time in _Consciousness Explained_ (and his other books) deconstructing both why Searle's Chinese Room argument is misleading, and also why is seems intuitively appealing (he calls it an "intuition pump"). I suspect you would find Dennett's treatment of Searle very much aligned with your skepticism!
bnjmn··on Don’t Tax Options and RSUs Upon Vesting
Thanks for the thoughtful feedback!
bnjmn··on Don’t Tax Options and RSUs Upon Vesting
Here's the statement I plan to read to my representatives:

I live in Greenpoint, Brooklyn. I work for a small startup of less than 30 employees.

Three years ago I left a large, publicly traded tech company to take this job, because of the potential I saw in the work the startup was doing.

I took a significant pay cut when I joined the startup, a decision that was justifiable only because of the distant future value of the incentive stock options that the startup promised me.

It is my understanding that the current tax bills on the floors of the house and the senate would tax these options when they first vest, long before I could possibly sell them to cover my tax bill.

If I am wrong about this, I would appreciate any clarification you can provide, ideally written into the text of the bills themselves.

If my understanding is correct, I honestly don't know how my startup, or other small companies like it, will be able to compete with wealthy public tech companies as we try to hire new employees in an already extremely competitive market.

I struggle to understand how draining talent from the most innovative small tech companies can possibly be construed as a good thing for our economy.

Thank you for your time.

bnjmn··on How We Built the ARKit Sudoku Solver
This! Still, admittedly, this is a drawback of my strategy.
bnjmn··on How We Built the ARKit Sudoku Solver
Since the author spent so much time on the optical character recognition step, it's worth mentioning that you don't even really need OCR for this task.

You can just find the squares, group the characters in the squares into visual equivalence classes, assign each class an arbitrary number, solve the puzzle in terms of those numbers, then fill in each empty square with the (average?) image of the equivalence class it matches.

This would allow you to solve a Sudoku puzzle with letters or WingDings instead of numbers, and the output font would naturally match that of the original puzzle.

bnjmn··on Meteor allow/deny vulnerability disclosure
It sounds like you've built some interesting technologies!

Some food for thought:

* RethinkDB was a startup that both collaborated and competed with Meteor, until it [went defunct](https://rethinkdb.com/blog/rethinkdb-shutdown/) last year, leaving its open-source technologies to be maintained by the community.

* Your gun project depends heavily on graphql-anywhere, which is a project maintained by none other than... Meteor Development Group.

Given the ways in which you've benefited from the community in general, and MDG in particular, I'm surprised you took this zero-sum, winner-take-whatever, nakedly self-promotional attitude. Let's be better than that?

Disclosure: I work on Meteor.

bnjmn··on Scala Native v0.1
Just what I hoped to hear, thanks!
bnjmn··on A Fight to Fix Symplectic Geometry
Yes! Wayyy too clickbaity without the niche adjective.
bnjmn··on Lake Nyos suffocated over 1,700 people in one night
I'll take "horrifying and lovely" any day!
bnjmn··on Lake Nyos suffocated over 1,700 people in one night
you're welcome :)
bnjmn··on Lake Nyos suffocated over 1,700 people in one night
I wish I'd known about this image when I wrote the poem I posted above!
bnjmn··on Lake Nyos suffocated over 1,700 people in one night
Thanks, that does mean a lot to me. I wish I had more to share, but writing poems takes such a long time, and my day job (programming) seems to deplete my motivation for other kinds of writing. But those are just excuses.
bnjmn··on Lake Nyos suffocated over 1,700 people in one night
A few years ago I wrote a poem loosely based this event.

I know unsolicited poetry from strangers on the internet is almost always awful, but this poem still holds up for me, which is pretty unusual for anything I've ever written.

So I hope you enjoy it, too:

  NYOS

  You took me in on dusky breath,
  tasted me, tasted nothing,
  gathered by my easy take
  that I was oxygen enough
  for idle inspiration.

  How swiftly my lack became your
  lack; my misgiving, your mistake.
  Your eyes flashed a baffled
  petition as you fell limp in a
  thousand different doorways,

  cribs, embraces, fits and fields,
  yet I pressed after whatever it was
  I thought to find in the lowest
  parts of Cameroon, as foolish in
  love as a gas trapped in a lake.
bnjmn··on A Wait-Free Stack
Make sure you check out Appendix A at the end of the paper (Asymptotic Worst-Case Time Complexity), in case you imagined the name "stack" implies constant-time push/pop performance. This data structure is only a "stack" in the sense that it provides last-in-first-out access.
bnjmn··on Inky the octopus breaks out of New Zealand aquarium
In 2016, Inky Dufresne escaped from Shawtank Prison. All they found of him was a muddy set of prison clothes, a bar of soap, and an old rock hammer, damn near worn down to the nub.
bnjmn··on Introduction to Boosted Trees [pdf]
Perhaps I am, but the problem of accidentally embedding stereotypes in machine learning models is all too real:

https://www.fordfoundation.org/ideas/equals-change-blog/post...

http://www.pri.org/stories/2015-12-06/are-algorithms-racist-...

bnjmn··on Introduction to Boosted Trees [pdf]
It actually does bother me, and I wish I had the time to fix it, but fixing it involves much more than the comically large amount of work you're asking of me: it means convincing this PhD student that he made a genuine mistake that betrays his sexism, and getting him to acknowledge that mistake by fixing his own slides.

I wish the instructor was more aware of his own sexism when he created this material, so that neither he nor I nor anyone would have to fix this problem now. He chose this careless example, and so it's his responsibility to fix it, no matter how busy he may be as a PhD student. We're all busy, but these slides have his name on them, and demonstrating that he knows better now is something only he can do.

bnjmn··on Introduction to Boosted Trees [pdf]
In the first example, being male is one of two features that predict playing video games, and (surprise!) only the boy and the old man are classified as gamers. Talk about casual sexism! Can you imagine taking this class as a woman (who maybe, just maybe, happens to enjoy video games) and having to forgive/ignore the instructor's cluelessness in order to get through the material? So incredibly tone-deaf and lazy, ugh.
bnjmn··on Meteor 1.3 Released: ES2015 & npm Support, App Testing, Mobile Improvements
As one of the people intimately involved in this release, I completely agree with the "stuff that should be there from the beginning" part of what you said :)
← PreviousPage 2 of 3Next →