HNHacker News
TopNewBestAskShowJobs

zardeh

342 karma · joined December 29, 2013

submissionscomments
zardeh··on Coconut – Pythonic functional programming language
True, but I've used python a lot, and I've used reduce maybe...twice? (well, twice that I can find on my github at least)
zardeh··on “autocomplete=off is ignored on non-login input elements”
SHOULD is not SHALL.
zardeh··on Typosquatting programming language package managers
True, but two simultaneous and specific typos is much, much less likely than a single one.
zardeh··on Typosquatting programming language package managers
Perfect, if each of those 10K hosts downloads the library 100 times you can now typo-attack the zope.event (working in python) library, which gets ~100 downloads per day, many of which are automated and so invulnerable to your attack. Your attack vector gets you, we'll say 1 new hit every 2 days at most, and likely only one a week or so (according to some math, on `requst` vs. requests)
zardeh··on Typosquatting programming language package managers
I know, I wasn't suggesting this as a namespacing solution, but instead a typo-prevention one.
zardeh··on Typosquatting programming language package managers
Well, you could also solve it by saying that the post slash names are unique. ie. There can't exist zardeh/serde if erickt/serde already exists. Then the author-name works as a logical checksum, and you aren't any worse off than you were with a global namespace.
zardeh··on Typosquatting programming language package managers
I feel like "pick the more popular package" is a good enough solution in this case.
zardeh··on DeepText: Facebook's text understanding engine
>No, I assure you that stockfish definitely does not look "all the way down (ie to states where the game is over)" in most positions, even with pruning - that would take too long. It uses a static evaluation function to evaluate positions without searching down the tree further

Ehh alright, it calculates the value of a node exactly (or exactly assuming its evaluation function). MCTS does not.

To be clear, what I meant, that you pulled out of context is that Minimax requires an exact calculation of the value of each child before choosing the best and returning it, this requires either simulating every possible game, or an evaluation function (depth-pruning). MCTS uses monte-carlo methods to estimate the best paths instead of trying all of them naively. At t=infinity, MCTS is equivalent to Minimax, because it will try all nodes. But it can run in time constrained systems better because it picks a much smarter order to run in.

>You described the "Monte Carlo" part of MCTS but not the "Tree Search" part. It's right in the name! See for example the first paragraph on page 20 of https://gogameguru.com/i/2016/03/deepmind-mastering-go.pdf, which clearly describes AG constructing a search tree.

yes. That has nothing to do with minimax though.

Which is what I said. AlphaGo uses pruning. AlphaGo doesn't use minimax with or without pruning, it uses MCTS (with pruning).

zardeh··on DeepText: Facebook's text understanding engine
No, stockfish uses minimax with pruning.

Minimax and MCTS are just different algorithms.

MCTS simulates games randomly and creates a distribution of expected value for each one based on some cool math. It estimates the value of each state based on a sample. Minimax actually calculates them all (or all of them except for some pruned ones).

That's the whole reason MCTS can work on Go and minimax doesn't. Minimax actually needs to look at all possible branches all the way down. MCTS can just say "well its been 100ms, let's use the estimates I have".

zardeh··on How to assign partial credit on an exam of true-false questions?
Simply state "The test will be (bell)curved". Now all students have the same goal: perform better than as many other students as possible.
zardeh··on DeepText: Facebook's text understanding engine
But that isn't minimax, its pruning by some evaluation function, where in this case the evaluation function is "blackbox Deep neural net trained on many games"
zardeh··on DeepText: Facebook's text understanding engine
Yup! Their additions to the algorithm are actually super simple, it was really surprising to me, its just very computationally intensive on the front end.
zardeh··on Jury in Oracle v. Google finds in Google's favour
>So yes, it most definitely isn't about copying code.

Exactly?

zardeh··on Gchat Was the Future of Messaging, but Google Didn’t Know
>Using one market (gmail) as leverage in another market (chat) is basically the definition of monopolization

No its not. Unless you're claiming that this is a form of product tying, but that would require that

1. Gmail held a monopoly or near monopoly over the email market at the time 2. hangouts was a service that people needed (or that to use gmail, you also were forced to use hangouts)

Neither of those were the case.

zardeh··on DeepText: Facebook's text understanding engine
AG doesn't actually use minimax, it uses montecarlo tree search, which doesn't even try to simulate a good opponent, it just plays lots of games very fast.
zardeh··on Gchat Was the Future of Messaging, but Google Didn’t Know
Wait are you saying that XMPP was forced out (solely) by Google, and that Google at some point monopolized the "instant messenger" market?

Not to mention that XMPP is a standard, not a competitor.

This is like saying that modern companies are monopolizing against SOAP or YAML.

zardeh··on Jury in Oracle v. Google finds in Google's favour
What?

This case had literally nothing to do with copying code.

zardeh··on SignAloud: Gloves That Translate Sign Language into Text and Speech
The people I'm talking about were also students, (University Hackathon). A group of four, I want to say 2 sophomores and 2 freshmen.

This was 2-3 years ago, so machine learning wasn't advanced (and even now, getting a functional NN running and trained in a weekend isn't feasible) so they only had it recognize a small set of commands, but it could do that.

zardeh··on Python Concurrency Decorators
Asynch and concurrency are very different. Python 3.5 allows first class aysnchronous calls, but concurrency is still "hard".
zardeh··on SignAloud: Gloves That Translate Sign Language into Text and Speech
I always find these cool, and then remember that I know someone who built a working (well working!) prototype of this in a weekend no almost no fanfare, from practically scratch.
zardeh··on Neural Networks Are Impressively Good at Compression
>If you count the number of connections on that last picture you will notice that there are fewer than there were in the first network. There were 55=25 at first, now there are 52*2=20. That reduction would be larger if I had more input and output nodes. For 10 nodes that reduction would be from 100 connections down to 40 when inserting two nodes in the middle. That’s where the compression in the title of this blog post comes from.
zardeh··on 20 lines of code that beat A/B testing (2012)
What I love about this example is that its the same algorithm applied in two completely different areas. This algorithm (or variants of it) can be used in place of A/B testing. The same algorithm can be applied to game playing and you get Monte Carlo Tree search, the basis of AlphaGo.
zardeh··on 20 lines of code that beat A/B testing (2012)
Indeed, otherwise you might be guilty of have Simpson's paradox either way.
zardeh··on Terra Bella
You also don't have to worry about things in geostationary orbit crashing into each other ;)
zardeh··on AlphaGo beats Lee Sedol again in match 2 of 5
Reading the paper, it doesn't at all sound like AlphaGo uses anything that TD-Gammon used.

It uses MCTS, which is unlike minimax. It doesn't use temporal difference learning, although they say that the policy somewhat resembles TD.

That doesn't sound like 'essentially built on', its sounds maybe like 'slightly influenced by'

zardeh··on AT&T sues Louisville to stop Google Fiber from using its utility poles
>and now we have a situation where a little girl gets raided for selling lemonade without a license.

Source, I'm actually curious?

But the real point is that yes, overarching government power is bad. But so is overarching corporate power. There seems to be this belief that the free market is perfectly self correcting, which in theory is true, but in practice isn't. There's a power differential between (large) corporations and individuals in their ability to organize, act, and finance their goals. As a result, large corporations will always be implicitly more powerful than groups of people, even quite large groups.

>And how does the government know more than the consumers? Are they privy to secret information?

Well, yes, essentially. A corporation has no reason to allow me, a private consumer to inspect it. I will have no way to verify that this cheese I'm getting is real or filled with sawdust. So you say, a corporation pops up that inspects the cheese and gives it a seal of approval. Perfect now I can trust my cheese. But how am I to know whether or not the cheese manufacturer is bribing the cheese-authenticating-company?

Or how will the cheese-authenticating-company even get off the ground, it needs to buy thousands of dollars in chemical testing and lab equipment. I have no incentive as a consumer to pay such a company if I think my cheese is fine. So the only chance such a company has of being successful is once consumers realize there is a major problem with the quality of cheese, by which point its too late.

Compare this to the FDA, who can say "let us inspect you and your food" and have the force of legal consequence for saying "no". Then, they're funded by taxes, meaning that even if I personally have no reason to pay them because I think my cheese is fine, I still pay a few dollars to keep the FDA around making sure my cheese is up to snuff.

>manufacturers are responding to consumer demand by removing transfat from their products without needing to be coerced by the government.

And the only reason we know about the contents of food is because of government mandated ingredient lists and labels. Without such government regulations, we'd be even more in the dark about the contents of food.

>which they are in a free market

Sure, if they are willing to spend time and effort to compare and investigate a (technically infinite in a true free market) number of products. Most consumers aren't willing to do this, even if they are able. Take open source software. I'm a user of various pieces of OSS. I've never audited any of the pieces of open source software I use. I could, but I don't have the time and am unwilling to develop the expertise to judge the python interpreter or LibreOffice or openSSL. And that's in an otherwise perfect situation, where I have full access to information. What about when I don't, or when there are 500 minor forks of openSSL that all do things slightly differently and I need to choose the best one for me. A perfect free market isn't feasible, there's too much information to work with.

I mean hell even if openSSL develops as the major one and everyone uses it, now we've got a near-monopoly, and someone who realizes that there's a problem in openSSL that they are unwilling to fix, and lets say its a fundamental issue, well now they have to build something new from the ground up, but it also has to match openSSLs api exactly because otherwise it will never be adopted, the cost of conversion is too high. And it'll take time for information about the problem to spread, and time to develop the alternative, and time for adoption to occur, and...this person needs to make money while developing this alternative somehow.

And that's in a situation with perfect information, extremely low market barriers, and a generally good natured monopoly who isn't acting to intentionally stump out competition.

>you need to clarify this.

Entering a market that has a natural economy of scale is nearly impossible unless you have a huge amount of resources already to throw at the problem. Take telecom. Telecom sucks in the US. We've broken up telcos because they're such natural monopolies, and awful. The only real company competing with telcos right now are internet based alternatives, and google, which is also an enormous company and piggybacked off of prior work by the telcos (dark fiber) to do what it is doing.

There's no way I could compete in that market, even if I promised services thousands of times better than current options. I don't have the resources to purchase 100ft of underground fiber cable, much less lay it.

zardeh··on AT&T sues Louisville to stop Google Fiber from using its utility poles
>This is the same argument as communism has never been tried argument and that Mao, Stalin et al were doing 'bad communism'.

No, sure some regulation is bad, but a lot of regulation is very good. For example, a lack of food safety standards led to terrible things for consumers in the early 20th century.

The problem with the whole idea of government regulation being always bad is that that's only the case if consumers have perfect information, are able and willing to choose any provider, and there aren't economies of scale.

zardeh··on Justice Department Wants Apple to Unlock Nine More iPhones
but 'me' is a personal pronoun
zardeh··on Discover Flask
So, ndg-httpsclient, mrjob, numpy, scrapy, monotonic and possibly others do have python3 compatibility.

Then there's a bunch of deployment tools like ansible, fabric, and marionette which don't really need to be python3 compatible for use.

Then there's tools like MySQL-python, python-gflags, python-cjson, INITools, filechunkio, which have python3 compatible alternatives, often in the standard library.

That brings the total up to 331

That leaves basically 3 categories of things left: mozilla tools, database connectors (many of which probably have python3 bindings though I haven't looked), and single sign on tools. Those account for 12 of the remaining 29 libraries.

That's pretty ready.

zardeh··on Issue and Pull Request templates
or like .dotfiles/ or .repoconfig/
← PreviousPage 5 of 6Next →