HNHacker News
TopNewBestAskShowJobs

maldeh

519 karma · joined September 6, 2013

submissionscomments
maldeh··on Adam Neumann will determine WeWork's fate
Benchmark is a major investor, actually. But it sounds like they might be getting the short end of the stick on this deal, like almost everybody else.

This case could be quite different from Uber per the thesis of this article by Stratechery: https://stratechery.com/2019/neither-and-new-lessons-from-ub...

While Benchmark was looking to preserve its assets from being squandered on possibly the startup-catch of a lifetime, Softbank on the other hand relishes being the Big Stack Bully, leaning hard on their portfolio companies to aim for home runs everytime, and their excessive leverage allows them concessions from companies most VCs can only dream of.

maldeh··on And 'Lo!': How the internet was born
Also recommended: "Lo and behold! Reveries of the connected world" by Werner Herzog.
maldeh··on Naming and shaming companies who kowtow to Chinese censorship requests
I wonder what sort of allegation it would take that could top what's already been attributed to this administration.
maldeh··on Facial recognition: School ID checks lead to GDPR fine
I fully understand your point that GDPR hasn't been good for investment or startups, and that that the likelihood of a startup succeeding to reach sustainability in the european field has significantly diminished. What I'm rejecting is that that's necessarily an undesirable state of affairs. Does the world really need the likes of Blur or DiscountMugs to succeed, when they have proven woefully incapable of protecting the most basic forms of user data?

The tech economy seeing such upheaval right now could be construed as a signal demonstrating how dependent it was on fundamentally unhealthy and untenable data practices that were previously endemic to the industry.

I'm sorry that you have had to spend a tonne of money to attain GDPR compliance. I imagine most "incumbents" have had to spend a good deal as well; I can only hope that the next generation of companies have learnt from your company's mistakes and to structure their data processes from day 1 to avoid accruing such sensitive data in the first place.

At any rate - a tech sector is possible. A thriving one, that can sustain as much employment? Maybe not quite, there'd have to be some adjustments; but the people would be better off. A tech sector with the same market cap? Unlikely, but we need to get over ourselves and question if preserving the techocracy's wealth is more important here.

maldeh··on Facial recognition: School ID checks lead to GDPR fine
That presumes all early stage startups (as made in the US) are created equal, and any regression in their success rates is bad.

Given that 9/10 startups fail even without GDPR, it's not surprising that early stage cos form the Lion's share of failures, and it surely can't be good for any data that was slurped up during whatever these experiments at market fit were doing.

And given that the ultimate goal of GDPR is to protect privacy, it doesn't make sense to exempt startups, especially when the early stage stakes are high and a failure to squeeze out every drop of value legally possible out of your data (while your competitors do) could mean the death of your venture.

As such, comparing to American standards or even current European standards doesn't quite work when there's a clear shift of the moral bar for GDPR compliance.

maldeh··on Paternity Leave Was Crucial After the Birth of My Child
Whoa nelly, that is a tough argument to make in the face of a global rate of natural increase (RNI) of 1.06 (that's 6% being added year-over-year on top of the last census population of the 7.53 billion we've measured in 2017).

If you consider making sure your own particular genepool sustains, as your prime directive, then sure, power to you. When it comes to sustaining humanity, then there is something to be said about sustaining global resources, relaxing supply, improving literacy levels, maximizing public funds, human happiness indexes, etc. that have all long correlated lower population densities with better outcomes.

maldeh··on The fungus that devastates the Cavendish banana has now arrived in Latin America
Not sure of the specifics here - one reconciliation of the two statements could be, that Cavendish proved to be a good transport vector for this particular strain of fungus to spread rapidly on a global scale, notwithstanding its broad virulence.
maldeh··on A Cold Take on IBM, Red Hat and Their Hybrid Cloud
How would hosting the site partially on-prem make things better?

Or do you mean "multi-cloud"?

maldeh··on H-1B visa: Government says work ban for H-4 spouses coming this month
There's currently a 32% selection rate on the H-1B lottery, regardless of whether your application's even eligible or not. The best case (with OPT extensions right after college, not applicable for already-working people; and also assuming selection rates don't further decline) is 69% selection rate after 3 attempts. I don't imagine hundreds of thousands of H-4 holders like those odds.
maldeh··on Open-Sourcing Ax and BoTorch: New AI Tools for Adaptive Experimentation
Cool, MIT licenses this time.

* https://github.com/pytorch/botorch/blob/master/LICENSE

* https://github.com/facebook/Ax/blob/master/LICENSE

It's interesting seeing Facebook's evolving open sourcing strategy, and a bit heartening not to find they're not still pushing the BSD+Patents licenses on everything.

maldeh··on The Fargate Illusion
Fargate's purpose was never to magically make container orchestration go away, though - it's just a managed compute layer you plug into via the other orchestration frameworks (ECS or EKS). Really the apples-to-apples comparison should be DigitalOcean's managed Kubernetes with EKS (and there are a lot of valid points to raise there, to be sure).

The real value-add of Fargate is pay-for-what-you-use, when your task queue is idle and your bill drops to zero with no extra orchestration (something that the article seemingly completely missed).

maldeh··on Bringing the best of open source to Google Cloud customers
As I see it, we have these open-source / core products, and the companies maintaining them are looking to provide certain value-added features (particularly critical to the enterprise crowd) in exchange for some vendor lock-in.

- Amazon's tactics are to either offer a proprietary alternative to those enterprise features (Elasticache, DocumentDB), or to vend a competing open-source alternative with the aforementioned features (ElasticSearch, Corretto)

- Google's approach looks to be partnering with the maintainer to offer a shared walled garden (so while you're still locked-in, you can easily migrate between blessed partners).

I'm not yet convinced GCP's tactic is overall good for opensource as a whole, since they are setting themselves up as de facto benevolent gatekeepers; although admittedly they will allow these projects to thrive under their current stewardship in the medium term.

Edit: I'm also interested in seeing what happens here when there are multiple competing vendors who have equivalent claims to stewardship of the underlying project - ex:- Hortonworks and Cloudera pre-merger, or Cloudera vs. MapR. Also, how vendors with vast proprietary ecosystems work out in this "native fully managed" setting, ex:- Databricks which has managed to get on by on the Azure and AWS marketplaces.

maldeh··on Meta-Reinforcement Learning
While a contextual bandit could be learning using more parameters about the environment than the multi-armed bandit posed in the example, it still has the goal of finding the best reward given the specific domain (while by and large using the same strategy as the multi armed bandit). But placed in a different context, the state space and underlying model parameters could be completely different, so the previous reward for any given state, or sequence of states, could be irrelevant.

The goal here is to reward the agent for the search strategy they employed to arrive at their answer, not the quality of the answer itself.

One possible use case (directly related to their example with multi-armed bandits, possibly learnt by a contextual bandit but requires a good deal more modeling) could be retail pricing, where different categories of products have drastically different demand curves. A meta-algorithm has the promise of generalizing better and rapidly arriving at the optimal pricing across a wide range of similar price curves.

maldeh··on Microsoft says encryption laws make companies wary of storing data in Australia
Assuming it could be proven, surely?

And there have to be limitations as to how far an individual could go as to subterfuge, so if your company enforces a 2-person code review and there aren't other authorized Australian nationals at hand, you could point at process preventing you from doing so without others' knowledge (how naive this defense is, I have no idea)

maldeh··on AWS API Performance Comparison: Serverless vs. Containers
The author didn't spend much time discussing the application load balancer, wonder if it was the source of misconfiguration (besides their http server itself).
maldeh··on I tried to keep my unborn child secret from Facebook and Google
"If you don't want period-related ads, then you must not want to be treated like a woman?" That feels like a bit of a leap.
maldeh··on Google Cloud Firestore NoSQL database is in GA
That would be a really weird user experience given how long this has been in Beta. Google has had over 1.5 years to get access control right, so I bet (hope) this is not an issue for new users.
maldeh··on Google Cloud Firestore NoSQL database is in GA
By "In these kind of systems, your app users also become database users", do you mean by RBAC or some other session-based control on the backend, or does Firestore perform some intelligent user entity resolution I'm missing?

As far as I can tell, any naked frontend-only client serving content without its own access management sounds quite open to a single bad actor spawning a large number of sessions on the same host, and on a small number of distributed hosts, to eat up all available resources. I'm not inclined to ever trust a number of users behaving nicely at n>100 personally.

Given that Firestore appears to advocate "serverless" client-driven applications, unless there's some foolproof DDoS mitigations I'm finding it a hard sell, especially on the whole "if you need more control and protection, do X" argument - you rarely need that until somebody straight-up exploits you, and when you do, you're not going to be particularly sympathetic towards Google's marketing speak.

maldeh··on Dive into Deep Learning: Berkeley Course
Your point is understood, and reiterated in other posts here. I just wanted to add context to the course as being presented by an author of MXNet, who is free to present it in his own framework.

Anyway, there's no dearth of courses in the popular frameworks already, it doesn't really help enrich the community if all educational resources only taught the current "market winner", we might otherwise be missing other opportunities or ways of thinking. Imagine if all schools only taught Javascript or Java!

maldeh··on Dive into Deep Learning: Berkeley Course
The instructors include core contributors to MXNet. I don't imagine one would expect, say, François Chollet to do a course on Pytorch :)
maldeh··on An Apology and an Update
All the same, it's good they took responsibility when they did. I don't use Slack at work at present, but being an emigrant worker myself (not iranian, but honestly it shouldn't really matter), the past week's events left enough of an impression on me that I would have likely strongly protested any attempts in my organization at getting a slack channel going. This response is at least well-written enough that it gives me pause, that I should research the company's track record more carefully before jumping to judgment. (they're still on thin ice as far as I'm concerned, but it's a step in the right direction)
maldeh··on Invisible no longer: Google’s shadow workforce speaks up
that, in fact, was dismissive?
maldeh··on Google Tried to Patent My Work After a Job Interview
Considering all the other corporate strategies Google employees have taken a public stand on lately, I wouldn't be surprised if there were serious fatigue on both ends - on corporate, for dealing with the constant rebellions killing major revenue sources (ex:- project maven's death being the end of any more government contracts); and for employees, who are likely sick of protesting an endless stream of obviously bad stuff while risking career alienation (ex:- dragonfly).

What's not immediately publicly evident is the opportunity cost vis a vis attrition and new hires - how many people have left from disillusionment? How many prospective employees are actively shunning Google directly because of these policies? (While a number might provide explicit feedback to recruiters about their reasons to reject, it's probable a silent majority is silently avoiding all contacts from Google recruiting.) At least personally speaking, I had a vastly more positive view of google around 2012 when I started grad school (more or less my dream company to work for at the time) than what I had when graduating (sufficient to decline any recruiting requests).

maldeh··on Horizon: Facebook’s Open Source Applied Reinforcement Learning Platform
Wow, that's... fantastic and generous, I wasn't expecting this response. The change makes this framework so much more inviting to use.

(I'm also surprised Facebook's legal / open source policy would allow such flexibility around licenses.)

maldeh··on Horizon: Facebook’s Open Source Applied Reinforcement Learning Platform
Ah, it has the fabled BSD+ license: https://github.com/facebookresearch/Horizon/blob/master/PATE...

I guess that's not too unusual from their opensource contributions at this point.

maldeh··on Data Warehouse Benchmark: Redshift, Snowflake, Azure, Presto, BigQuery
That actually sounds like a solid improvement. Do the head-to-head comparisons change with sort- and dist-keys? What was the tuning used?
maldeh··on The International 2018: Results
> We don’t believe that the courier change was responsible for the losses. We think we need more training, bugfixes, and to remove the last pieces of scripted logic in our model.

It is interesting to consider that manually scripted behavior may have been responsible for some of the more strange moments in the matchups; possibly they may have had hardcoded place-ward() routines that resulted in all of that clumped-together warding... or maybe even "check-roshan-pit()" without adequate training data to identify Roshan's respawning patterns?

maldeh··on Are Blockchains the Answer for Secure Elections? Probably Not
This feels a bit backwards to me - is the goal to describe and implement a "secure election", or to apply blockchain to X?
maldeh··on Ask HN: Have you shipped anything serious with a “serverless” architecture?
That sounds like a very pertinent thing to disclose when offering a critique of a competitor, particularly when you're not able to substantiate (for legitimate reason or otherwise)
maldeh··on Gnome has moved to GitLab
A "WONTFIX/NOTABUG" closure at least gives confidence that a bug was investigated. Do we know how many bugs were still pending investigation here, prior to being closed due to this migration?
← PreviousPage 2 of 5Next →