4mitkumar

You are not alone.

I'd guess at least over 50% of folks here would have gone through a similar phase at some point in their lives (learning block wise).

Based on my experience, especially over the past couple of years, I've realised that even if I can power through some of the subjects/topics which I find arcane (e.g. linear algebra, tensor calculus), I quickly lose what I've learnt if I don't get to apply the learning on a real project, not a toy one.

In your case, I think once you've familiarised yourself with the absolute basic syntax structure if python, go find couple of real world projects and just try to get them done, without worrying too much about code elegance. That would give you the necessary confidence and understanding for moving up the next step.

You can find small volunteer projects it participate in competitions (on platforms like kaggle) which may require you to use at least a portion of python.

But most importantly, don't lose heart. Learning takes time. It's just as important to take a step back from it for a while as it is to keep at it.

I strongly recommend a series of 4-5 video lectures on Coursera by a MIT prof I think. Course was called "Learning how to learn". You don't have to take everything in it but I still believe the overall approach may help you.

Cheers and all the best!

Nice little tool.

You may want to provide some details/examples on input data formats for those of us not from graphing community

Orgzly is really grast in terms of feature support and minimalistic UI. I only wish it supported time logging as well, which I use fairly regularly.

Along with SyncThing and Emacs in laptop, that's my preferred organising system for the last 6 months or so.

Could also be that there is little to discuss since they are planning to launch the course only in June.

Folks might just have found it interesting enough, registered for the course, upvoted it.

There are links to some interesting (basic) content on ML on graph data though

> this setting causes problems with google captcha - the number of challenges that you will need to solve will drastically increase

No kidding. I'm talking about ~30-40 clicks (1 click per task in the captcha grid)

Just wondering how weird would it be if someone at the UN proposed to break up "monopolies" like China it the USA itself. And yet, I suppose a good parallel case can be built in support of this.

NOTE: I'm not dissing the monopolies act and understand the dangers. This is just a thought pointing out different yardsticks for corporations and countries.

Similarity and suitability are two radically different scores. But for various reasons - including, but not limited to, the ease of calculating a similarity score - we end up conflating the two in a lot of cases.

While calculating similarity scores is getting easier by the day across a lot of content formats (think image classifiers, sentiment scores, etc), the same is not true for suitability scores.

Technically, I'd think calculating suitability would need more than just matching patterns based on some selected criteria, which is how essentially all recommendation engines work today.

I wasted almost an hour trying unsuccessfully to book a cab using Uber yesterday only to realise it was 1.1.1.1 in my Private DNS which was the problem.

I had internet access just fine (browsers) but Signal, Google Maps, Uber and Ola just wouldn't work with 1.1.1.1. Haven't gotten around to think about the reason

This. Is how an informed discussion should be. Based on original references and nuanced positions. For hyperboles, we always have the mainstream media :)

But you can! Mostly in Android though. I can confirm MIUI - the stock rom on Xiaomi phones - have this exact feature built in.

You can block all hidden numbers, all numbers not in your address book, specific numbers, etc.

Or one can use Tasker - an Android automation app to achieve the same functionality and more in any Android phone. I suppose these days, you can use Iphone's scripting functionality as well for the same, can't you?

Privacy concerns regarding MIUI /Android can be handled by installing a firewall like Netguard.

There is some good advice in here, some which I myself stumbled across after trial and error.

But 39 points?!

May be it's just me but whenever I see a list with over 10 items, I just get so tired when thinking about it. Ultimately it just gets filed away, never to be referred again.

I wish the post started out saying "9 things to get your writing style better" instead of "...39 ways to make your blog..."

In fact, even 10 is a long list. These days I try not to make lists more than 5 items long. Even shorter when it comes to list of things to do. Of course, each item can have some sub-items, if needed.

Edit: auto-correct mistakes and rewording

Wondering if there is a corpus of "stop-phrases" like stop-words which will include gems like "passionate about software development", "good communicator", "truth-seeker" and the likes.

It might be interesting to do a basic pre-processing of the JD to remove all such stop-phrases before evaluating the role being offered.

As Prof. Bohnet herself points out, the HR dept. should do more A/B testing of its practices instead of relying on the gut feel. Hence the gradual increase in the prevalence of "people analytics" (apart from allure of sucking in all the data-points you can).

But there are a few problems: 1. A/B testing in talent mgmt practices is costly in terms of time-investment. - Final, long term impact of such experiments are not visible to the org for, well, a long while. Sure, may be some short term impacts like increased diversity - may be visible in the short term. But is that worth all the experimentation for all the firms except the largest ones like FANG?

	   The experiment cost is probably not worth the value for the smaller orgs.
	   

2. Vetting people for skillsets and qualifications only, masking the personal details - like in the OP on Helsinki, is a good start. But it does not address the biases on the historical disadvantages like not coming from a pedigreed organization because of not having been to a pedigreed school. Think of being ranked lower in the applications to Netflix because you are not coming from Google because you didn't go to Stanford. Blind -audition technique would be the best in situations like these as the focus would be on demonstrating the skill needed and not the "qualifications" (read - pedigree).

Such skill tests can probably be done easily for programming and other situations but a lot of roles do not have a strictly defined evaluation criteria as the KPIs are...well... fuzzy. Or at least not something which can be demonstrated in a short "test". Think sales / pro-management. You can not demonstrate in a test if you are a good enough sales person who can meet the revenue projections ("sell this pencil to me" doesn't count).

Issue here is lack of a easily reproducible and reliable signal for your future success for such roles, thereby increasing the dependency on past history - and by association, your pedigree.

Of course there will be other issues like isolating performance factors into "chance" and "skill" in a noisy dataset that is usually available, if at all. Or hiring mindset which focuses on the required skillset (what is needed to get the job done) but the "best person" for the job which muddies the waters by adding evaluation factors which are really not material to performance on the job at hand.

But if the cost of experimentation can be reduced by devising quick tests for generating reliable signals independent from historical background, it would be awesome!

It probably would be more helpful on the pull side rather than the push side of the channel. Useful for consumption of content in certain situations by certain segments of readers.

Say, someone scanning through a list of legal docs to identify the most relevant ones. A short blurb would be pretty helpful.

Recently I've been thinking about inherent biased towards applicants based on their prior work / educational history, in terms of organizations they have been to.

I get that it acts as a moderate signal of how tenacious and inherently able that applicant might have been in the past if she went to XYZ school or ABC company. But there will be so many false negatives!

A lot of folks who probably would be otherwise capable of the job at hand would be dismissed for not having the right pedigree. Also, past performance is not always the best Intuit if future success.

And then of course is the issue of how do you even define success, especially in case if university graduates. How does a Yale or Harvard define our measure success for its students... Probably in terms of potential impact they have on the world or how good their names may look on the rolls 10 years down the line.

A lot of the issues, especially in terms of job applicators but also applicable in admissions process, ultimately comes down to information and numeric asymmetry between the organisation and the applicant.

If there was a way to somehow capture the readings - however brief - for rejection or selection of applicants or a documented scoring methodology, that itself would bring a lot of change.

I would love to know if some resources to further explore the alternative models of applicant assessment which does not place as much importance on the past organisations / schools they have been to.

Any suggestions would be much appreciated

Recently I've been thinking about inherent biased towards applicants based on their prior work / educational history, in terms of organizations they have been to.

I get that it acts as a moderate signal of how tenacious and inherently able that applicant might have been in the past if she went to XYZ school or ABC company. But there will be so many false negatives!

A lot of folks who probably would be otherwise capable of the job at hand would be dismissed for not having the right pedigree. Also, past performance is not always the best Intuit if future success.

And then of course is the issue of how do you even define success, especially in case if university graduates. How does a Yale or Harvard define our measure success for its students... Probably in terms of potential impact they have on the world or how good their names may look on the rolls 10 years down the line.

A lot of the issues, especially in terms of job applicators but also applicable in admissions process, ultimately comes down to information and numeric asymmetry between the organisation and the applicant.

If there was a way to somehow capture the readings - however brief - for rejection or selection of applicants or a documented scoring methodology, that itself would bring a lot of change.

I would love to know if some resources to further explore the alternative models of applicant assessment which does not place as much importance on the past organisations / schools they have been to.

Any suggestions would be much appreciated

In my experience on a similar configuration, most of the times such aliens is because of i/o bottleneck.

Antivirus, Windows app optimisations, and device manufacturers' softwares running at predefined schedules, choking the disk might be a problem.

You frontotemporal should check the process explorer next time you gave the issue. In the meantime, may be check the task scheduler to see if some heavy read write tasks are scheduled

Vector magic runs perfectly fine and does a great job even in high Res photos on Win 10. About 90 seconds for 2048x1433 image with some intricate details at high quality setting (as compared to offered medium and low quality settings) with unlimited colours.

But results are still pretty nice even at low quality setting with limited colour palette. Hardly takes 15 secs.

This is really interesting! Have been doing something similar myself on my own phone (Android) for about a month using the wonderful NetGuard app ( https://f-droid.org/en/packages/eu.faircode.netguard/) which allows one to see the source and destination for each request from all apps.

It doesn't display the content type enough, afaik

Working fine for me as well on Android (Oreo) with Firefox Focus with all tracking disabled. No pop-ups on couple of articles I tried

It would be really amazing to have an option to engage with suspected telemarketing callers using Google Duplex.

User experience: * Get a suspected telemarketing call * Ringer silent while the in call UI waits for user to pick up * If ignored, call us picked up and Duplex engages the caller to know about the offer and get other details * Call transcribed; Optionally summarised and saved in cloud for later access by user

What Google gets: * More widespread use of Duplex with associated improvements * Data from telemarketer, voice as well as offers information which can be further analyzed for any beneficial purpose

And maintaing the balance at those speeds. I can hardly control mine at 50 kmph downhill

I am about 4 hours in and still into it! Thank you for sharing the lectures.

Just finished the series on Face Recognition. I was wondering if the fusiform face area (FFA) get activated while thinking of someone's face instead of actually looking at one? Might also be interesting to explore the memory of faces in persons with prosopagnosia?

Haven't actuallly googled for these yet. Will do so in a bit.. after the next series of lectures may be.

Would love to discuss. You can also reach me on the email - considering the timezone difference (I couldn't find yours on the profile page).

The news media has a way of using the "wave". The barrage of articles with FB caught with it's hands in the "cookie jar" will contiue piling up. And a lot of them are just trying to get on in the action but some - like this one here - are genuinely upsetting and need to be reported.

The issue here is not the business model of FB but the sneaky ways it tries to make it seem something completely else. While the audience here at HN and similar forums are more than aware of the business model, I'm pretty sure if I asked my mum about it, she'd draw a blank or would not be completely informed. It's to a population like this who are most likely to fall for dark UX patterns and other shady ways FB uses for getting an uninformed consent.

And that's not right. I think, the outrage is not about the business model as much as it is about (un)informed consent. Sometimes - like in this story - almost while cocking a snook in the face of regulatory authorities.

This is actually a very good point. I was thinking myself something along similar lines that intuitively speaking, death by unnatural causes like terrorism/homicide/suicide are of course more "newsworthy" than the ones due to "heart-disease".

Although the study is trying to do a larger scale comparison between the perceived and actul leading causes of deaths, it would serve it much better if it were to segment the results by age group.

I wonder if this also has something to do with their planned support for AMP in Gmail. If yes, then this might get tricky ...

Why does the atom look so big in the bicture? Atomic radius of Strontium is 219 pm, so that small spec there in the picture should be about 438 pm across.

I'm assuming the two ball-pen nib shaped structures on both sides of the spec in the picture are "two metal electrodes placed about 2mm (0.078in) apart". So, the space between the left tip of the electrode and left edge of the spec is about 1 mm. Based on some "visual calculation" (zooming in the picture and doing some approximation), the spec seems to be closer to about 0.03 mm across, which is orders of magnitude larger than 438 pm that it should be. What gives?