dswalter

Honestly, I view Google as a large impediment to python3. If they'd bit the bullet and spent 10MM on transitioning over back in 2012 or so, the perception that 2.7 was still 'the real python' wouldn't have persisted nearly as long.

When tensorflow was released (november 2015), it was python2 only!

> no end in sight for Python 2.7

That's not correct. Numpy, for example, will not be supporting python2.7 starting January 1st, 2020 (https://docs.scipy.org/doc/numpy-1.14.2/neps/dropping-python...), and there are others in the process.

Python itself will stop supporting 2.7 on January 1st, 2020. (see https://devguide.python.org/#status-of-python-branches)

Has it been poorly managed? Yes.

Will a number of companies skip upgrading from python2 to python3 and instead rewrite it in Go or Rust? Absolutely.

But "no end in sight" is not an accurate depiction of the current state of affairs.

> I get the impression from that the author is an ancient backend

Thanks for this clear example of an offhanded dismissal.

It's not entirely clear what you're saying here. Does your business use these solutions as a way to enrich the information needed used by someone assembling your products, or do you use them to ensure everyone's eyes are looking in the correct direction while on the assembly line?

Nobody's perfect, but BurntSushi isn't someone who typically does things by half measures.

The other thing is that the 'gg' in ggplot2 is the Grammar of Graphics, which describes a high-level grammar for composing visualizations. Having an underlying grammar makes it possible for the syntax to have logical consistency, where once you learn how it works, you can incorporate and generalize simply. (aside, I used to use ggplot2 but I switched to working in python, and I miss ggplot2 so, so much).

The syntax being shown here demonstrates an incremental way of layering on the components of the final visualization, which shows the power of the underlying grammar of graphics.

In the wake of the Theranos debacle, I would expect to see the FDA take a stronger stance against medical startups that are taking actions that have real potential impact without evident clinical support.

I couldn't find any license information for BlazingSQL. Where is that available?

Presto's set of functionality ( ANSI-compliant SQL plus so much more) is a pleasure to use as a data scientist who does a bunch of analysis. JSON, arrays, maps, dates, HLL, spatial stuff, rich aggregation functions, etc.

The only real problem with it is that it's easy to get into some thorny-looking transformations because there is so much to work with.

This really undersells what Quasimondo does with his GANs.

What he does is not just firing up a GAN and sampling some pictures.

He's an explorer navigating uncertain realms. https://mobile.twitter.com/quasimondo/status/106311763402625...

I expect this raise to result in a number of enhancements to the user experience which make it a more pleasant and enjoyab[Open this link in the Reddit App?]le place to discourse.

From an ease-of-use perspective, it certainly had the room to improve.

The wisdom that comes from running things in production makes open-ended questions about those gotchas useful interview questions.

'looks like you used kubernetes for a major project recently. What were some of the unexpected warts?'

Someone with experience running a service on k8s will have at least a few stories, and how someone describes their own responses to those hurdles can shed light on their personality.

Since it's part of what looks like an overall project (ixy) to show how it's possible to do this sort of low-level programming in a variety of languages (the roadmap includes js and python), the developer experience is an important component of what they're trying to shed light on.

Per a talk from Joshua Liebow-Feeser at the Rust Belt Rust 2018 conference, Google is heavily using Rust for networking in their new Fuchsia operating system.

This is a rough approximation of a quote he gave during the talk:

"What makes Rust different is not that you can write high-performance, bare-metal code. People use C and C++ to do that all the time. What makes Rust different is that when you write that code, it is safe, clean, and easy to use, and you are confident in its correctness."

If even Google is using Rust for performance-critical development, that seems pretty promising to me.

The underlying reasoning behind PCSA/HLL (that if you squint, you can use the zeros and ones from a hash as analogues to coin-flipping, and can use the probability of the rarest event seen to estimate the total number of events seen) is excellent.

Count-min-sketch is also beautifully simple.

Google is heavily investing in cloud; in 2015 they embarked on a sizable shift, where they decided it was worth using their considerable head starts in databases and machine learning.

They open-sourced Kubernetes and Tensorflow that year, both with a view toward cannibalizing AWS's lead. They released Apache Beam the following year, but it hasn't caught on the same way.

At this point, Kubernetes is winning everywhere, so why not use Google's infrastructure for it? , and tensorflow has enough adoption that you might go to Google Cloud just because they integrate so tightly with it.

More recently, Google purchased Kaggle, and has turned Jupyter notebooks into Colab, as well as hiring Jake VanDerPlas out of UW. They're trying to turn GCP as the easy, first choice solution for running and versioning your darn notebooks. But if you're already there, why not use BigTable to pull the data INTO the notebooks...?

There are definitely tone-deaf steps they're taking, but Google is building serious advantages into their cloud offerings.

I think the parent comment was saying that none of these were primarily made by cloudera.

I think Google has pretty good deep learning-based word prediction for their Swype-style Android keyboard.

There was a period, maybe 1.5 years ago, during which text input prediction got substantially worse, then gradually improved. Along with the change came the ability for text input to change the estimated word after you entered the next word, using the combination of your entries to both words to estimate both simultaneously.

If they have language models which perform that task at a level worth pushing out to consumers they can do some smoothing of entries in a list.

PSA for anyone reading this, the phrase is colloquially "burying the lede."

"The lede" is a journalistic term roughly meaning the gist of the story. http://grammarist.com/usage/lead-lede/

A half-decent software engineer can easily and quickly produce a lot of data products.

If you divide the responsibilities of a data scientist into two broad categories:

* Can I get the answer I need/build the product I need?

* What is the question I should be answering? What do I need to build?

Software engineers tend to do quite well at the former, but are not necessarily guaranteed to be successful at the latter.

If your target is a data scientist role at Google, you're probably going to want more schooling.

But if you've been producing value in a DS/ML role for years, you have experience, which is even more rare than some of the qualifications people are listing here.

If you can say "I created an anomaly detection system using isolation forests that 5,000 clients relied on for detecting market changes", there will always be places that want your skillset: it is kind of a new field, after all.

So the credential bloat at entry-level shouldn't really be an issue for you.

Cars weigh thousands of pounds and routinely drive upwards of 60 miles per hour.

The success of the first iPhone or Facebook app didn't depend on using it to navigate through life-and-death situations not only for the users but also for everyone around them.

There are places for 'move fast and break things'. But cars move fast already, and they can really break things.

You can also view Facebook messages on mobile without Messenger by switching to the "Desktop Site" mode for your mobile browser.

I don't think there is anyone in the python ecosystem who is as good as Hadley Wickham at understanding data analysts and designing API's that a budding analyst can reason through and understand.

ggplot2 is of course the prime example. It simply and elegantly allows characteristics of your visualization to either vary based on data, or remain fixed to your specifications. Since so many data visualizations in science and industry are really variations on that theme, you can do the vast majority of what you want.

That's a good term. I'm sure you've found this collection of links on 'streaming algorithms'. It's a gold mine of resources in this space: https://gist.github.com/debasishg/8172796

It always makes me happy to see approximate algorithms/data structures like hyperloglog being used.

That went gloriously off the rails.

I know net neutrality is no longer law, but I think if you interviewed a thousand people on this site, you might be the only one miffed that Netflix isn't paying ISPs for faster routes.

Put another way, it is my impression that consumers are already paying for bandwith. Getting Netflix to pay for faster routes would mean ISPs are double-dipping.

I agree that the first two three body problem books (I've just finished dark forest) have some weak points on character development and dialogue.

But I've really enjoyed them so far because the books have surprised me at least a dozen times.