Mikio L. Braun

ML and AI expert

@mikiobraun

@mikiobraun Twitter Memorial

19,827 tweets · 2008–2024 · 1046 threads

2019

RT @bigdata: Thread: In this episode of #TheDataExchangePod my collaborator @mikiobraun and I speculate about AI and Data Trends in 2020. F…
Replying to @thvasilo
@dhh I‘d say start with DHH‘s book. Werner‘s book is more about learning a craft and stepping aside to let it happen.
1
Replying to @mleznik
@dhh I read „It doesn‘t have to be crazy at work“ and think that will always be relevant ;)
1
I find it amazing that this gets so many irritated reactions. Also, shouldn‘t you people be WORKING? Joking aside, the book Effortless Mastery by Kenny Werner (on making music) opened my eyes a long time ago on that topic. twitter.com/dhh/status/120…
7
Replying to @tkasasagi
I wouldn’t even blame it all on politics and people, the staggering complexity of what the company has built is another factor that makes anything new complicated.
2
Thread (3/3)
DHL customer support was also extremely helpful, I mean totally clueless. It seems that fact that the parcel was waiting at a distribution center wasn‘t even in the system. Every time I called I got different information. 🤷🏻‍♂️
Replying to @baggerspion
@therealpadams Murder mystery, pretty sure I heard that one before /cc @truemped
1
Replying to @ChiefScientist
Join the Siri team at Apple and make Siri smart enough to teach children how to math!
Replying to @VaroonMathur
Interesting how smugness/angriness is reversed between Hillary and Trump! 🙄
1
Replying to @baggerspion
@therealpadams Unfortunately I clicked and saw the video of the beach covered in these 😳🤯😫
Just found out the final season of Silicon Valley is starting to be available on Sky in Germany. 🙌
So after 2 years my son is finally able to run that diesel train in minecraft after I found how to add lava as possible fuel to the config file for the mod. This feels painfully close to work! But his joy was worth it ;)
11
Replying to @kimknilsson
Actually it is mostly a concert hall, and then there's a spa on the side... . I know, what an odd combination!
Replying to @lzamparo
Cannot imagine how it is nowadays, last time I was there was 2012 or so, just before Zuckerberg showed up and everyone was like whatishappening!?! I just remember endless oxygen free poster sessions.
Replying to @tdhopper
Who is this thought leadership that is also a provider of guidance it seems?
1
Replying to @voidmaze
@joaomcsantos @therealpadams @ZalandoTech @oldJavaGuy That's a great talk. Funny I've never seen it before!
2
Replying to @FcoJerSanchez
Turns out, the stickers are still good and are now on my own laptop. #whew
1
Nothing like diving deep into midlife crisis level critique of modern life than a couple of friends you‘ve known for 20+ years.
3
I don‘t think it is too early for Till Brönner‘s Christmas Album yet... 🎺 🎄 💕
Replying to @RachelStewart04
Don't get me started on Germany's approach to cashless commerce. There's also a range of money up to 10€ which can you cannot pay electronically or with large notes.
1
@totopampin Once we have created the new math it will also be packaged for brew, sure thing!
Replying to @NanaYamazaki
Well if they ain‘t getting ready for the robot uprising I don‘t know who is.
1
Idea: service that tracks a number of colleagues to figure out who is free to join for lunch. I call it LunchMaster. Enterprise version can be configured for non-declinable random lunches across the company. For "networking."
8
Replying to @noelwelsh
Well that is a perfectly good question. I guess I would need to get up for that, though.
Thread (2/2)
Well, the prospect of building AI was what first got me into studying computer science, but the way we attribute things to AI is a bit on the too optimistic side if you ask me.
4
Replying to @mucio
@peteskomoroch @earino @chrisalbon Well my hunch is that the credit limit sits in some trusty old database and is populated by the estimate when accounts are opened? And can be overwritten when needed.
Replying to @peteskomoroch
@earino @chrisalbon And yeah, they probably cannot talk about other people‘s credit scores as well.
Replying to @peteskomoroch
@earino @chrisalbon I agree that this is how it happens (and scaling up CS is hard). In the end, for the customer, the effect is the same.
Replying to @peteskomoroch
@earino @chrisalbon I‘d say the fact that it is treated as if it is black box to people affected is an issue related to giving AI too much power over our lives, and that is concerning.
Replying to @francoisfleuret
I always thought the biggest danger of AI overtaking our lives is if we let them.
RT @francoisfleuret: If you have theoretical concerns about corporate decisions outsourced to "artificial intelligence", read this. https:/…
@sebastianarnold Brackets on PC/Linux was sort of okay with German layouts, but on the Mac it does not work. And don‘t get me started about \ and |!!
Thread (2/2)
ah gosh there was a collection if meeting stock photos with the caption „am I in this meeting“ on each of it, but of course I cannot find it now in this stream of consciousness that is Twitter...
Replying to @mikeproppe
@zebulgar I guess this happens if you autoscale each panel separately.
Replying to @rasbt
@c_a_schiller @theShiftPR0JECT @lexfridman Algorithms, yes, maybe also much more energy efficient deep learning computation hardware.
Rewiring my muscle memory from German to English keyboard layout... takes time! I celebrate each ~,+,^ and () that I get right the first time!
7
@c_a_schiller @rasbt @theShiftPR0JECT I wasn't yet aware of the digital sobriety concept, but I also fully agree that this isn't sustainable. Finding more efficient algorithms has always been one of a goals of computer science, but currently we're going through a phase of adding more power and machines.
3
Replying to @earino
But only because I studied in Germany where there are no student fees (or they were minor like less than 1000€ per year).
1
Replying to @bobbruno70
I fear I would need to be the one who creates that... but that would be a good abstraction!
So tried looking into getting the lapack libraries for jblas to build, but I realized I am past sinking hours into fixing versioning issues across languages and operating systems... and probably time.
So personally, I‘d rather do something because I know why than because I don‘t know why not but maybe that‘s just me.
3
@Lauri_Apple Timing is everything... you wouldn‘t want that that dopamine hit from a code commit messes with that.
Replying to @madsh
Now joking aside, didn‘t Microsoft have some cloud version of Word where you can collaborate?
I have to say writing software with many people (I try not to say „at scale“) is a hugely fascinating topic.
3
Replying to @nikovarga_hr
To me that combination of visuals, goofyness, and existential doubt is also deeply Japanese.
Replying to @rorcde
@RodrigoRivr I was, but for some reason I never watched it before. Being half Japanese I also spent many summers in Japan watching anime there. Still... ;)
1
I saw Neon Genesis Evangelion for the first time this summer when it was on Netflix and it is amazing, and deeply disturbing. Reading this brought it all back. Font as visualized language does have more impact than is usually obvious. twitter.com/iA/status/1186…
5
Full house with @matei_zaharia at @Zalando‘s Data Engineering meetup with some insights on differences between classical coding and ML projects.
Media
14
Replying to @DmitryKan
And some professors just had insane writing speed even on a chall board. Faster than me with a pen!
1
Thread (4/5)
The interesting about mathematics is that is so far on the edge of human capability that there exist a handful of people who are still able to parse this and won‘t get why others don‘t.
Replying to @semiDL
@snikolov @fhuszar Well, you could argue that calculus was indeed inspired by the needs of Newtonian mechanics, IIRC.
2
RT @jessetanderson: If you are in London and going to Flink Forward next week. I have a huge favor to ask to bring some items I left in Lon…
So most cash machines in Germany will give you only 50€ bills (if the amount is large enough), and most shops refuse to take cards for amounts under 10€, so in effect it sometimes just difficult to do commerce in DE.
1 4
Replying to @johncutlefish
Are the fields on the left tabs that show more panels when clicked or is this it? It is all so confusing...
@vetal_don Totally agree. And yet some companies are pushing bringing automation etc. to notebooks.
Replying to @DmitryKan
Totally agree. I think notebook code should mature and move into libraries and then future work is built on that. MATLAB was great at auto-reloading code, it is possible in Python, but most people I talk to say they rerun the notebook which can be impractical.
1
Thread (5/5)
Oh, and the whole topic of how to effectively collaborate on notebooks, or whether they are good enough for production code is also still open IMHO.
2 3
Thread (4/5)
I remember with big data there were such possibilities as well, in the end we settled on Spark and friends, which is again close to SQL in what it can do. Maybe the same will happen for end2end ML, too?
3
Thread (3/5)
Likewise, training might involve not just the model but other preprocessing/feature generation steps and pipelines, backtesting, etc. And there is the whole monitoring topic.
3
Thread (2/5)
For example, one version is to deploy "pure ML" models from notebooks, but there are use cases where the model is part of a bigger system, including feature computation, post-processing, and so on.
5
My impression from last week's #stratadata was that there are many startups on either data cataloguing or end2end ML. Reminds me of a few years back when there were many big data startups. I think what is interesting is that end2end ML is not fully understood yet.
8 25 View thread →
Thread (2/2)
What is the tone? How do you react to difficult situations? How to be inclusive? I believe that the more „personal“ AI becomes we can‘t help but to relate to the technology we create, so these are very important questions.
So #StrataData is all about data (sometimes even big), but only SAP had the mental presence to bring in the only true Data. 🤩
Media
4 37
RT @NanaYamazaki: I am so so proud that the DATA SCIENCE x ZALANDO video series that Yvonne and myself worked on just got shortlisted for t…
Replying to @MartinSonneborn
@FacebookDE „The system considers unusual even if you didn‘t mean it“?!?
1
Replying to @filip_woj
@Aaronontheweb I also tend to not take notes, and it also helps to listen bettee IMHO.
Replying to @AmazonHelp
Sooo, I followed the link but I get forwarded to general FAQ pages, or could contact a support person via chat. Not sure that his will be more helpful... So I'll just hope this'll eventually cool down. Thanks anyway!
Replying to @AmazonHelp
Yes, tried the reinstall already. It was a higher version than what I had, but then the same problem... do you know what the latest version is?
Replying to @AmazonHelp
Sorry for the confusion, this is not the web player, but the app that you can download. Where would I check for pop-ups? Do you mean notifications?
Replying to @AmazonHelp
Hey there, just a follow up. I re-installed the app, I'm now at 7.7.0.1903 and it keeps telling me there's an update and after updating, I'm still at that version. I'll try and ignore the notification for now... .
Replying to @DRMacIver
True. Nobody ever means the question literally. There's always a "and it is WRONG that nobody thinks about it!"
Replying to @AmazonHelp
No, not yet. It is not super distracting but I found it a bit odd. Usually I get updates only every 1-2 weeks. I will if it keeps up, or try reinstalling.
Replying to @AmazonHelp
Sorry, I didn‘t mean the web plaayer but the desktop player that you can install on your machine.
Hey @amazonmusic, your web player wants to update every 2-3 hours since Friday. Lots of bug fixes or am I stuck in a loop?
1
Replying to @x0rg
I only briefly did some Ruby and I still have the sticker and still go to irb for quick calculations. #theanswerisyes
1
So I established with a colleague that BS means bold solutionizing and now I cannot go back.
4
Replying to @adichad
I still posit that #doglevelAI is a good proxy to aim AI research at. 🐶🤖
Replying to @adichad
Yeah, it definitely is progress. I'm just saying we shouldn't equate test performance with actual intelligence.
1
RT @monsoon0: Q: What's yellow, linear, normed and complete? A: A Bananach space. #mathjoke
Replying to @tenderlove
The one franchise that does not understand the meaning of the word „final.“
Replying to @cagan
@truemped Thanks for sharing this, this gave valuable context to some experiences I've had with the product role in big projects.
Replying to @baggerspion
@therealpadams I tell you if you tell me... *looks at sketchpad* what you wanted to become as a child and who your favorite teacher was.
Replying to @baggerspion
@therealpadams You want to reverse engineer my security questions or what?
1
The great thing about friends is that they know stuff about you you are not aware of and care enough to tell you.
1 6
When I said to my daughter „have a great start into your weekend“ she replied that the getting started part is not the problem with weekends, it is the getting out of them again part that is. ;)
11
Thread (2/2)
@RodrigoRivr Well, okay, unless the skill they want is to focus on the (end-)customer, then your customer is the company that employs you and the customer of the company ;)
Replying to @rorcde
@RodrigoRivr Yes. And what you should do as employee is to optimize your skillset to uphold employability. There's a lot of talk about everyone should be customer centric, but at the end of the day, the company is the customer of you.
View thread →
Thread (3/3)
Okay, like how divorced the creation of value is between the company and the employee. The company creates something that people are willing to pay for, but for the employee it is all about capabilities and skills. (at least when you take an extreme view)
Replying to @IgorCarron
@joinstationf @MrsCaroline_C @Microsoft Just gonna burn a few PetaFLOPS to train those models!
1
RT @johncutlefish: (1/11) So you want to fix something in your company? You're not happy, and want something to change? Here is a quick t…
Replying to @aschrock
@ElizabethFlux Version 0.99 patchlevel 13 on a machine with 8MB (!) of RAM. Starting X11 so I could watch ghostscript render the tiger head demo. Working on the console because X was too demanding.
Replying to @rorcde
@RodrigoRivr Functional! ;) No idea to be honest, but I think we got better at knowing what good coding concepts and approaches are, so maybe there is a good language within PHP, but my experience with it are limited.
1 View thread →
Replying to @InkmiHq
@codemonkeyism @dehora Well I guess there is lesson here about listening to people ;)
1
Replying to @srchvrs
@FedorShabashev @lemire @richardstartin Did I only dream this or was there a recent paper around language models that bragged about number of petaflops sustained during training?
1
Replying to @dehora
One of the argument back then was that programming would be visual and super easy. I have yet to see one visual tool (eg interface builders) that does not require deep understanding what is happening.
Replying to @dehora
I still remember being asked why the heck I was studying computer science back in the 90s. People told me coding would get so easy, everybody could do it.
3
Replying to @lemire
@srchvrs @richardstartin One could argue that that‘s also okay. You have to keep exploring to make new „useful“ discoveries.
1
Replying to @lemire
@srchvrs @richardstartin I probably made a bold claim there. But there is a specific kind of research (eg NLP models) that involves lots of data and insane computing resources that seems like it can currently only be done by companies like Google, FB, or Baidu.
1
Replying to @lemire
@laurencetratt @johnregehr Yes, the state as the primary supplier for research money is fine as long as it seems to work. There might be better systems, but how would you test that? In AI some big companies are outperforming publicly funded research. Maybe there is a lesson there.
Replying to @tonyveo
@johnregehr Sometimes authors also need to pay a fee after their paper is accepted to be published. Including an open access fee...
Replying to @rorcde
@RodrigoRivr @johnregehr And travel restrictions are getting worse...
1
Replying to @tonyveo
@johnregehr That, and a substantial part of the work involved (the reviewing) is done for free.
@johnregehr Also limits on papers per issue/conference are so arbitrary. Even meeting in the same space seems arbitrary. Twitch but for scientific conferences.
2
RT @CommodoreBlog: Let's see how many Amiga owners are really on Twitter? Comment and retweet if you are or was one back in the day👍🤗 #Onl…
Replying to @agibsonccc
@Kingwulf @kubeflow And then people make decisions based on those blog posts and you end up with a set of tools that don‘t really make it easier.
2
Replying to @agibsonccc
@Kingwulf @kubeflow Oh I didn‘t mean to dismiss your concern. I think it is actually worse. I often get the feeling these tools aren’t really fit to how a data scientist would actually work but just tick some boxes.
2
That point in summer where you wear shorts just to survive the train microclimate. 🚃😓
Replying to @agibsonccc
@Kingwulf @kubeflow I think that could be said about quite a few things.
1
Replying to @semithinking
@SusanPotter @Twitter „Categories and functors, without admitting it?“ 😂
Replying to @semithinking
@SusanPotter @Twitter Yep. I tried reading a book on it, my professor saw it and said „isn‘t that just abstract non-sense?“
Replying to @semithinking
@SusanPotter @Twitter Yeah some of the other stuff also looks a lot like probability theory. I thought category theory was all commutative diagrams.
1
Replying to @shuheikagawa
Won't you need your BVG ticket so you don't have to ride your bike?
1
RT @bamafangrl: Do y’all remember, before the internet, that people thought the cause of stupidity was the lack of access to information? Y…
Replying to @xamat
I heard Miles Davis would sometimes show up and mention that thta C# in bar 37 was a bit flat...
1
Replying to @ssdpd
First a button. If the AI's expectation is met consistently, automagically (but notify me). If that works, automagically!
1
What I need is a GMail feature that notifies the sender that I have read the email and will respond... just not now...
5
I keep hearing "they shipped their org," but I wonder whether a better interpretation of Conway's law is that teams will inevitably ship their org, so you have to get the org set up right.
2
Replying to @octonion
Yes, definitely. I really don’t have a lot of technical understanding but it seems TPUs are substantially different from CPU tech.
Replying to @octonion
Intel‘s whole PR approach to AI amazes me these past years. They also come up with all kinds of creative metrics like Watts per epoch to try to make a point.
1
Replying to @klauso3
You know how hard it is to find these people... that‘s why I wanted to start with slightly superlinear... Still thinking whether x log x would already be worth it...
10x and 1x are O(x) anyway. You should really look for x^(1 + ε) for ε > 0 people.
1 22
RT @QuinnyPig: "Famously, you're not allowed to work at @awscloud unless you can eat two entire pizzas yourself. We're horizontally scallin…
@Lauri_Apple @KhaledH_ @therealpadams Actually, bm1 would be later in the afternoon... I have to check but I have meetings running till 2pm.
Replying to @dehora
For what it is worth, I think there is a natural tendency in companies to have very clear roles, and also to split responsibilities up to make them small enough they become manageable. But each such decision puts in structures which may or may not be fitting later on.
1
RT @jannis: 👎 Yesterday afternoon somebody broke into our apartment and stole all electronics. ❓Anybody in my Berlin network selling a Mac…
Replying to @chrisalbon
Still considering whether the „unmade bed look“ is my conscious choice or not.
Replying to @HABrazeau
If you need to get something done, you can be quick if you know beforehand exactly what to do, why, and how.
Replying to @acedrew
@aphyr Ah, the young ones aren‘t any good. There was a time to define processes and we‘re done with that.
1
@aphyr That‘s because all our processes are golden, perfected more than 100 years ago and handed down from generation to generation. Then again, we Germans also consider everything inherently broken and fix-worthy. So, yeah...
1
Replying to @chrisalbon
I once interviewed a guy who implemented decision trees in SQL (with JavaScript UDFs) because all the data was in BigQuery and he wasn‘t allowed to export it.
7
Replying to @baggerspion
@therealpadams @truemped @Swarm64 Actually I was quoting Back To The Future. (Didn‘t find the right gif, but this seemed appropriate)
Replying to @chrisalbon
That‘s a bit like everyone on Tinder seems to travel to exotic places.
3
Replying to @neuroserve
@bigdata @ZalandoTech @Zalando Sorry if I misunderstand your point. Are you saying ML does not work for logistics, because RL didn't really work in that one use case?
Replying to @neuroserve
@bigdata @ZalandoTech @Zalando Slides are posted out of order and context. The slide with the „none“ shows how you bootstrap a project on the product, ML, and data side iteratively. Logistics is optimiztion based on prediction models. Hope this helps.
Replying to @_fogfish_
@fogfish3 Traditional time series has pretty strict assumptions on the generative process. With DL you can be more flexible and adapt to structure in the data. The DeepAR paper from Amazon talks about challenges for sales prediction.
RT @ginablaber: AI and retail: advanced forecasting models help optimize retail operations, Improving efficiency and timely package delive…
Replying to @alisonjudy
@josh_wills Thanks for the clarification. I was also mainly impressed because I for one do not have a personal backup device... .
We didn't know "put a dent in the universe" was a Steve Jobs quote. We felt it was maybe not the right scope. With the help of thesaurus we found a few alternatives: scratch - a modest start scar - if you have unfinished business crater - if you are really angry
2
Replying to @clairikine
Someone once told me the best way to sleep in the office (assuming there is a door), is to spill paperclips and then lie down such that you block the door.
1
@totopampin @j_c_cabrejas You had wifi at home as a kid? ;) *old man voice* all I had was a 56k dial-up, that was shaky if there were sun spots.
3
I have to sit in my brothers room because there isn‘t enough wifi in my room to video chat with my friends is my daughter taking it to the next level in parenting challenges.
4
Replying to @baggerspion
@therealpadams @rustlang What are your other projects called? „1on1“? „Jourfixe“? „Purposeworkshop“?
11 years! Holy moly! And people still ask me why I am on Twitter. At least food updates are now covered by Instagram.
Media
4
RT @QuinnyPig: So @awscloud announces a new service about which you’re probably the world’s leading expert. What service are they launchi…
Replying to @CMastication
Trying to remember what the most amount of V related to Big Data I‘ve ever seen was. Pretty sure someone made it to 10...
1
Replying to @IgorBrigadir
@animesh_garg @IgorCarron Well, the DDoS is already happening with all the humans involved, it seems.
2
Thread (2/2)
@asknbid @IgorCarron @animesh_garg Joking aside, I find it hilarious that with all that cloud stuff that we have cmt still breaks down on NeurIPS submission night... .
3
Replying to @asknbid
@IgorCarron @animesh_garg Yeah yeah, request limit, open connections, concurrent transactions, point is *old man voice* we were lucky to have 100 back when I was submitting my papers.
2 View thread →
Replying to @animesh_garg
@IgorCarron *old man voice* 6400? In my days we were lucky if we had 100 concurrent database connections!
2 4
Me in my childhood travelling in Japan: „That‘s some intense public transport going on here.“ As of lately, I am increasingly getting that feeling in Berlin.
2
Thread (2/2)
@octonion Now I am thinking what a private data science license would look like... no production stuff, data is always clean, focus on beautiful charts instead of impact... actually, this sounds pretty good! ;)
Replying to @QuinnyPig
@stu @paulrteich Just day 2? My Twitter feels like it has been going on for at least half a week already? ;)
2
Replying to @QuinnyPig
When I hear "cloud native" I'd like to think of flying whales drifting through the clouds with their offspring.
2
Replying to @baggerspion
@therealpadams Consent on the blockchain! Also, I wonder what consentcoin will be useful for...!
Replying to @tonyveo
@mleznik And „in production“ is just that fine line that marks the transition, right?
1
Whoops, accidentally e-bought the tickets to #AvengersEndgame for tonight instead of tomorrow. #IAMNOTREADY
1
What is it about programmers (me included) that they consistently put "let's write a tool for that" into any form of problem solving?
4
Replying to @stefan_will
@johncutlefish Maybe I am! ;) I think what I am looking for is a visual way to track progress across multiple, mostly unrelated projects. Each swimlane being a project you see where you are and what is upcoming. The columns are good for wip limits. It works okayish.
Replying to @stefan_will
Depends on your setup, could be both. I personally use something like it on a physical board to manage all my different pieces of work. I'm adding a time dimension by having a "this week", "next week", "future" column (and a "done" column of course).
View thread →
Replying to @baggerspion
@therealpadams @ZalandoTech Don‘t listen to what anyone says about your bum!
3
Replying to @mariaInTech
@marsty5 Oh my. Good luck with that. Maybe one of those biting things will help...
Dual classing seems to be a theme for me: - computer science / mathematics - engineering / music - academia / industry You gain perspective, but I personally struggle feeling at home. Anyone else feeling the same?
1 15 View thread →
Replying to @mariaInTech
@marsty5 Better go see a doctor. I had inflammations in the ear, but they didn‘t radiate to the neck...
The increase of morning temperature by 10 C, the lack of S Bahn‘s heating adjustment and my poor choice of coat. 😓🔥
In Lüneburg to give a talk at the university. Reminds me that slides printed on sheets of plastic by a color laser printer was once the state of the art of presentation technology.
5 View thread →
RT @lorisdanto: As per tradition, the last question of my compilers final was: Write a haiku about compilers. 0.5 points if funny, 0.5 poin…
RT @stochastician: I’m home in Idaho visiting my mother, and she gets ~10 spam calls a day. The last one was supposedly about breast cancer…
Replying to @mucio
Yeah, I read the article. Somewhat mind-boggling how much perceptions of freshness and durability diverge.
Replying to @superglaze
Battery life is okay (something like 8h), the only issue is once they start running low, they last for only another 10 Minutes. Mine don‘t power off after a while, so that has been an issue ;)
Replying to @superglaze
Got it, too, and getting rid of the cord makes such a difference!
1
Replying to @turtletrafo
@cloudHPC Not sure whether that was my intention, but thanks! 😅
2
Replying to @CMastication
Once we have basic derivatives, let‘s repackage them in layered commodities so everyone can get a piece of the cake 🎂
2
TFW you check into the hotel wifi so you can check for routes to run in the morning. 🏃🏻
Most people don‘t know the joy of crafting some elegant piece of code that does nothing useful... but boy do I miss those days sometimes...
1 19
Well the weather matches my mood ☔️ but I have also forgotten to bring an umbrella 😞
RT @MichelleSPhelps: Professors: I spent the last 10 years of my life studying this topic. Also Professors: I wrote this talk on the plane…
Replying to @arpad_ryszka
@shuheikagawa Some combination of software deployment and actual shipping containers would‘ve also been very cool!
1
Replying to @shuheikagawa
I have a friend who for a while thought docker is really about shipping containers.
2
Replying to @baggerspion
@therealpadams This is what I get from tweeting video links in meetings...
1
Replying to @fx86
*cough* not yet. The way we've been building systems in the past was very team centric, we're now building up more central structure like a notebook server.
Replying to @fx86
Currently... different things in different teams, from Cassandra to Solr 😂
Replying to @CMastication
„I‘m just short five exposures for my rent, so thanks for considering me?“
4
Replying to @baggerspion
@therealpadams @LisaNoeth It ain’t day drinking if it‘s brunch!!
Thread (2/2)
Also, asking people for their email password is just wrong as hell. Reminds me of some payment mechanism where you‘re supposed to give your banking id and password... just... no.
5
Okay maybe there is a version of reality where fb was unintentionally downloading your contacts... but isn’t that equally bad in terms of how they check what goes into production.
4 View thread →
Replying to @Henrikop
Ah, glad to hear that ;) I didn’t really mean it in terms of being frustrated with people, but I think the expectations towards people dealing with people should reflect that you cannot make people do whatever in general.
1
RT @seanjtaylor: I just realized one of the most useful life experiences I've had was my first programming job. I was working on a large le…
Replying to @sszuecs
@ebowman Yeah, I can get how it works, but this for-loop-that-only-executes-once-to-yields-something feels counterintuitive.
1
Replying to @CMastication
@bertil_hatt @DataCamp You could implement the „pull“ such that people who alread paid keep the material. Or otherwise signed up for it if you go with a subscription model.
1
RT @tef_ebooks: kills me that i've seen hundreds of tinder photos outside that glass pyramid and not one says "looking for louvre in all th…
TFW you heard a word so often its syllables start to melt into pure sound and you have to think „now that is peculiar“ every time you hear the word. For me, that word is „question“ as of lately. #vacationtime
Replying to @CMastication
@earino Same here. Job, kids, parents, rage on Twitter. You are not alone!
1
Replying to @geomblog
@IgorCarron First I was like „evangelical does not mean what it used to with all those tech evangelists.“ Then I was like „Oh.“
RT @truemped: Yes, THAT @conways_law like THIS. Seriously, if you don't know what wardleymaps are, go forth and read... Life changing. http…
RT @QuinnyPig: "We're focusing on spending more time talking to customers." I think the stronger phrasing there would have been "listening…
Replying to @Henrikop
Yeah... I know about privacy etc, but right from a consumer point of view, expecting people to have money in the right denomination... if we go all in on cash maybe don‘t start your day with 25€ of change?!
Replying to @vivekjuneja
I usually don't mind using cash, but today it was particularly infuriating... .
Thread (2/4)
On top of it all judging from the tone of the conversation it seems to clearly be my fault that I am not better at managing how my money breaks down into cash.
Germany and its obsession with cash: you can only pay electronically if it is more then 10€. And oh, by the way you cannot pay that with a 50€ bill. 😤
15 View thread →
Incidentally, solution to technical problems are sometimes not technical, and solutions to non-technical problems are seldom some new tool.
1 1
Replying to @baggerspion
@therealpadams @github Oh yeah, do you remember, @oldJavaGuy? ;)
1
Replying to @DJCordhose
And then, the last step, some NLP GAN that autogenerates the talk from some seed topics. And scans slideshare and blogs to stay on top of things.
2
Replying to @DJCordhose
Would probably work well with that AI powered email client I was thinking of.
Replying to @DJCordhose
I had this idea of using Deep Learning trained on slides and videos to automatically show the matching slide to what I am talking.
Replying to @noootsab
@MKarnitschnig @RachelStewart04 That picture has people professionally wearing suits... and Mark... And I am saying that with a lot of sympathy for him cause I‘d probably look the same in a suit.
RT @jasonkarns: Bought a mechanical keyboard with more resistance so my code will be strongly typed.
Replying to @agibsonccc
@suneelmarthi @deeplearning4j @ApacheFlink @ApacheBeam Let's not make this political ;) I have no interest in competing. But people still seem to use jblas and it could use a brush up. Sure, dl4j is the much much more complete library, but telling people to migrate is also harsh ;)
Replying to @agibsonccc
@suneelmarthi @deeplearning4j @ApacheFlink @ApacheBeam @jruby I think the ship has sailed for Ruby and ML... Unfortunately ;)
Replying to @agibsonccc
@suneelmarthi @deeplearning4j @ApacheFlink @ApacheBeam I meant like eigendecomposition, SVD, etc. Althought IMHO people are not using them these days anyway because they are too costly no matter the hardware in terms of big-O.
Replying to @suneelmarthi
@agibsonccc @deeplearning4j @ApacheFlink @ApacheBeam Which brings me back to the original plan to have jblas and then use @jruby to build a JVM based ML platform.
2
Replying to @suneelmarthi
@agibsonccc @deeplearning4j @ApacheFlink @ApacheBeam Well one thing that comes to mind are more advanced linear algebra operations... .
Replying to @suneelmarthi
@agibsonccc @deeplearning4j @ApacheFlink @ApacheBeam @agibsonccc once told me the DL4J API was inspired by jblas. I wonder whether jblas 2.0 is still required given more advanced libs like DL4J...
Replying to @suneelmarthi
Ha! Thanks, man! Usually people's reaction is "YOU wrote that! Man, I spent all of Oct 2014 getting it to compile!" ;)
My jblas needs some love... . Last release was from 2015. The build system is a nightmare (I know I wrote it), Win64 should exist now, it does not know about GPUs... . But it also has 450+ stars, so people seem to be using it.
6 View thread →
Realized you could do dips while waiting for the meeting on those uncomfortable bar stools.
1
Replying to @baggerspion
@therealpadams @DominikDiamond I would run with it, too, if I still had the stamina 🏃🏻
1
Replying to @MonarchFisher
@tef_ebooks My theory is also that we think stars look like this ⭐️ because of this. There really just dot shaped lights in the sky.
RT @MelissaBowers_: My husband told me I write in “long gazes and furious bursts.” When I asked what that meant, he did...this. Also, ple…
Replying to @DJCordhose
Well I used to work in an environment where that wasn‘t the case but you actually tried to understand what it is...
IMHO in most cases where people talk about „<adjective> AI“ we should probably first talk about what AI is.
1
RT @StallmanFacts: Vendor lock-in is when vendors lock themselves inside of a building out of fear of Richard Stallman's wrath.
Replying to @bobbruno70
Thanks Roberto. My current thinking is along the lines that as a principal you need support for those kinds of challenges. You often hear about „lateral leadership“ but that is more about helping people solve technical challenges.
Replying to @bobbruno70
Well as I said asking for a friend, but let us hypothetically say, it has already been tried and the main blocker is organizational, what should a principal‘s role be?
So obviously asking for a friend how do you principal engineers out there see your role if there could be a solution but the issue is mostly team setup/dynamics?
4
Did I write that blog post how industrial software development is a bad idea? Or was that just a dream?
2
Replying to @suneelmarthi
Well as long as it is good for interviews ;) No, but seriously, I hope you will be alright. Take care, man!
1
Replying to @suneelmarthi
What a horrible experience that must have been. I am glad you managed to go on and got some help from strangers!
1
RT @ZalandoTech: The Zalando Data Science Days is a two-day internal conference organized by our Data Science Guild to foster knowledge exc…
RT @squarecog: Mistyped "Data Scientist" as "Data Scientish" and now convinced this should be a thing. "As far as product people go, she is…
Thread (2/2)
In entirel unrelated news, no I am not shopping for another bass guitar but apparently it took me 20 years to figure out that the P bass has the sound I was looking for.
2
The great thing about having stuff happen to you in your forties is that you‘re this close to midlife crisis anyway and therefore... ah, forget it, it is horrible.
3 View thread →
Replying to @shuheikagawa
I was honestly thinking what JS framework/new programming language is raffaelo?
3
Replying to @DJCordhose
My involvment with Zalando is exactly what makes this an unpublishable blog post ;)
Replying to @DJCordhose
BTW, that post would be about how microservice architectures make it HARD to build great software because of technical, organisational, and cultural aspects. Recently it occurred to me you might have thought it is an argument for microservices ;)
Replying to @j_c_cabrejas
Well, I don‘t have any comparison whether recovery would be easier if I were younger, just started a year ago ;) But I am indeed happy that „even at my age“ I am making progress!
Replying to @evilttwinn
@SergeyVAlexeev @stephens_ben x% of y is x / 100 * y and it doesn‘t matter which order or when you do the divide by 100.
1 13
Replying to @j_c_cabrejas
Yep, that might be what's happening. I once ran 3 5km over 4 days and felt crushed, then started reading about recovery times and proper nutrition 😅. It seems the running part is not the biggest problem right now.
View thread →
Replying to @mariaInTech
@marsty5 The first time I "accidentally" ran more than 10k, I just laid down waiting for lunch to be ready. It was okay for a while, but lately I'm still hungry days later. Good luck! ;)
1
The problem when you manage to run 10km is that even after a full dinner, you're still hungry.
2
Replying to @davidwhite_ai
@intel I have seen quite a few talks of Intel at big data conferences that imply it is already happening.
1
RT @XandaSchofield: Me: Machine learning is overhyped, even anthropomorphized, to a point of being dangerously misinterpreted by the public…
Replying to @DJCordhose
@RobinDrexler @YouTube For me it was a winery for a while! 🍷 🕺🏻
1
RT @DJCordhose: Only today my "Deep Learning Crash Course" is being featured as Manning's Deal of the Day, along with Grokking Deep Learnin…
Replying to @baggerspion
@therealpadams @RedisLabs It is just so hard to find the right legal setup to outsource development to unpaid hobbyists and professionals, you know.
1
Replying to @drewconway
If she really did what she allegedly did, that could also be read as really ill advised backup plans.
1
RT @maaminian: I assume someone has made this joke already... @RachandRolle #MachineLearning https://t.co/ydl1tZAjdt
Media
I‘m pretty happy that the statement that got the most laughs at my last talk was when I answered the question „given all these advances in AI, what do you recommend to your children they should learn“ with „they should do whatever makes them happy.“
3 28
RT @ian_soboroff: I can't decide which is more needlessly complicated: deep learning or LDAP.
@gergrubi Thank you! And yes, it works. You also have to balance out the person who looks confused and the 30% who are on their laptops ;)
Replying to @iamtrask
Yes! In this time and age there mustbe ways to make this work virtually: discussion boards under posters/presentstions AMA style for poster sessions.
View thread →
Replying to @DJCordhose
Can‘t say more without disclosing my favorite interview questions 😅
1
Replying to @DJCordhose
It‘s not like specifically ask about it, but it predictably comes up when discussing dealing with html pages (as one kind of data).
1
Replying to @thatferit
@FokusMan @Medium I‘d also say this is very out of character for @medium ;)
1
Replying to @shuheikagawa
But don‘t you have to be early in bed on Sunday to be early at work on Monday? ;)
1
TFW you‘ve already moved on and everyone is still on the Peak of Inflated Expectations.
6
Replying to @ivan0yu
It‘s a bit like the const in C++ good idea but it really is a lot of visual noise and has sometimes surprising side effects (in C++ ;))
Replying to @CMastication
Personal favorite: Python - Python. They seem to be the least delusional.
2
RT @CMastication: Brutally honest hot take here. This also confirms my own experience thus rendering it correct. #python #Stata #spss #sas…
Replying to @baggerspion
@therealpadams @jargnar @truemped Yes in the end it‘s all bits and bytes, but programming languages give structure, inform and enable our thinking. Some things are easily expressible in one language and painfully hard in others.
1
My colleague: "If I may go meta for a moment, I really like how we're in a constant retro and always looking for ways to improve things. Wait, now this was meta-meta." Me: "The only way to fix this is to stop talking about it." Him: "That's meta-meta-meta." Me. "Stop!"
1 7
Thread (7/7)
It is either that or I am getting old and stuck in my ways. @truemped speaks high praise of clojure, @therealpadams looks into Rust. But I have maybe already found my one true love. Programming language wise.
Thread (6/7)
There was something about Ruby that allowed me to focus on the essence. Scala got close. Python despite its awesome ecosystem never did this for me. #confessionthursday
Thread (5/7)
Gawd this sticker taunts me, reminds me of a time when I would pick up my computer as if it were an instrument to sketch fragments of songs. #havingamomenthere
Still contemplating whether to put a Ruby sticker on my laptop. Thing is, I always felt like Ruby is the script language that embraces its scriptiness the fullest, hacks‘n‘magic all the way. But it is also so out of fashion...
2 View thread →
Replying to @badnetworker
Reminds me of Isaac Asimov‘s stories. Didn‘t they have a system to predict the course of society?
Replying to @duergner
@therealpadams @fuenfrhythmen I thought that‘s what the 20s and 30s are for ;)
Replying to @noootsab
Yeah, once you enact all the participant things run much more smoothly imho anyway ;) But seriously, I think you need to focus on (a) nice people that (b) can represent stakeholders instead of just adding anyone who might have a stake.
Replying to @baggerspion
@therealpadams You think it is safe to teach her about satire already?
Meeting anti-patterns! So you are clear about the purpose and have an agenda. Congrats! Now please: - don’t just invite everyone with an opinion or who has a stake, - use visual aids to track the discussion (eg whiteboard, post-its), - decide on next steps. Thank you!
1 3
My 11yo daughter has started to ask me „so what are the benefits of Brexit“ every other day now. I think she is just learning how to troll... ;)
16
Replying to @mkuegi
Yeah, and then there are other factors... like designing for extensibility... *cough*
1
Replying to @superglaze
A few years back my impression was that in the US people were mostly concerned with the govmt collecting data on their citizens. Good to see that is changing...
2
RT @markusandrezak: They have some grim sense of humor over at facebook, calling the app “fb Research” like in User Research like “we want…
Replying to @baggerspion
@therealpadams @rustlang I read somewhere that these two are often mixednup because they came out at around the same time but actually rust is for systems stuff and golang is more for building services.
That one time I built a Twitter search engine with bespoke b-trees through handcrafted memory mapped files ranked by a citation inspired metric and held a whole year of 50 tweets/s on my laptop.
7
So I did this thing where you sit around in an Apple store while people magically appear out of the crowd to have my battery replaced. No need to get nervous at 1% anymore, @truemped!
2
Replying to @baggerspion
@therealpadams @EmilyGorcenski It is snowing?! *jumps out of bed rushes to window*
Replying to @noelwelsh
It's not like I believe everybody thinks about me all the time! 😅
Replying to @noelwelsh
Yes, I agree. But let's say you give a presentation and are worried about people in the audience. It's all in your head, you don't even know them. But do you think they are potentially getting mad to help you? What is the importance of that to you?
I've learned this the hard way, whenever you put ideas into something concrete - writing, code, a talk - choose the medium to match the amount of change you expect to have. Rough sketches in the beginning, fill in details and polish at the end.
2
Replying to @noelwelsh
I hope that sometimes it is yes! I've started to ask me that question more regularly, and the answers... are difficult ;)
1
Okok this seem to be off topic, but for all of you concerned about what others think of you: do you believe what they think is meant to help you?
2
Replying to @AstroVivi
@adilaijaz @mnandi92 @stpyang @frlazzeri You're welcome! I'm glad you found it helpful! :D
2
So TIL you shouldn't slack on getting rid of your Xmas tree as spiders might lay a nest in the tree and some day two dozen little spiders might hatch. #sparingyouthepictures
2
Replying to @try_except_
Xubuntu has always been my fav ubuntu distro, too! Even made myself a xubuntu T-shirt once.
2
Shout out to my colleague who is so kind that he keeps on insisting that MF stands for "my friend."
9
Replying to @soblom
@pawelbrodzinski Now I‘m intrigued to know how that continually improving coffee table would look like... 🤔
Don‘t even know what that broom-box-like thing in the last pic is 😂 https://t.co/c0NThKy2wY
RT @noelwelsh: @mikiobraun This is why real programmers use mechanical keyboard---no electronics to hack
Replying to @datitran
Yes, I think that's the beauty of being in a growing and fresh field, as roles expand everyone can find a place they enjoy working in.
Replying to @datitran
I‘d probabl argue that not all of those 90% must be considered non-fancy. Deploying models and telling business folks about ML for example. ;)
2 View thread →
Replying to @aschrock
When I left, people were like „but you still come to the university one day a week, right?“ Never regretted it.
1
RT @stadtlegende: About a year ago the @ApacheBeam community went out on a quest to add Python and Go support to its Runners, which could o…