Mikio L. Braun

ML and AI expert

@mikiobraun

@mikiobraun Twitter Memorial

19,827 tweets · 2008–2024 · 1046 threads

2018

RT @wardleymaps: Let’s clarify the lockin myths a bit, as they are often used as an argument against a solution we do not like.
Replying to @vambenepe
I yet have to figure out the magic by which Linux can do updates while continue to run. #maybesanekerneldesign
1
Replying to @noootsab
@EASPORTS I had a student once who did all his Ph.D. research in one gigantic C++ project. That was the first time I regretted not to have taken a look at code earlier.
Thread (3/3)
Then again a friend who is a prof for programming languages told me he once saw a poster on a compile time interactive debugger for C++ template statements.
Replying to @earino
Then don‘t! If you have the powrs necessary to become a supervillain you certainly have it in you to stop you! ;)
2
Replying to @noootsab
Does anyone do data science in C++? I mean the actual data analysis part?
Replying to @jessitron
@avdi I read somewhere that there is no language in the world where "why" is not accusatory at least a bit. They suggested to ask "what did you hope to accomplish" instead.
1 4
Replying to @DJCordhose
Ah you mean a role is more like a set of activities etc. one is doing at a time. Hm. More like a position in soccer maybe? 🙈 #hereitcomes
Replying to @DJCordhose
That brings us to the question when to create specific roles at all. My guess is if it is either very specialized (or too confusing otherwise) and people need to focus, or when you can parallelize work if it is separate roles.
1
Replying to @DJCordhose
I think it gets a bit better when data scientists either move closer with product management people or with engineering, at least they should work closer with another role.
1
Replying to @DJCordhose
I think on some level it is not different from other Xfunc teams (like frontend vs backend). But then again the work of a data scientist can take weeks to finish - always a challenge for any kind of planning ;)
1
Replying to @DJCordhose
Yeah, I think it can be part of a team but as always with cross-functional teams, making sure everyone has something to do at all times is a challenge.
1
Replying to @DJCordhose
Yes, I agree. Whether you have data and/or it will allow you to learn what you need is often an overlooked risk. The hype around the power of AI doesn’t really help either ;)
1
Replying to @DJCordhose
It depends ;) Definitely most of the uncertainty should be addressed. I‘ve seen projects which started with how to present results to the customer before even training a model.
Replying to @Springcoil
It is tricky. I think in the end, you need to combine the patterns you've learned over the years with enough detail knowledge to still be able to make reasonable predictions. So "yet another framework for X" is not enough, but saying "framework for X, but good at Y" is good.
2
Replying to @Springcoil
Well not sure about the list, but I believe that (1) there is pressure to keep up to date with too many things and (2) the principles will outlive the concrete tools we have today.
My 11yo daughter on me explaining that the raspberry pi is a full computer: „You know, if I didn‘t know you well I‘d think you‘re crazy. A COMPUTER WITHOUT A KEYBOARD!!“ 😆
Media
5
For 2019, contemplating to go with initials only => ML Braun... too late? Too early?! Btw, could also be Dr. ML!
5
RT @whyevernotso: Names for collections of code in various languages: A pile of JavaScript A crystal of Haskell An undefinedness of C++…
Replying to @noelwelsh
You‘re not. Although it barely lines up as nicely as in your case ;)
1
Replying to @baggerspion
@therealpadams @ZalandoTech @truemped I have to ask him! Oh my, did I not ever ask him yet?!
Replying to @baggerspion
@therealpadams @ZalandoTech @truemped I guess it is worth a try....
Replying to @baggerspion
@therealpadams @ZalandoTech Who will I be chasing helicopters with? 🚁
Replying to @baggerspion
@therealpadams @ZalandoTech You‘ll be painfully missed! Who do I spontaneously bad-cop-good-cop-dont-mind-we-didnt-talk-about-this-well-figure-it-out with?
With 2018 winding down, I think I never wrote this much... text... I barely remember what words mean 😅
1
Replying to @markusandrezak
Holy moly that guy is savage. „Okay so that‘s the cold shock response... it‘ll pass“ 😱
1
Replying to @bobbruno70
Also a very interesting point. Yeah I think our mental model of just how much we can incrementally influence customer behavior on a website is probably too optimistic ;)
1
Thread (5/5)
A less favorable explanation is that Germans don‘t believe in creating opportunities so every problem solved is one less possibility to prove yourself and we‘re just jealous. ;)
2
Thread (3/5)
Never quite got to the bottom of it, but it feels like the comfort comes from the fact that there is still work to do. Very different from - ahem - other cultures where everything is always „awesome“ ;)
4
Thread (2/5)
Now in light of the ongoing NeurIPS, a typical discussion with fellow German colleagues is that „none of this is actually AI.“ And then everyone nods and smiles in agreement.
1
One thing I (as a German) consider to be very German is the understanding that everything is a bit broken. There‘s no shame here. Everyone will agree. I feel the main driver for innovation is not joy of creation but the satisfaction of having done at least one thing right ;)
4 View thread →
"Show me your flowchart and conceal your tables, and I shall continue to be mystified. Show me your tables, and I won't usually need your flowchart; it'll be obvious." -- Fred Brooks, The Mythical Man Month (1975)
1 3
RT @DJCordhose: @mikiobraun a more general thought: I guess our actions are more limited in reach than we hope they are. We can not make pe…
Replying to @DJCordhose
I am thinking maybe one needs to test customer touch points for how much impact they make before trying to improve them.
1
Replying to @DJCordhose
Maybe. That could be taken into account before putting months of work into something maybe ;)
Replying to @mdespriee
It seems hard to do something differently but still get almost exactly the same performance.
Replying to @mdespriee
Hehe yeah. But still it makes me wonder. Because there are sometimes huge differences in how things work and what they do. Still no effect on the customer - why? They don‘t care? They don‘t notice?
Replying to @mdespriee
Yes. So what I ask is that often you see small changes only. Like a new recommendation algorithm gives you 5% (relative!) uplift for example. That seems so small. ;)
Replying to @mdespriee
Usually people make sure tests run long enough... but then again we all know classical statistical tests become significant eventually...
1
Replying to @DmitryKan
Interesting. You mean for more drastic changes, it takes time for people to discover and change their behavior?
Anyone else wonders why A/B tests so often show only tiny improvements in numbers? Laws of large numbers? Taking averages over a very diverse population? Changes are too small to be noticed?
1 1
You know how teachers always stress that writing skills will be essential for your professional life? Guess what, they were right 😅
2 12
Replying to @vielmettialt
@vielmetti Thanks for sharing! Haven't seen it before. But I always wondered that there is so little discussion of "how to find proofs." It seemed you're expected to discover through trial and error and maybe by osmosis.
1
One of the best pieces of advice for solving problems from one of my math profs: Take a blank sheet of paper. First thing, write down "what I know" and "what I want" at the top. In some way this is still my template to this day.
1 18
Replying to @adichad
Never thought I‘d see gerrymandering and peanut-buttering in the same sentence. ;)
Replying to @adichad
I think Conway‘s law talks about what happens *after* you decided to build it with N teams. I meant the urge to go for N teams. Maybe instead of starting with 1 and grow.
Replying to @skamille
Yeah, it also shifts the responsibility away from the person handing out the readme. "I don't know what the problem is, I wrote everything in the doc!"
6
What is the name for the architectural fallacy to build something for N teams because that's the number of people you have. Is it a corollary to Conway's law? Is this Conway's law in action?
1
@marawanot For me, one of the biggest challenges was to learn which direction the train goes. You'd have to learn about the relative location of all destinations around Berlin for that. Is Ahrensfelde west or east? And so on...
Replying to @basche42
@petervan @Pocket I forgot how Steve Yegge‘s rambling rants are just the best.
1
Replying to @munterluggauer
@thinkberg @EdwardACurry @vavoida @vardi In principle yes. It is also an iterative process, so you can often not say beforehand what exactly you need. Often, which features are important also can be said only after you fit the model an did some analysis. So it‘s complicated ;)
Replying to @munterluggauer
@thinkberg @EdwardACurry @vavoida @vardi Depending on the richness of the data. If you have user click data, for example, you can look into all kinds of questions. With GDPR you should of course collect as little data as possible and needed so this will probably change.
1
Thread (2/2)
FFXV was the only FF title I ever finished. Although I got close with FFXIII - twice, a second time after the hard disk broke that contained my save file.
1
Thread (2/3)
@tarah @EFF It has happened occasionally, like with the Journal of Machine Learning Research (JMLR), but having to figure out the technical infrastructure is maybe still a pain point.
1
Replying to @tarah
@EFF If there is one industry that needs to be disrupted SF-style, it is academic publishing. There are not technical limitations, if the leaders of a community would decide that from now on they don't need the big editors, it would be over.
2 View thread →
What irks me to this day is that platform is written Plattform in German. (where "platt" means "flat" and I suspect flat is actuall "platt" after some consonant shifting, and btw we also have the word "flach" which means "flat" 🤷‍♂️). #IANAL #LasinLinguist
3
Replying to @fhuszar
Yeah I can imagine. I also didn't mean to blame people, I think it wasn't worse than in any other academic community. And there were always exceptions. I fondly remember eating with McDonald's with Jürgen Schmidhuber in the basement of the Hyatt in Vancouver. ;)
Replying to @MaineC
I remember when I was attending conferences without a laptop. I was listening, and taking notes! Sometimes it was boring, but at least I was mentally present.
2
Is it just me or is Twitter becoming more about Big Data, AI, and the Blockchain again? Or have I just found the right set of words and accounts to block?
2
Thread (5/7)
Update: but turns out the machine thinks the game is a pre-order for a yet to be released game. Although it is already released. Seriously, can MS get its act together... 😂
Thread (3/7)
I call support (M$ for xbox). They say it is a „global issue“ and I should keep retrying. I wonder what that is code for. They need to mint more digital copies? Blockchain?
1
So I ordered a pkg in CH, sent from NL. Put in a DHL station, but delivery is by UPS. They call me (how?) and I give them my work addr. Pkg still in delivery. Resent to a bakery. I go there, but there is no name on the pkg because they replaced my name by the co name. Success!
2 View thread →
Replying to @lalleal
It is half a joke, but the idea would be a bit like "schema on read," you make sure deletion etc. is handled at read time, not when you write the data. No idea whether that is even feasible, though.
And this completes my Fall 2018 Conference tour ;) New York, Heidelberg, London. Thanks for everybody who showed up!
2
David Barber: great progress in „Perceptual AI“, but limited progress in „General AI.“ #TheAIConf
2
„AlphaGo was an amazing feat, but s dead end in terms of AI“ - David Barber at #TheAIConf
7
„Learning from trillions of random interactions - that‘s not AI, that‘s just dumb!“ David is on fire! Too many truths told to keep up with tweeting! #TheAIConf
2 3
UCL's David Barber: "If something better than neural networks comes along, we'll drop it like a hot potato!" XD #TheAIConf
1 6
RT @vincenzomanzoni: #TheAIConf Last day of conference. After the keynotes, let’s start with the sessions. The first is @mikiobraun from @Z…
Replying to @alansaid
If he were right I don‘t know why I am regularly having those discussions ;)
1
AWS‘s speaker at #TheAIConf: „Working with ML is very similar to normal software development, it is all about iterating quickly.“ Well I agree with the iteration part. But the rest I think there is a lot more to this.
1 9
Is Intel‘s keynote speaker at #TheAIConf holding a paper cup of coffee in his hand?! 😚☕️
1
RT @Zalando_Press: Caroline Carlqvist today shared how #Zalando creates 24m Zalandos for our 24m customers by giving a truly #personalized…
RT @bigdata: AI systems are often hybrid, clever combinations of sensors and measurements, specialized algorithms, and some heavy machine l…
Thoroughly enjoyed attending #data2day. Great talks and overall just a great open, helpful vibe. Thanks to the organizers, my co-programm committee members, the speakers, and everyone else who contributed!
2 10
Thread (2/2)
What is behind this? AI for the sake of it? One idea I heard is the notion that "as long as it is not fully AI" a lot of human intervention is necessary. Is that even true?
1
The keynote by @ted_dunning reminds me that for some reason, the number of times I had to discuss whether collaborative filtering /really/ is ML/AI is increasing lately.
8 View thread →
Kicking off the #data2day conference in sunny Heidelberg with a keynote from @ted_dunning
Media
3
Replying to @baggerspion
@therealpadams 😱 but what about that extra box they draw around a string of messages?!?
OH: „That guy? I don‘t think he is on that project anymore, he became lead of some other team“ „But you would still accept PRs from him right?“ „If I‘d accept them? I‘d just write ‚Honored to be on this PR 👍‘!“ 😁😆
2
Colleague, seeing me getting a sandwich and an energy bar the vending machine downstairs: "so I see you went with the lunch of the champions." 😜
1
Replying to @truemped
@therealpadams @duergner You own @therealpadams nothing, he was just paying back the debt he owed his own lead who raised him!
2
Aaaaaand it‘s a wrap! #stratadata Next up: data2day in Heidelberg in two weeks and then AI conf in London in a month ;) #confautumn
2
RT @CMastication: I caught up with @mikeloukides for coffee and I explained to him that the way I solve business problems is to write crapp…
And now, my colleague Francesco Mucio on „What is this? BI for ants?!“ 🖥 🐜 #StrataData
Media
2
My favorite #stratadata quote: „before you‘re live, every failure is a failure of management“ by @jessetanderson
2 5
Very thought provoking keynote talk by @JuliaAngwin #stratadata on bias and data science.
Media Media
1
Really interesting talk about AirBnB‘s Zipline feature management project. Can‘t yet tell whether really cool or just complicated because they built on top of their DWH (did they?) #stratadata
@mmmpork @CMastication @mikeloukides Yeah never go in and just solve the problem. Nobody will cite you as there is nothing left to do. Best is to have bad solutions for promising problems!
Replying to @pacoid
It's like saying that the car industry "democratized mobility." Would one say that?
1
Replying to @pacoid
Yes, "democratization" sounds like giving people a say in influencing things they didn't have before. But data engineering right now is more about providing access to techincal capabilities and infrastructure. It is also a form of power, but isn't it different?
3
A lot of talk about "democratization of ML." I always wondered whether that is really the right metaphor... .#stratadata
2 3
RT @data2day: Mit unserer 2. Keynote ist das Programm der data2day komplett! Wir freuen uns auf @mikiobraun und seinen Blick auf die mensch…
Replying to @kostas_tzoumas
@Ellen_Friedman @dataArtisans You're motto is "living the stream"?! ;) I like it!
1
Replying to @noootsab
Btw, that‘s the same fallacy every new software framework steps in. „I have a great simple idea...“ We know how it will end ;)
Thread (2/2)
@noootsab Essentially I said each new mathematical theory started with someone realizing that things weren‘t as easy as they thought they‘d be ;)
1
Replying to @noootsab
Well yes, I said they had too much time on their hands, got bored, hoped to get some nice clean answers about the world, then discovered irrational numbers, and that‘s how it all started.
1 View thread →
Hey ho, remember we might be the product, but at least we get to write the content. Have a great start into the week, everyone!
3
So my Internet at home broke and that was the most mind shattering experience (well not really). Also my constant reliance that everything will be mirrored in the cloud. I was amazed that Xcode could compile a HelloWorld without downloading the Internet first.
4
Replying to @chrisalbon
I get this with many new games I play. NO IDEA WHAT ALL THAT IS BUT IT‘S SO SHINY. Takes a while to learn what is really relevant information.
2
Replying to @rwhitcomb
Programs = Time Machines for Text! Messages From The Past! I like it!
1
Me: installing suspicious mods for ancient versions of minecraft which crashes on startup with a huge stacktrace because it’s not Java 10 compatible. Kid: Wow look at all this text! Just image you would have had to type all that!
3 View thread →
Replying to @vodafoneservice
Bin inzwischen durchgekommen und hab einen Termin... aber das hat jetzt insgesamt 40 Minuten gedauert... Aber Danke für das Hilfsangebot!
@vodafoneservice schon dreimal heute aus der Kundenhotline geflogen nach 8-10 Minuten warten... Was soll ich anders machen?
Replying to @baggerspion
@therealpadams Also, I wonder what the Vermont State Colleges did to get on this list...
1
Replying to @xfceofficial
xfce has always been my favorite window manager, keep up the good work!!
1 11
That time of the year to try out another Social Network... If you want, follow me @mikiobraun@mastodon.cloud
1
Replying to @stadtlegende
Yeah, luckily the office building I‘m in is pretty well AC‘ed.
Who would have thought all that Java experience boils down to troubleshooting my son's minecraft installation... .
6
Replying to @munterluggauer
Oh I mean humans can definitely do this. I just find it so tedious personally ;)
1
Replying to @fs111
„Mach ein Ziehbegehren, wenn du mit der Vereinigung fertig bist!“ X-D
Thread (2/4)
Also, colleague said the phrase „no kanban boards in the kitchen“ is part of his regular interaction with his girlfriend. #Iguessthelivingroomisfinetho
1
So I am wondering, are we witnessing the massive failure of scaling up our innate abilities for social networking or is there hope we’ll find a way to organize bn+ individuals?
@_hkropp Yes I think that is right. And everything related to making that happen (e.g. what kind of teams do you need, how do you manage those teams).
Thread (5/5)
@_hkropp Of course, only if it makes sense to apply AI or ML ;) I‘m too pragmatic to push AI for the sake of it... Hope this gives an idea, thanks for asking!
Thread (4/5)
@_hkropp Not doing all of that myself, of course, but my idea is that even if I lack the expertise or capacity, I should know who to connect to or bring in to support. There are more things I am working on right now, like an „AI maturity“ index to self-evaluate and show a path forward >
Thread (3/5)
@_hkropp So I see the AI architect as a continuation of that, but across a while department, while focussing on anything related to AI and machine learning. It is not just technical arch but also talent, processes, product dev, long term strategy, etc. >
1
Thread (2/5)
@_hkropp The question is how to best do that, of course. We‘re doing a mix of ad hoc requests, projects, and more long-term technical vision and roadmapping. With the latter being the biggest challenge given lack of time, of course ;) >
@_hkropp Hey there, sorry for getting back so late! That is a very good question! Prior to this role I was architect in search. We were two ppl in that role and could gather some experience! So essentially we are there to support teams in hard and/or team crossing issues and challenges. >
View thread →
Replying to @mike_julian
@xaprb @nicolefv So I‘ve started to read that book and it is truly, life changing good! Thanks!!
1
@gergrubi @stefan_will It is just a vanity title, of course, officially I‘m still a principal engineer ;)
1
So after 5 years of CS, 5 years to get a PhD in machine learning, 10 years as a PostDoc and 3 years since joining Zalando I started my new role as „AI Architect“ today. Honestly, I was expecting sentient robots and flying cars by now, but you gotta do with what you have I guess.
8 114
Replying to @stochastician
@muratk3n You can always rewatch westworld with what you now know in mind.
RT @therealpadams: Not only does @ZalandoTech allow it, the company /expects/ it! Involvement with #opensource makes us better engineers! h…
RT @softwaredoug: Duncan Blythe from Zalando Research sharing their recent research in deep learning for search #mices2018 https://t.co/gP…
Replying to @mleznik
Not to trigger an cat persons (considering myself more of cat person btw), but I think dogs are more intelligent and social wrt to us humans than cats.
1
Thread (2/3)
Things a dog can do that computers cannot: (1) get super excited when they hear your steps dozens of meters away (2) enjoy being pet on the head (3) pick up frisbees out of the air Of course I am just extrapolating #notadogperson
1
RT @nagaraj_arvind: @zacharylipton If I have seen farther than others it's because I've git cloned repos and pip installed libraries of gia…
@majortomm Not that I say you shouldn't ;) but fingers also work fine. And the latest iPad also supports the pencil... . But the iPad Pro is also very nice. Snappy, and that screen! So many colors, so smooth :D
Replying to @kirel
Indeed, I was surprised, too ;) if only there were some way to edit tweets IN 2018. ;)
Replying to @JonnyDaenen
Anything you use for journalling? I‘m using Notability for that right now.
1
Replying to @JonnyDaenen
Yeah I use paper53. It is rather simple (no layers) but I like the way it looks.
Replying to @saeedamenfx
Thanks! I think towards the end of my PostDoc I spent a lot of time thinking about these things to decide which way to go.
Replying to @DRMacIver
I‘ve been there, jblas has my own config tool with dependency management between configs written in Ruby 😭
1
"I don't want to talk about GDPR... also because it's probably too late now" ;) IBM's Jean-François Puget at the #stratadata keynote this morning.
10
@semibogan You know I never meant to see you again But I only passed by as a friend, yeah All this time I stayed out of sight I started wondering why Now I, I wish it would rain down, down on me
RT @ZalandoTech: "I like learning, helping people & solving problems. In the best cases, all of these come together & I lose track of time.…
Replying to @DRMacIver
Hey who else can claim 10+ years working experience with Scala? ;)
1
Replying to @Major_Grooves
Ah well, if only I still cared enough about academic publishing ;) But let me know if you start something! ;)
1
Replying to @Major_Grooves
😂 everything already „exists“ on the Blockchain. Well, I‘ve mostly started to think of blockchains as distributed databases and the associated coin as the financial incentive to participate. Saying IT DOESN‘T HAVE TO BE on the blockchain.
Replying to @Major_Grooves
Hm.... maybe this screening is a service you can offer for really high submission volume journals.
Replying to @Major_Grooves
I think the only other real work with Nature is screening the thousands of submissions. That is definitely a full-time job...
1
Replying to @Major_Grooves
Yeah yeah. I wrote about this somewhere, acadmia is a global system so it is hard to change. Full editorial boards need to decide to make the switch. They don‘t seem to realize that Nature is just a platform and a brand.
Replying to @Major_Grooves
Finally, a working open source community would also potentially help with the development.
1
Replying to @Major_Grooves
Also, compared to what institutions are paying right now, everything will look cheap ;)
1
Replying to @Major_Grooves
Definitely you‘d go open source, but maintaining a web site installation forever is hard at universities (PhD students as admins, frequent job switches) so maybe they‘ll be happy to have a hosted version.
1
Replying to @Major_Grooves
One could have a flat fee or pay per publication model, and probably get public funding as well. And no, I haven‘t been thinking about this ;)
1
Replying to @Major_Grooves
I personally worked with the JMLR system which is seriously Web 1.0 (with all respect), that could be done so much nicer with current technology.
1
Replying to @Major_Grooves
Yeah, traffic volumes are moderate, you‘ll need support for configurable review flows, but could experiment with discussions / marking, integrate with mendeley or do it yourself, etc.
Replying to @Major_Grooves
Unfortunately nothing. But in this time and age, what would it take to host your own journal. All you need is something like wordpress, but for scientific publishing ;)
Replying to @mariaInTech
@marsty5 In fact email is the main comm channel here. I don't know slack, but everything's on fire for me ;)
Replying to @mariaInTech
@marsty5 Also, I think email is being used as a universal notification layer. These emails must be treated completely differently, like a stream, auto archived, and maybe have more structure to be machine readable. And so on... ;)
Replying to @mariaInTech
@marsty5 I think Google did the classification okay, but it also only goes to the thread level. But I think one could also do some graph analysis and figure out sets of people that collaborate right now and tie everything in (docs, links, etc.)
Replying to @mariaInTech
@marsty5 About email, I feel like (1) it should deal with humans vs. notifications much much better, and (2) automatically extract not just threads but teams and groups.
2
Replying to @paidicreed
I don't! I'm thinking you need a kanban board but with horizontal lanes as well or something... . It does not seem to exist. Maybe because it's a bad idea, too ;)
At this point in my career I just care about three things: people, how to fix email, and what is the right multi-project kanban-like tool.
1 11
RT @DRMacIver: Me: Learning new programming languages is a lot easier than the analogy with natural languages would suggest. Flatmate (Medi…
Replying to @vivekjuneja
@naval @farnamstreet 2nd of all, I read somewhere that there is no culture in the world where „why did you do that“ will trigger the urge to justify.
View thread →
Replying to @vivekjuneja
@naval @farnamstreet Well first of all, don't be an inquisitor ;)
1
RT @drewconway: Data is the new oil...in the way that oil is a ubiquitous commodity that requires incredible resource allocation to extract…
Replying to @baggerspion
@therealpadams My favorite one: '"I know how to spell bananana, I just don't know when to stop," says the Markov Chain'
4
Must stop watching Comedians in Cars Getting Coffee... at night. Just makes me crave a good coffee. ;)
2
At this point in my life, I can't be bothered to remember which programming languages print stacktraces inside out or outside in.
5
Replying to @tgrigoryan
I can recommend it. Modern OSes seem to assume they are talking to a SSD anyway. They also got much cheaper, paid about 230€ for 1TB.
1
Replying to @drewconway
@mikedewar For a few moments I was assuming they are messing with me. Then I realized they actually haven't been exposed to it yet :D
Replying to @drewconway
@mikedewar The other day I learned there are people who haven't seen it before!
1
Replying to @xamat
Does the DS literally say „n weeks“ or does he give a specific estimate? ;)
1 2
It is 2018, and as I ask the audience „Do you know Twitter?“ they go „duh, 🙄.“ #ThankYouGermany
1
Replying to @ChrisDiehl
Thanks, Chris! I found out for myself that I want to work with people in ways that feel right to me, not just focus on results.
1
After a solid week of sub-zero temperatures (Celsius!) the river Spree starts freezing over.
Media
4
Today someone asked me what my tech stack is, and after a moment of confusion I said "people."
18 61
Replying to @baggerspion
@therealpadams @ZalandoTech One more of those and you'll break my heart for good! 😢 Have a great (second) first day in Open Source land.
1
Replying to @PLT_cheater
@DRMacIver @flippacpub Yeah if you‘ve established what you‘d like to change and they don‘t... it‘s time to „let them go“ I guess.
When I unfold the quotes at the bottom of an email and see the whole thread is in there I think "this is the original blockchain." (minus the crypto, of course)
4 11
Replying to @struberg
@MaineC Not sure. I think the actual smart contract payload is rather small. Don‘t know what ethereum‘s mining task is. Some graph problem? But yeah, all that computation power...
1
Has someone already combined a cryptocoin with SETI@home, that would at least help advance the state of science.
1 8
Alright, I‘m saying it: Cars 3 was one if the most inspiring movies I saw about leadership in 2017.
1
Replying to @DRMacIver
Yeah, IIRC, modules are somehow classes where every def defaults to class level, and requiring a module is more like C‘s #include...
Replying to @DRMacIver
And I never can remember whether import makes symbols available or adds them to the current module or whether it physically includes source code or...
Replying to @DRMacIver
Oh yeah Ruby, I have a blog post somewhere ranting about Ruby modules...
Replying to @DRMacIver
Yes! Also: * what are the scoping rules? * what is the package import model?
1
The best thing I read in a while is "without feedback, there is no transformational change possible."
1 8
My 8yo son keeps referring to the monitor of my old school PC as the “laptop” and says things like “don’t forget to turn it off, the battery will be empty soon!”
2
Replying to @alambert
Also, you say vendor lock-in is bad, because if you want to move out you‘d have to rebuild everything yourself. So instead you just build everything yourself, probably even spending longer because you have to learn first.
5
Replying to @zacharylipton
@nytimes Counterfeit economy? Seems like there are better ways to spend your money - like investing in cryptocurrencies ;)
1
Replying to @fchollet
Also, the underlying assumption that hours put in are the main indicator of success is highly questionable IMHO.
1
Replying to @baggerspion
@therealpadams @ZalandoTech Ah, that‘s precisely how the mining works!
Replying to @baggerspion
@therealpadams @ZalandoTech Let me check whether that does not already exist...
Replying to @baggerspion
@therealpadams @ZalandoTech What does it say about me that my first question would be "why would you want to do that?"
Replying to @baggerspion
@therealpadams @ZalandoTech I'd say the answer involves founding a cryptocurrency first to handle the computational load.
Replying to @baggerspion
@therealpadams @ZalandoTech That's Adi's favorite interview question, right?
@Quora the links in your best of digest emails have been broken for a while, get me to my home page, or a "you can have Quora in German" page. Any ideas why?
@LauritaApplez @therealpadams @duergner @ZalandoTech Everybody needs a dad at times. Let‘s not make a fuzz about it ;)
1
Replying to @fchollet
Bottom line: academic publications is not the best place to look for industry best practices and sensible defaults ;)
3
Replying to @fchollet
Totally agree. Academia what it is, you need to have something novel to publish. I tried „nice and small“ several times and only got „has been done before“ and „all parts are already known“ reviews...
RT @ProfFeynman: Study hard what interests you the most in the most undisciplined, irreverent and original manner possible. https://t.co/cX…