HNHacker News
TopNewBestAskShowJobs

troelsSteegin

821 karma · joined February 12, 2017

submissionscomments
troelsSteegin··on The AI Boom Has Silicon Valley on Another Manic Quest to Change the World
From the snippet: "How a seminal 2017 paper by Google researchers laid the groundwork for the AI hype cycle, resulting in a Silicon Valley frenzy not seen since the dot-com boomFind"

What was the paper they cite? "Transformer: A Novel Neural Network Architecture for Language Understanding" [0]? "Attention is all you Need" [1]?

[0] https://ai.googleblog.com/2017/08/transformer-novel-neural-n...

[1] https://arxiv.org/abs/1706.03762

troelsSteegin··on Two years of teaching high school CS
"The Skillful Teacher", by Saphier, Haley-Speca, and Gower, is an authoritative and comprehensive guide to the craft of teaching. Like the commenter says, there's a lot going on, and so potentially a lot to juggle.

https://www.rbteach.com/resources/skillful-teacher-7th-editi...

troelsSteegin··on Money, the Unit of Caring (2009)
I read the article as asserting that all action must be economically observable in order to be action. That action requires concrete allocation of resources, and that in order to care, you have to do something. Otheriwse, I agree.
troelsSteegin··on A prompt pattern catalog to enhance prompt engineering with ChatGPT
"systematic ablation" - https://stats.stackexchange.com/questions/380040/what-is-an-.... How do you ablate a black box? You work with the inputs. I agree that this article lacks rigor, but I don't mind some structured thinking about what seems to work. I don't think the authors over-promised here.
troelsSteegin··on Ted Chiang: ‘The machines we have now are not conscious’
"So if he had to invent a different term for artificial intelligence, what would it be? His answer is instant: applied statistics"
troelsSteegin··on More Than Just Algorithms
This is an interview with the authors of "Data Science in Context", https://datascienceincontext.com/, about their intentions behind the book, which I'll describe as a framework for applied data science. The book looks like a textbook for people outside CS, as written by luminaries in CS - Norvig, Wing...
troelsSteegin··on Should we take measurements at an intermediate design point? [pdf]
This is from 2000.

Off topic, but it would be interesting to a predict plausible "next posts" from a given HN user. When I see some interesting but random seeming post, like this Gelman paper, I often try to infer the context behind the post. Is there something in the zeitgeist that makes this post more relevant today? Or is it just noise? The user-posts predictor would look for topics of interest, sources of interest, and trends in topics. Is the user onto something? Should I pay attention? The worst/best part of HN is the interesting noise. I am sure there's a Gelman paper on that.

troelsSteegin··on Keith Lynch's timeline of net related terms and concepts (2006)
From the early '80s, I have nostalgia for "talk" and "finger".
troelsSteegin··on Yahoo will lay off 20% of staff, or 1600 people
That suggests that companies could have designed layoffs as accelerated attrition. They could make a prediction about who likely would have left on their own. They could make a prediction (or observation) about whom they were managing out. Then the first layoff target is the staff who would likely not be here in two years anyhow. That's a different approach than for example culling people with higher salaries who are not graded as extraordinary contributors. Either way, I think having an identifiable rationale for layoff choices makes a layoff less shitty to go through.
troelsSteegin··on 230 years since British pirates robbed the US of the metric system
"Beyond Measure" [0], by James Vincent, is a good well researched read on the history of metrology. He doesn't mention the Dombey story. He does conclude [p. 265] that the ultimate reason that the imperial system stuck in the US and Great Britain (100 years after Dombey) was that these economies were dominant incumbents and didn't /need/ to change. Great Britain had metric and imperial in parallel. For the US, the switching cost was too high.

[0] https://www.theguardian.com/books/2022/jun/07/beyond-measure...

troelsSteegin··on A journey in e-commerce search
Question: are they using "filters" as facets or what FAST called refiners? That's the impression I had on a read through. Endeca (now something inside Oracle) had some early success with faceted search for ecommerce.
troelsSteegin··on The last three years of my work will be permanently abandoned
His post mentions "Bean Machine."

https://ericlippert.com/2020/09/23/introducing-bean-machine/

https://beanmachine.org/

troelsSteegin··on Three common misconceptions about the nature and nurture of research
They are writing about re-inventing industrial research labs - what should a modern Bell Labs be? They advocate for application-inspired scientific research as the majority of a portfolio - but also leaving room for curiousity-driven work that may suprise.

They've a book, "The Genesis of Technoscientific Revolutions: Rethinking the Nature and Nurture of Research" [0], and a video, "The Nature and Nurturing of Research: A Modern Synthesis - Dr. Jeff Tsao" [1].

[0] https://www.hup.harvard.edu/catalog.php?isbn=9780674251854 [1] https://www.youtube.com/watch?v=ByCdXyak3Vs

troelsSteegin··on Earth now weighs six ronnagrams: New metric prefixes voted in
Via https://www.nature.com/articles/d41586-022-03747-9, "Extreme numbers get new names":

The prefixes ronna and quetta represent 10^27 and 10^30, and ronto and quecto signify 10^−27 and 10^−30. Earth weighs around one ronnagram, and an electron’s mass is about one quectogram.

This is the first update to the prefix system since 1991, when the organization added zetta (10^21), zepto (10^−21), yotta (10^24) and yocto (10^−24).

troelsSteegin··on Tech’s talent wars have come back to bite it
Across the executive mea culpas, there is this theme of "I over estimated demand". But, one could take the perspective that they correctly hired to extract value during peak demand, because on average each new resource is profitable, and are now are reducing resources because there is less value there to extract. An investor could argue that it is irresponsible not to hire for peaks and thus leave value inthe market. Tech jobs are always tied to the ups and downs of economic cycles. Perhaps the mea culpas are really "this cycle ended before I expected it to". Some roles and some firms are more buffered than others. It would be nice to know how volatile your value is.
troelsSteegin··on Hey, GitHub – Waiting list signup
What if the code in question is a DSL? Something say that is syntactically python, but with a namespace defined through a narrow set of imports. This would be interesting to explore for end-user scripting.
troelsSteegin··on Hans Niemann Suing Magnus Carlsen, Chess.com, and Hikaru Nakamura
FYI, she's still in prison: https://people.com/sports/brittney-griner-releases-statement...
troelsSteegin··on Selecting Meaningful Metrics with Goals and Signals
This is from the "Software Engineering at Google" book, https://abseil.io/resources/swe-book

Goals-Signals-Metrics seems like a nice way to organize thinking about performance. Predictive signals or metrics are the grail - lagging signals or metrics don't tell us much about what to do to improve.

Maybe it is interesting to compare metrics in the context of engineering centered processes (this) vs product development or product investment processes. The latter is more business focussed, and looks not our efficacy (SWE) but our business impact relative to business expenditure. Maybe that's a question of whether Product Management is sitting in Engineering or not. Are we getting things done well (SWE) or are we building the right things (product). The product POV never not has an impact on dev quality of life.

troelsSteegin··on Unfortunately for the Star, that person was their fashion editor
The suggestion is that in the face of something overwhelmingly weird, you cling to whatever structure you've got.
troelsSteegin··on Couples therapy for startup co-founders
Is “cofounder coaching" all that different from a service a management consultant might typically perform? In any event, getting in the same room together and listening without interrupting seems like a pretty fundamental part of improving a relationship, no matter what that's called.
troelsSteegin··on Categorizing design decisions in enterprise product development
I would start here: https://djpardis.com/
troelsSteegin··on Cognitive loads in programming
Would GOMS [0] be a useful framework for thinking about cognitive load in programming? GOMS is a framework for analyzing user workload in system interaction. Bad or gnarly code creates a much more complex interface for the programmer. A quick look did not surface a reference here, but I'd look further at work by David Kieras [1].

[0] https://en.wikipedia.org/wiki/GOMS [1] https://www.researchgate.net/profile/David-Kieras

troelsSteegin··on Ask HN: I want to dive into how to make search engines
I would include (5) the presentation of search results - document snippets or surrogates and affordances for query refinement like faceting. One salient acronym is SERP, for search engine results page.

A different "information retrieval" perspective overall would be to take a deep dive into the career of Susan Dumais, http://susandumais.com/

troelsSteegin··on The Hub+Spoke Content Strategy of 100k+ follower B2B creators
I am not familiar with "holotypes" in the "content strategy" space. From context, it looks like "genre". Elsewhere [0] it looks like "exemplar".

I am awaiting some future post on "holotype tips" ... How does any explicit content strategy in the hub+spoke vein not come across as inauthentic? Maybe this is a job for a Brechtian holotype.

[0] https://sciweb.nybg.org/science2/herbarium_imaging/typedefin...

troelsSteegin··on The great Miyazaki watch of 2022
Vocab word of the day: "eucatastrophes": "a sudden turn of events in a story which ensures that the protagonist does not meet some terrible, impending, and very plausible and probable doom.", via https://en.wikipedia.org/wiki/Eucatastrophe

I never would have connnected Miyazaki with Tolkien.

troelsSteegin··on The Four Dirty C-Words of the Internet
The empty c-words content, culture, community, and creator all up to the fifth: empty consumption. Why do we keep clicking?
troelsSteegin··on HBO Max pulls nearly 200 ‘Sesame Street’ episodes
Sesame Street feels like a important piece of American culture. "Mr Rogers", too. Like, if there was a National Film Archive for TV... What preserves these shows, what guarantees access to these shows? The shows are of course also someone's private property. This action by HBO Max reduces them to nothing more than a means to extract marginal value from a market.

Sesame Workshop and HBO Max/WarnerMedia/Discovery+ could better serve the public and their markets by committing to metered access to the full library. Two episodes a day should be enough for anyone...

troelsSteegin··on Nobody wants to teach anymore
As a parent, you can be well very well informed on the issue, and still not be able to solve it in your kid. I'd venture that this kind of futility is more prevalent than ignorance.
troelsSteegin··on More content by people, for people in Search
What reading the post suggests to me is that they are identifying high quality chunks of content as unique, and then ranking pages up or down in part according to what was seen to have published a given chunk first.
troelsSteegin··on Paper: Augmenting Decompiler Output with Learned Variable Names and Types [pdf]
This work reads to me like a neat smart hack. It's the kind of thing I miss reading about in more ambitious systems and learning papers. It "magically" reverses elements of a transformation. If I read it right, it follows a pattern of taking a structured input (source code), an unobserved but consistent transformation (compilation), and an output with latent structure (binary). They trained a transformer model in order to extract features from the output and label them with features that make sense in the representation of the input. This is harder if the transformation is noisy, perhaps as from different inlining behaviors. This is harder if the feature space of the input does not have labels that make sense to the model consumer. What they did seems like fun.
← PreviousPage 3 of 9Next →