carls

I've seen this being posted all over, but people rarely seem to realize that data here is from 2023.

I think it's reasonable to be turned off by a slick-looking website, but I imagine it's because the intended audience of the website is the general (dog-owning) public, likely for the purposes of soliciting participants.

Interestingly, through engaging with you I discovered that this is a cognitive bias called the "horn effect" and is the reverse of the more common "Halo effect": https://en.wikipedia.org/wiki/Horn_effect#:~:text=The%20horn....

This project is a research project out of the University of Washington, led by several of the professors there. I believe the lab is the Healthy Aging and Longevity Institute.

They share a list of academic publications that have resulted from the project, and their Team page lists the full names a sizable large number of people.

Their FAQ indicates that the cost of the DNA Kit and other things are covered by the project funding. [1]

What made you think that it's engaging in fraud? I'm genuinely curious.

I'm not involved in the project but just from looking at the site for several minutes, it seems to be a fairly reasonable research project.

Or did you say "fraud" less to mean "these are people who are stealing money and e.g., hoarding it away" and more to mean "these are people engaging in a research project I disapprove of"?

[1] https://dogagingproject.zendesk.com/hc/en-us/articles/441699...

There are probably several, but one of the ones that made the rounds a while ago is "The Man Who Killed Google Search" by Edward Zitron: https://www.wheresyoured.at/the-men-who-killed-google/

I skimmed the technical report: https://cosine.sh/blog/genie-technical-report

At the bottom, they noted the following:

SWE-Bench has recently modified their submission requirements, now asking for the full working process of our AI model in addition to the final results -their condition to have us appear on the offical leaderboard. This change poses a significant challenge for us, as our proprietary methodology is evident in these internal processes. Publicly sharing this information would essentially open-source our approach, undermining the competitive advantage we’ve worked hard to develop. For now, we’ve decided to keep our model’s internal workings confidential. However we’ve made the model’s final outputs publicly available on GitHub for independent verification. These outputs clearly demonstrate our model’s 30% success rate on the SWE-Bench tasks.

Their model outputs are here: https://github.com/CosineAI/experiments/tree/cos/swe-bench-s...

It seems to me that there's a natural tension in emerging fields of science and engineering between establishing clear guidelines and regulations early on to minimize harms, or instead allowing practitioners to experiment, tinker, build and create outcomes that may be potentially harmful.

What are some frameworks for how to think about navigating this tension in emerging scientific or engineering fields?

Some ones I'm mulling over:

1. Rate of innovation: In rapidly evolving fields, imposing strict regulations too early can hinder innovation and progress. In such cases, it might be better to minimize restrictions early on to allow practitioners to explore new ideas. Then, as the field matures, regulations and standards can be gradually introduced.

2. Adaptive regulation: Implement a flexible regulatory framework that can be updated as new information becomes available.

3. Self-regulation: In some cases, maybe we should expect and encourage the industry to use self-regulation via developing guidelines and codes of conduct. This may be one way to try and strike a balance between responsible innovation while minimizing bureaucratic obstacles.

What do others think?

I took a class with Professor Ousterhout. He would end every Friday's lecture with a "Thought for the Weekend", such as this one.

It was very entertaining and charming to hear him discuss his personal and professional life, and lessons he's learned throughout them often occasionally have very little to do with computer science.

I don't remember all of his "Thoughts for the Weekend", but I do remember one story he told about wishing he had apologized sooner to resolve some conflict he was in. That was a bit of wisdom that stuck with me from the class, beyond any of the computer science topics we covered.

The article does in fact mention this event in the below paragraph:

"The hypothesis took another hit last July when a bombshell article in Science revealed that data in the influential 2006 Nature paper linking amyloid plaques to cognitive symptoms of Alzheimer’s disease may have been fabricated. The connection claimed by the paper had convinced many researchers to keep pursuing amyloid theories at the time. For many of them, the new exposé created a “big dent” in the amyloid theory, Patira said."

I think the situation is more complex than just "we must assume they don't care about learning ... just fail the bastards." I think it's more important to first understand what were the social, environmental, cultural (and otherwise) causes of this behavior.

Specifically, different systems of incentives and permissiveness will produce different behavior. I taught high school computer science for 4 years, and I can attest that cheating occurred in the classes I taught. I've also been enrolled part-time in Stanford's MS in CS, and have taken a number of the core undergraduate curriculum for CS majors.

I also went to a hypercompetitive US public high school with a number of brilliant classmates, many of whom also cheated.

My experiences have showed me that there is a wide spectrum of "cheating", ranging from students sharing things like, "I was at office hours and heard from the TA heard from the professor that topic X is going to be really emphasized on the exam, so you better study for it!" to outright blatant copying of other's code or answers.

What I've noticed as qualities of a learning environment that seems to increase the likelihood of cheating are:

1. The technological ease of which it is to cheat: it's easier to cheat on an asynchronous online exam than when you're taking it synchronously in a large classroom.

2. How "high stakes" the course is for students: for students at institutions like Stanford, where they be used to a certain level of academic success, failing a course isn't just a blow to their transcript -- it's a psychological blow to their identity as a "smart student." They may find it easier to cheat and maintain their self-image (and projected image to their family/friends) as a great student than to take the honest hit to their GPA, and have to give up their identity.

3. How "legitimate" the course feels: classes where the instructor is widely perceived as "unfair" or "incompetent" seem to have more cheating. Students feel disrespected ("How could she put X on the exam? We barely covered it!") or unvalued ("He doesn't even bother giving clear instructions on the homework assignments. Why should we respect his test?") may try to 'retaliate' by cheating.

4. How permissive the academic culture is around cheating: if there is widely perceived to be little-to-no consequences to cheating, or if cheating is seen as, "well everyone does it", then you will have a lot more cheating.

I'm sure the above is not an exhaustive list. My broader point is that in order to address the issues around cheating, we need to be more encompassing than simply punishing the cheaters. If the stakes are high enough, and the incentives strong enough, cheaters will still exist even if they are aware of the severity of the punishment.

I recently took a CS class at Stanford with an interesting policy on cheating. While cheating almost certainly happened during the course, at the end of the quarter the course staff made a public post allowing any student who cheated to make a private message to the staff admitting they've done so.

If a student admitted to cheating, while they would face academic disciplinary action (i.e. receiving a failing or low grade), they would not be brought up to the administrative office that deals with issues of academic integrity, and therefore would not face consequences like expulsion or being on official academic probation.

However if a cheating student decided to risk it and not admit their guilt, they were at risk of a potentially even greater degree punishment. The course staff would run all students code through a piece of software to detect similarities between each other, as well as online solutions. Students who were flagged by this software would then have their code hand-checked by at least one course staff, who would make a judgement call as to whether it seemed like cheating.

I found this policy quite interesting. As a former high school teacher, I've certain encountered teaching in my own classes, and have historically oscillated between taking a very harsh stance, or perhaps an overly permissive one.

The one taken by the lecturers of this course offered a "second chance" to cheaters in a way I hadn't seen before.

I think a fundamental mistake you (and other commenters on this article) are making is judging the value of the "plain" English version by whether you think it's good writing.

However, the article's intention is to use plain language to be accessible to individuals with intellectual disabilities or other difficulties in language (i.e. recent immigrants).

As a fairly well-read and educated person, I also find the "plain" version dull and uninspiring. However, I accept that the article is trying to make the point that such writing may be more broadly understandable.

For example, I recently had some relatives immigrate from a non-English speaking country. I helped them set up internet and noticed there were multiple points during the company's signup and payment process their lack of English fluency created huge hurdles.

I think the $100bn number does not refer to the company revenue. According to the company, in 2016 revenue was close to $2bn [1]. It may instead refer to total assets under administration (which was $74bn in 2016, according to the same source).

-------

[1] https://www.oneamerica.com/newsroom/news-releases/oneamerica...

Although the author may be wrong about whether the 10-Ks he saw were the same ones being submitted to the government, I think his curiosity is still worth indulging.

Why did Netflix slowly stop designing their 10-K reports that lived on their own investor relation page?

My own guess is that there was some downward pressure to just produce the basic report, since that's all the government would see and investors wouldn't really value the design.

I'm not sure this is a sufficient explanation.

McCarthy left in 2010. But it looks like the decline in the "designy"-ness began in 2006.

It seems like the blog is referencing the documents posted directly on Netflix's website[1], the early versions of which do open up with well-designed cover pages.

In contrast, the link you shared is perhaps the direct filings that the SEC receives from the company, which may remove these designed elements prior to filing.

-----

[1] https://ir.netflix.net/financials/annual-reports-and-proxies...

This probably depends more on what the topics are, and how they're talked about.

I've done the Coursera course, Nand2Tetris, that inspired this website [1] and found it deeply intellectually satisfying and engaging.

If anyone is interested in learning about the logical primitives that build up to a computer, and how they're implemented using logic gates, I would deeply recommend the course!

-----------

[1] https://www.nand2tetris.org/

I teach an AP CS A class at a Bay Area school. In previous years, the class would have students write and run Java on their own computers.

This year, given the remote nature, the class has completely transitioned to using Repl.it. We publish assignments and projects on it (and have also tried syncing with Github classroom to give more in-depth comments for each commit), and also demonstrate new topics by pulling up a blank Repl.it and live coding.

While there have been hiccups and days when the service crashed during class, it has for the most part exceeded our (my co-teachers and I) expectations.

I appreciate that they have a vibrant community. There seem to be lots of tinkerers/hackers on the platform who are doing weird, interesting things (I saw a Python implementation of a text-based Among Us game the other day [1] and a turkey translator [2]). There are also other CS teachers who have published their exercises/projects on Repl.it free to use or remix.

There are also a number of tutorials (i.e., here's one for this year's Advent of Code Day 1 [3]). These give me favorable impressions of the active community.

I don't have any comments on their long-term path to profitability and growth, but the personal experiences I've had with it have been mostly positive so far.

------------

[1] https://repl.it/talk/share/GAME-ROOMS-Among-Us-ALPHA/79187

[2] https://repl.it/talk/share/Turkey-Translator/82208

[3] https://repl.it/talk/learn/Advent-of-Code-1-Walkthrough/8353...

This is the way that I wish more teachers taught. I love finding analogies or visual diagrams for CS concepts to share with my students.

Request: can others share links to books, courses, videos, articles or other resources that employ incredibly thoughtful pedagogy in explaining, rather than diving into the nitty-gritty details immediately? I'd love to start a collection of examples to share with others.

My question in the original comment is not meant to mock or poke fun at those who enjoyed it. I, myself, found it somewhat entertaining to read as a kid.

Rather, my question is more in reference to what seem to be incongruous facts: the article points out that the comic was not written with the intention to be funny, so then what explains its popularity as a comic strip?

I don't doubt that there are some (many!) who thought it funny. What I'm trying to understand is the discrepancy between its immense fame AND its apparent lack of humorous intention.

It'd be a little bit like if there was a famous band that was incredibly popular, yet in multiple interviews they reveal that they don't put a great deal of effort behind making appealing music. As such, I would be curious to understand what instead may be the other driving sources of their appeal (ex. good looks, marketing etc.).

For example, here are some other possible candidate reasons the Garfield comics may have been popular. I don't know if any of these are true, rather I would consider them hypotheses that I would be curious to hear others confirm/debunk:

* it was marketed very well and gave people the perception that it was _supposed_ to be funny, and if you didn't find it funny that was perhaps a result of _you_ not getting something. * there was a more limited selection of sources of comic strips, so the standards for what passed as an entertaining comic strip were lower than in our Internet age. * most people knew it wasn't funny, nor that it was meant to be funny. Rather they read it because it was the cultural meme of the day to do so.

I didn't know how much of Jim Davis' motivations were commercial ones. It seems very at odds with folks like Calvin and Hobbes creator Bill Watterson:

For years, Watterson battled against pressure from publishers to merchandise his work, something that he felt would cheapen his comic. He refused to merchandise his creations on the grounds that displaying Calvin and Hobbes images on commercially sold mugs, stickers, and T-shirts would devalue the characters and their personalities.[1]

I'm curious: what led to Garfield's immense popularity as a syndicated comic? Were there in fact legions of people who _did_ find it funny, or entertaining?

---------

[1] Wikipedia: https://en.wikipedia.org/wiki/Bill_Watterson#End_of_Calvin_a...

I spoke on a panel with Rick Sung, the Undersheriff who allegedly withheld permits until Apple coughed up the iPads.

At the end of the event, each panelist was offered a Starbucks gift card by the event organizers.

Rick gave his Starbucks gift card to me, saying, "since I'm here as a public official I can't accept this. Otherwise it would be counted as a bribe."

Wild.

I definitely think it's a good point to support Python/R for data manipulation, and allow to also JavaScript for complex visualizations.

Their current target for paid customers seems to be infoviz folks at news outlets (the CTO and co-founder was at NYTimes' data viz team for a number of years): https://observablehq.com/teams

That doesn't seem like a huge market, but perhaps they're starting with this audience segment and will be branching into more DS-type features (i.e. supporting a Python kernel, trying to be a more feature-rich Jupyter).

It looks like their current plan for monetization is a per user subscription fee for added features on the notebook: https://observablehq.com/teams

Given the marketing on the page, it also looks like it's targeted at business use cases (i.e. a bunch of data scientists who want to visualize something in d3 but also want to collaborate).

I've used Vega-Lite and d3 before and have appreciated both. Vega-Lite seems to be great for rapid prototyping and d3 for really refined, intricate and more complex plots.

I checked out the Falcon documentation on Github and currently don't have a great understanding of (a) what it would be like to "write in Falcon" and (b) what it's intended use case is and how it differs from existing libraries.

Do you mind clarifying?

Can you share how you were able to embed the notebook into your blog post?

It's a bit buried, but if you click through the link about how today is a big day for Observable, you'll notice it links to an article announcing that they raised their Series A today: $10.5M from Sequoia and Acrew.

A lot of the code you can write does look like JS, though. You can use the libraries that may be common to a JS dev. It's not perfect, but there's enough overlap and the differences are understandable and documented[1].

Also having learned a lot of D3 by reading through Observable tutorials, I'm curious what you've seen that doesn't seem to work when you port it to JS?

One thing I'm aware of is that you have to move a lot of the plotting code into the .then() method of promise when you're loading data.

------

[1] Observable's not JavaScript: https://observablehq.com/@observablehq/observables-not-javas...

I'm a big fan of Observable, having used it to prototype and learn a number of different visualizations. Note that you don't just have to use D3, but can use other visualization libraries as well (i.e. Vega-Lite, Highcharts).

I also want to shout out Mike Bostock, one of the company founders (and creator of D3). I emailed him randomly to ask for some help with a d3 package and he replied the next morning.

Busy creators who nevertheless still make themselves available to engaging with the community always impress me!

Shriram did an excellent job being a lively, energetic and engaging speaker who distilled some of the core concepts of the paper down without getting tangled in the formalisms.

I do not have an academic background in theoretical computer science and was nevertheless able to follow along his talk perfectly well.

For anyone interested in reading it, the paper by Matthias Felleisen that this talk is based on can be found on the author's website: https://felleisen.org/matthias/papers.html