Sunday, December 18, 2016

Quick links

Some of the tech news that caught my attention lately:
  • Humans working for the AI: How we get ground truth for machine learning ([1])

  • Deep learning helping on diagnostic medical imaging with accuracy at human level ([1] [2] [3] [4])

  • BHAG from Intel: "Intel aims to deliver up to 100x reduction in the time to train a deep learning model over the next three years compared to GPU" ([1])

  • Deep learning's success is mostly a lot of data paired with an algorithm that can take advantage of a lot of data ([1])

  • Fun! "A software platform for evaluating and training intelligent agents across the world’s supply of games, websites and other applications ... Agents use the same senses and controls as humans: seeing pixels and using a keyboard and mouse." ([1])

  • Details on Duolingo's learning algorithms, including that they found what worked best for students using A/B tests ([1])

  • A rant on hype-driven development ([1] [2])

  • Building finished products is hard ([1])

  • "There is an optimal newness for ideas -- advanced yet acceptable" ([1])

  • Massive expansion of Facebook in Seattle. Seattle is increasingly becoming a mini Silicon Valley ([1] [2])

  • Andy Jassy optimistic Amazon Web Services will become a $100B business ([1])

  • Detailed comparison of pricing on Google, AWS, and Azure. To summarize, it's complicated, and what your cheapest option will be depends a lot on what you're doing. ([1])

  • "Google has been carbon neutral since 2007, and [in 2017] we'll be powered by 100% renewable energy as our newest wind and solar farms come online" ([1])

  • Likely to see truly massive wind turbines in the near future ([1])

  • "The Waffle House Index also stands for something less obvious. It is an indicator of how complex and long supply chains are — for food, for fuel, for power — and of what it takes to plan around infrastructure that can be fragile in unexpected ways." ([1])

  • Xkcd: "Of course, 'Number of times I've gotten to make a decision twice to know for sure how it would have turned out' is still at 0." ([1])

  • "Not one, nor two, but five major VC funds reached out about investing in Rocket AI ... The ultimate fake AI company ... AI is at peak hype, and everyone in the community knows it." ([1])

Saturday, December 10, 2016

Book review: Chaos Monkeys

Cynical, mercenary, and dark, this book aptly serves as an opposing view for any idealism you may have been feeling about Silicon Valley startups or their bigger brethren. Some of us work in technology to make a difference. That is not what you will find in this book.

It is a tale of a startup that wasn't really a startup, three people with no real product acquired after 10 months. It is a tale of sales and personal marketing, spinning unfavorable realities into golden-sounding tales capable of jumping the next hurdle and moving on to the next deal. It is a tale of greed and personal ambition, everything viewed through a Wall Street lens of climbing a hierarchy of wealth and power, some in the world of venture capital, and particularly detailed at Facebook.

Facebook comes out of the book particularly poorly, as if Zuck is a some kind of fickle boy king holding court with his sycophants. During his time at Facebook, the author appears to try to join this clique, only to grow bitter when entry is rebuffed.

Most interesting is the description of Facebook's struggle with advertising revenue, especially after its IPO. As the author describes it, Facebook couldn't figure out how to make the promised revenue. Eventually, in mid-2013 or so, they found a way, not by using data on what people do, but knowing who most people are, which turned out to be particularly important on mobile ("basic targeting like age and gender was a godsend to data-starved marketers ... data-wise, you have a first-party relationship with [only] a few apps"). The real value of Facebook turned out not to be its data on what people are doing, but merely being able to identify most people consistently and willing to exploit that to its fullest.

It helps if you know at least a few of the personalities featured in the book. Paul Graham, Sam Altman, Chris Sacca, Greg Badros, and many others make at least brief appearances, usually to get splattered with the slime that drips from these pages. Many VCs and people at Facebook and Twitter are also mentioned, mostly described as the amoral who's who of the rich and powerful of Silicon Valley.

Like many who got lucky, the author confuses luck with skill. Sure, that pitch meeting went well, but that meeting almost didn't happen. Success often was a result of a chance connection at the right time. In cases where the author angered someone with his arrogance or foolishness, someone should have killed the deal, and might have had they been in a slightly different mood that day. This startup was almost stillborn, barely making it into Y-combinator. The acqui-hire almost didn't happen, almost killed by lack of customer growth and shenanigans by the author. That everything worked out even as well as it did was mostly good fortune.

To his credit, the author realizes some of this in the end. In the acknowledgments, he writes, "Let's be blunt: ours was a relationship of pure convenience, and I exploited you as much as you did me." But he also writes of some he encountered, "In a Valley world awash with mammoth greed and opportunism masquerading as beneficent innovation, you were the only real loyalty and idealism I ever encountered." I'd like to think mammoth greed and opportunism have much smaller representation than idealistic innovation.

Some may call me wishful, but I think pushing for that idealistic world to be true is part of making it true. This book is not going to stop me from thinking that tech companies should be a force for idealistic innovation and promise for the future. At least in my circles, most people I talk with are awash with idealism, a genuine belief that what they are working on can make things better for others. It saddens me to see that the author's perception of the tech industry is so different than my own.

Friday, November 25, 2016

Quick links

Some of the tech news I found interesting lately, and you might too. Heavy on the comics this time to lighten the mood:
  • Jeff Bezos: "Good leaders ... seek to disconfirm their most profoundly-held convictions, which is very unnatural for humans ... Anybody who doesn’t change their mind a lot is dramatically underestimating the complexity of the world we live in." ([1])

  • Amazon is hiring 120k employees just for the holidays. I can't believe how our baby is all grown up. ([1])

  • On building products: "Keep it extremely simple, or two thirds of the population can’t use your design" ([1] [2])

  • "The problem isn't the users: it's that we've designed our computer systems' security so badly that we demand the user do all of these counterintuitive things." ([1])

  • Fun AI experiments from Google. Don't miss "Quick, Draw!" ([1])

  • Interesting new phone design, screen taking up the entire front: "Hands down, the best looking smartphone ever" ([1])

  • Great article on Netflix recommendations, tidbits on the importance of reacting immediately to new data, using immediate intent, freshness (esp. new releases), and perceived quality (difference between online evaluation and offline). ([1])

  • Opinionated summary of RecSys 2016, and also somewhat of a summary of recommendations and personalization research as of 2016 ([1] [2])

  • Xavier Amatriain on lessons learned from building recommender systems ([1])

  • YouTube is now using deep learning for recommendations, more than just embeddings, includes a ranker with heavily engineered features ([1])

  • Ex-Facebook employee: "News Feed optimizes for engagement. As we've learned in this election, bullshit is highly engaging." ([1])

  • Pfeffer: "You need to be careful with what you measure, because you are going to get it, and often you don’t really want it." ([1] [2])

  • Obama: "Traditionally, when we think about security and protecting ourselves, we think in terms of armor or walls. Increasingly, I find myself looking to medicine and thinking about viruses, antibodies." ([1])

  • Surprising, just set up a hotspot, and the interference from people's fingers moving in the WiFi signal is enough to catch most of the passwords anyone enters while connected ([1] [2])

  • "An entire company’s product line has just been turned into a botnet that is now attacking the United States" ([1] [2])

  • short URLs hid malicious content that was then used to get at Colin Powell's e-mail ([1])

  • Carefully picked textures on eyeglass frames to fool face recognition, pictures in the paper are amusing ([1] [2])

  • AI guru Andrew Ng: "We're lucky the AI community is very open, and top researchers freely share many ideas and even code. This helps the whole field progress. Hope we can keep it that way." ([1] [2])

  • Love this: "Being able to go from idea to result with the least possible delay is key to doing good research" ([1])

  • Two new massive labeled open data sets from Google, one for images, one for videos ([1] [2])

  • "Translations that are vastly improved compared to the previous phrase-based production system. GNMT reduces translation errors by more than 55%-85% on several major language pairs" ([1])

  • Google CEO Sundar Pichai: "Our goal is build a personal Google for each and every user." ([1])

  • I got a mention in The Guardian for some of my past work: "Greg Linden may not be a household name..." ([1])

  • Data on what Amazon Echo is actually used for. Mostly playing a song, it appears. ([1])

  • Like at the last dot-com boom, there are a bunch of delivery services cropping up with models that don't seem like they're likely to be profitable. Uber, which was in a better position than most to do this profitably, just shut their food delivery service down, which doesn't bode well for the others. ([1])

  • Current state of virtual reality: "None of these uses are particularly compelling right now, especially given the cost of buying a VR headset. This may change in the future." ([1])

  • "Giving employees hours, days or even months in which to work without close scrutiny has enhanced productivity instead of harming it" ([1])

  • T-mobile's CEO on leadership: "Listen to your employees, listen to your customers, shut the f*** up, and do what they tell you" ([1])

  • SMBC comic on survivorship bias ([1])

  • SMBC comic on work, dark but funny: "The important thing is to find the low low bar that works for YOU." ([1])

  • SMBC comic on eliminating security risks ([1])

  • SMBC comic being a scientist. Don't miss the mouseover: "Hopefully your kids don't drink as much as research scientists" ([1])

  • Love the mouseover text on this SMBC comic: "Studying social science has completely obliterated my ability to enjoy pleasant human behaviors" ([1])

  • SMBC comic on political economy. Don't miss the mouseover text on humans: "In economics, they're robots. In political economy, they're all jerks. In sociology, they're all misunderstood." ([1])

  • Brilliant comic by Matthew Inman (The Oatmeal) on happiness and meaning ([1])

  • Xkcd on a CS degree: "That just means I understand how everything went so wrong" ([1])

Sunday, August 28, 2016

More quick links

A tightly curated list of what has caught my attention lately:
  • New Yorker on AI: "A lot of what people are calling 'artificial intelligence' is really data analytics -- in other words, business as usual. If the hype leaves you asking 'What is A.I., really?,' don’t worry, you're not alone .... Intelligent software helps us interact and deal with the ... [information] onslaught ... winnowing an increasing number of inputs and options in a way that humans can’t manage without a helping hand .... A set of technologies that try to imitate or augment human intelligence .... [But] we are a long way from creating virtual human beings ... In the meantime, we're going to have to deal with the hyperbole surrounding A.I." ([1])

  • Tim O'Reilly: "Humans are increasingly going to be interacting with devices that are able to listen to us and talk back .... [Alexa] demonstrates that conversational interfaces can work, if they are designed right .... Smaller domains where you can deliver satisfying results, and within those domains, spend a lot of time thinking through the 'fit and finish' so that interfaces are intuitive, interactions are complete, and that what most people try to do 'just works'." ([1])

  • Netflix: "We think the combined effect of personalization and recommendations save us more than $1B per year" ([1] [2] [3])

  • "The main reasons cited for using ad blockers include avoiding disruptive ads (69%), ads that slow down their browsing experience (58%) and security / malware risks (56%). Privacy wasn’t the top answer. So Facebook thinks if its can make its ads non-interruptive, fast, [useful,] and secure, people won’t mind." ([1] [2])

  • According to the NYT, Uber lost $1.2B on $2.1B in revenue in H1 2016 ([1] [2])

  • "Amazon reaches new high of 268,900 employees — skyrocketing 47% in just one year" ([1])

  • Amazon's going hard for Netflix on their key vulnerability, strength of the catalog ([1])

  • Great example of how Bezos sees failure as just a step toward success, following up on their $170M loss from an expensive Amazon Fire Phone with another (and I think very promising) attempt using existing cheap phones ([1] [2])

  • Talks from ScaledML 2016, including Jeff Dean, Qi Lu, Ilya Sutskever, and more ([1] [2])

  • Great paper on the data pipelines at Facebook and some of their design tradeoffs ([1])

  • Good article on Facebook's approach to research, not separate from engineering, not part of engineering, but just open ([1] [2])

  • Great article in ACM Queue on Amazon's microservices, which allows for "permissionless innovation" and has many benefits for testing, deployment, debugging, and reliability ([1] [2])

  • Nice example of fine-grained control of data center power and cooling using machine learning to save electricity ([1])

  • Precision agriculture using GPS, self-driving tractors, and crop and nutrient sensors ([1])

  • Pew Internet study of Amazon Mechanical Turk (MTurk), lots of remarkable details, including that most workers are making less than $5/hour, almost all less than $8/hour ([1])

  • "The line between outright deception and poor user design is often hard to distinguish" ([1])

  • "[The] many confusing design decisions made us wonder if projects were assembled entirely from poor stackoverflow posts" ([1] [2])

  • Amusing story of what happens when a geolocation is missing ([1])

  • On education: "A feeling of hopefulness actually leads us to try harder and persist longer -- but only if it is paired with practical plans for achieving our goals, and specific, concrete actions we’ll take when and if (usually when) our original plans don’t work out as expected." ([1])

  • On management: "We have to give them the space to fail in the short term so they can succeed and grow in the long term ... There is that magical moment when we delegate and allow an emerging leader to grow into their new responsibilities, and they end up being way better at it than we ever were. That’s real management success." ([1] [2])

  • On teams: "The best teams respect one another’s emotions and are mindful that all members should contribute to the conversation equally ... A shared belief that it is safe to take risks and share a range of ideas without the fear of being humiliated." ([1] [2])

  • Comic on being data-driven and how it sometimes feels ([1])

  • Xkcd on self-driving cars: "This car has 240% of a horse's decision-making ability" ([1])

  • Xkcd: "Is this a normal bug?" ([1])

  • Xkcd on code quality ([1])

  • SMBC comic on statisticians ([1])

  • SMBC comic on economists and the golden goose, don't miss the mouseover text: "A physicist would figure out how the Goose was transmuting elements without getting to a high temperature, then use the trillions of dollars to build a really sweet fleet of quadcopters" ([1])

  • SMBC comic that perfectly captures why I love talking with geeks, it's the infectious enthusiasm ([1])

Thursday, June 02, 2016

Quick links

A tightly curated list of what I enjoyed in the news recently:
  • Bezos: "Every single important thing we’ve done has taken a lot of risk, risk-taking, perseverance, guts, and some have worked out. Most of them have not." ([1])

  • Bezos: "You need to select people who tend to be dissatisfied ... As they go about their daily experiences, they notice that little things are broken in the world and they want to fix them. Inventors have a divine discontent." ([1])

  • Page: "Is it going to affect everyone in the world? Very few ... think this way." ([1])

  • "More than anything else, the rise of the bots signals the death of the mobile app ... The whole app thing didn't really work out." ([1] [2])

  • "As it turns out, the mundanity of our regular lives is the most captivating thing we could share with one another" ([1])

  • "This is the most demonically clever computer security attack I've seen in years ... insert a nearly undetectable backdoor into the chips themselves" ([1])

  • "Most Android vulnerabilities don't get patched. It's not Google's fault. It releases the patches, but the phone carriers don't push them down to their smartphone users ... This is a long-existing market failure." ([1])

  • "It’s not like iPhones have somehow gotten worse. Other phones, though? They’ve gotten a whole lot better. And they’re cheap." ([1])

  • "Google, with its tech chops and its control over digital ad delivery, is positioned to do what individual publishers and their associations can’t do on their own, though, by requiring that ads are not obtrusive or annoying — a main reason people choose to block ads." ([1])

  • "How quickly cars can learn to do the really hard parts of driving ... navigate congested cities in the pouring rain where humans, pets and rodents run into the road" ([1] [2] [3])

  • "With so many advances in machine learning recently, it’s not unreasonable to ask: why aren’t my recommendations perfect by now?" ([1])

  • "Developers’ speed mattered ... only to the extent that we made effective product design choices ... It didn’t matter how fast they were moving if they were moving in the wrong direction." ([1])

  • "Building and growing startups may appear glamorous from the outside ... It is anything but that from the inside." ([1])

  • "% of pitches for bots and/or AI companies approaching 100%" ([1])

  • "Tech firms are plundering departments of robotics and machine learning ... for the highest-flying faculty and students, luring them with big salaries ... The field was largely ignored and underfunded during the 'AI winter' of the 1980s and 1990s, when fashionable approaches to AI failed to match their early promise." ([1])

  • The FizzBuzz Tensorflow interview "will probably only make sense to people who have gone through really terrible CS interview processes" ([1] [2])

  • Remarkable, deep networks trained on artistic style, then used to apply those styles to video ([1])

  • A good summary of the state-of-the-art in deep learning ([1])

  • "There are limits to the predictive abilities of even tremendously superior intelligence (due to partial observability, chaotic behavior, or sheer randomness)" ([1] [2])

  • SMBC comic: "Once you realize there is no hope, you can relax and just enjoy the progress in machine learning." ([1])

  • My favorite old T-shirt from, Earth's Biggest Bookstore ([1])

Saturday, May 14, 2016

Code Monster from Crunchzilla is now open source

Code Monster from Crunchzilla is now open source, free to use and modify.

Code Monster is a tutorial that has been used by hundreds of thousands of children around the world to learn a little about programming. It's a series of short lessons where each lesson involves reading and modifying a small amount of code. Changes to the code show up instantly, students learning by example and by doing.

The lessons content for Code Monster from Crunchzilla is in a JSON file that can be modified fairly easily to create your own content. By open sourcing Code Monster from Crunchzilla, I hope three things might happen:
  1. Translations. Taking the current content and translating into languages other than English for use in more classrooms around the world.

  2. New lessons and new content. By adding new messages and example code to the JSON lessons file, new tutorials could be created for teaching programming games, working through puzzles or math problems, or perhaps a more traditional computer science curriculum aligned with a particular lesson plan.

  3. Entirely new tutorials. Some ideas and techniques used by Code Monster, such as how Code Monster provides informative error messages, how it does live code, or how it avoids infinite loops in students' code, might be useful for others creating web-based coding environments.
Code Monster from Crunchzilla has been used in computer labs and classrooms around the world. One of the most common requests is translations into languages other than English. Now that the code is open source, I hope that makes it easier for translated and modified versions to get in front of even more children.

If you use the code for anything that helps children learn computer programming, I'd love to hear about it (please post a comment here or e-mail me at

Saturday, April 02, 2016

Quick links

What has caught my attention lately:
  • "We simply don't know how to securely engineer anything but the simplest of systems" ([1])

  • Impressive at their scale: "Facebook ... releases software ... three times a day" and makes configuration changes "thousands of times a day... every single engineer can make live configuration changes." ([1]) 

  • Pew Research report on global internet and smartphone usage ([1])

  • Cute idea for telepresence: "We propose projecting [2D] virtual copies of people directly onto (potentially irregular) surfaces in the physical environment" ([1])

  • For those of us tracking virtual reality, a detailed review of the Oculus Rift ([1]), a review of Hololens ([2]), and a fun TED talk motivating augmented and virtual reality ([3])

  • For disk to be the new tape "custom disk designs uniquely targeting cold storage" are required that are "much larger, slower, more power efficient and less expensive." ([1]) Related, Google seeks new disk designs ([2])

  • Lessons from building AWS, including automate everything and favor primitives over frameworks ([1])

  • In the AWS service terms: "However, this restriction will not apply ... [when] human corpses to reanimate and seek to consume living human flesh, blood, brain or nerve tissue." ([1])

  • Google says, "With multi-homing ... failover, recovery, and dealing with inconsistency ... are solved by the infrastructure, so the application developer gets high availability and consistency for free and can focus instead on building their application" ([1] [2])

  • Remarkably successful contest: "The winning team exceeded the power density goal for the competition by a factor of 3 ... Some of us at Google didn’t think such audacious goals could be achieved." ([1])

  • "Welcome to the Internet of Things... and its tradeoffs" ([1] [2] [3])

  • Netflix's catalog has dropped to 5,532 titles from 8,103 titles in about two years ([1] [2])

  • "The James Webb Space Telescope will be a major advance ... primary mirror will be 50 times [larger] ... eight times the resolution" ([1])

  • "The price of planetary insurance, it turns out, isn’t all that high." ([1] [2])

  • Teaching math: "In most people’s everyday lives ... what [people] do need is to be comfortable reading graphs and charts and adept at calculating simple figures in their heads ... Decimals and ratios are now as crucial as nouns and verbs." ([1])

  • He's the "‘seagull of science.’ He used to fly in, squawk, crap over everything, and fly away." ([1])

  • Good answer to the question, "What are the most important things for building an effective engineering team?" ([1]) Related, similar advice from Amit Singh ([2] [3])

  • An old office map from early 1997 (back when Amazon only sold books, "Earth's Biggest Bookstore"). My "office" was a card table in a kitchen. ([1])

  • What If comic: What would happen if you tried to squeeze all the water going over Niagara Falls into a straw? It's worse than you'd think. ([1])

  • Xkcd comic on bots: ""Python flag: Enable three laws" ([1])

  • Good Xkcd comic on Celsius or Fahrenheit ([1])

  • SMBC comic: "Philosophy tip: Make any sentence profound by adding 'true' to it" ([1])

  • Dilbert comic: "No need for conversation. I know everything about you." ([1])

  • Comic with a Calvin and Hobbes crossover into Bloom County, brings back memories ([1])

Saturday, March 05, 2016

Virtual reality hitting the mainstream: The next $100 bet

Virtual reality is hot again, with dedicated hardware headsets launching from multiple manufacturers intended for general use.

The world is substantially different than the last time this happened. In particular, there's more computing power available in our smartphones than the most powerful graphics workstations had back in the 1990s. Google Cardboard and others take advantage of that, using a smartphone and little else for a quick-and-dirty virtual reality experience.

But, for a product to appeal to a broad market -- to get beyond early adopters with disposable income seeking to show something cool to friends a couple times -- it needs to survive the harsh judgement of busy people. It isn't enough for virtual reality on expensive dedicated hardware to mostly work. The experience will have to wow repeatedly at a price people like.

So, Daniel and I have another bet: "Virtual reality hardware (not counting cardboard) will not sell more than 10M units/year worldwide before March 2019." I'm saying it won't. Daniel says it will. Loser donates $100 to the winner's choice of charity.

Daniel already posted his side of the bet. In brief, he thinks three years will be enough time for someone to get it right.

I think that mainstream adoption of dedicated hardware for virtual reality requires breakthroughs in usability and price that are too difficult to achieve in the three year time frame. The experience just isn't good enough yet for it to be anything other than a toy for early adopters. Current virtual reality hardware is bulky, expensive, not fully immersive, and not addictive or compelling beyond the initial wow. I expect even the next generation will just be a niche market (low million units per year) until we see major developments on price, form factor, and quality of the experience.

There are several wild cards here. For example, it is possible that much cheaper units can be made to work. It's possible that someone discovers very carefully chosen environments and software tricks fool the brain into fully accepting the virtual reality, especially for gaming, increasing the appeal and making it a must-have experience for a lot of people. As unsavory as it is, pornography is often a wild card with new technology, potentially driving adoption in ways that can determine winners and losers. A breakthrough in display (such as retinal displays) might allow virtual reality hardware that is much cheaper and lighter. Business use is another unknown where virtual reality could provide a large cost savings over physical presence. I do think there are many ways in which I could lose this bet.

Like Daniel, I'll add some constraints to make my side of the bet even harder. I'd be surprised if dedicated virtual reality hardware sells more than 10M total over all three years. I'd also be surprised if virtual reality using smartphones (like Google Cardboard) goes beyond a toy, so, is used regularly by tens of millions for gaming, education, or virtual tourism.

And, like Daniel, I expect virtual reality to be big eventually, am frustrated by our current computing limitations, and think we should work to have much better from our computing devices today.

Saturday, February 27, 2016

Tablets replacing PCs: Resolving the $100 bet

In 2012, Professor Daniel Lemire and I bet $100 over the question of whether tablets would replace PCs.

Specifically, the bet was, "In some quarter of 2015, the unit sales of tablets will be at least twice the unit sales of traditional PCs, in the USA." Loser donates $100 USD to the charity of the winner's choice.

It's 2016, and tablet sales went far higher than I ever expected, approaching PC sales, roughly 60M/year units for both tablets and PCs in the US. But tablet sales seem to have peaked, with Q4 2015 unit sales worldwide actually 14% lower than the previous year, which is worse than the 8% decline in PC sales.

There are other surprises. One of my concerns was that a very cheap tablet would dominate the market, and Amazon did come out with a $50 tablet that got relatively good reviews and nearly tripled Amazon's market share on tablets. There hasn't been enough time yet to see what happens with very cheap tablets, but tablets this cheap are a different category than the tablets that were around in 2012.

Another concern at the time was hybrid tablets, so tablets with detachable keyboards that function a lot like laptops, and whether they'd blur the line between PC and tablet. Hybrid tablets have done very well -- a major category in tablets -- and look likely to continue to grow over time.

The last concern at the time was whether tablets could thrive despite the pressure from increasingly larger and more powerful mobile phones. That seems to have been the biggest issue. Phablets are getting as large as early tablets, and tablets that try to be much bigger than a smartphone proved too unwieldy and sold poorly. After all, who needs a tablet when you've got a mobile that's almost as large?

The broader question in the bet was whether people would stop using PCs. PC sales have been in decline, though the pace of that decline has slowed recently. What seems to be happening is that people are continuing to use multiple devices, which was a visible trend back in 2012.

A phone is great when you want to do something quickly on the run. A bigger screen is good when you need to do a lot of reading. A keyboard, mouse, and large screen become useful when you're producing instead of consuming. If you need to do all of these, there's no reason to only have a phone, only a tablet, or only a PC. Instead, people often have all three and more.

Even though I technically won this bet, I want to congratulate Daniel Lemire on this getting much closer than I ever expected. I also admire the bravery he had to take the bet, especially with such favorable terms, and appreciate what I learned from this. The terms were that the loser donate $100 to the charity of the winner's choice, and I'd like to match the donation. Daniel and I will both be donating $100 to the Wikimedia Foundation, which runs Wikipedia.

Update: Daniel's post is up: "Lost my bet: the PC isn't dead... yet".

Sunday, January 17, 2016

Quick links

What caught my attention recently:
  • "Big ideas emerge from spills, crashes, failed experiments and blind stabs .... As people dredge the unknown, they are engaging in a highly creative act .... the habits that transform a mistake into a breakthrough" ([1])

  • Lots of details on recommendations, personalization, and experimentation at Netflix in a new ACM paper ([1])

  • Fun and interesting Slate article on how Facebook selects posts for the news feed ([1])

  • New paper claims the filter bubble for news is much stronger in what people self-select and on social media than in search and recommendations ([1])

  • "Bayesian program learning (BPL) framework, capable of learning a large class of visual concepts from just a single example and generalizing in ways that are mostly indistinguishable from people" ([1] [2] [3])

  • NIPS 2015 paper on problems that accumulate in machine learning systems, such as dependencies between features, dependencies between models that build off each other, and complicated and fragile data preprocessing ([1])

  • "Should they teach [self-driving] cars how to commit infractions from time to time to stay out of trouble?" ([1])

  • Wal-mart is doing poorly against Amazon, which is surprising, I think ([1])

  • Good article on product management. I particularly like the points that most products fail (so you should expect to experiment, adapt, and iterate) and that a good product is about experiences not features ([1])

  • "People keep mentioning how different things are to the period just before the AI winter" ([1])

  • "Smartwatches still have a long way to go in terms of proving their usefulness, necessity, and style" ([1])

  • "CYA security: given the choice between overreacting to a threat and wasting everyone's time, and underreacting and potentially losing your job, it's easy to overreact." ([1])

  • A new $7M XPrize for autonomous undersea drones ([1] [2])

  • Simulating the World in Emoji is a very fun educational simulation, similar to the Artificial Life work a while back, great for kids ([1])

  • From the Exploratorium Museum, a demo of how wave motion arises from swirling smaller movements in water ([1])

  • Dilbert comic on tech jargon ([1])

  • Pearls Before Swine comic on clickthrough agreements ([1])

  • SMBC comic: "Update has been a test of your loyalty." ([1])

Saturday, January 02, 2016

SwipeLingo and Javascript Notebook

I've been working on a couple educational projects since Google, SwipeLingo and Javascript Notebook. SwipeLingo is a quick matching game for touchscreens. Javascript Notebook is a tool for writing coding tutorials, exercises, and examples.

I'm unable to fully finish them and get them exactly where I wanted them before starting at Microsoft. But I'm launching anyway in case they or the ideas in them are useful to others.

SwipeLingo is a game-with-a-purpose, a quick matching game that is both fun and helps with memorization like flash cards do. There are example games — particularly interesting is Chinese numbers, where you learn the characters pretty quickly after starting with wild guessing — and it's also easy to create your own. I was motivated to create SwipeLingo by loving Duolingo but wanting the vocabulary memorization in it to be more fun, and also wanting to try to build a non-native touch web app game that works equally well across desktop, laptop, tablet, and phone.

Javascript Notebook tries to make it easy to write and share coding tutorials, coursework, examples, exercises, and experiments. It was heavily motivated by Stanford's CS101 class and their content. Here are some examples: "Getting Started", "Introduction to Programming", "What You Can Do". It's a bit like a simple Javascript-only IPython Notebook in feel, but runs entirely in the browser, requiring no configuration or set up, just write and share. Others can modify the code, run it, and save and share their own copies.

Please let me know if take a look and have any comments or suggestions. And please tell others who might be interested about them too!

Update: About three years later, I shut both of these down. There are no longer available.