Four articles for the month of May 2013. One of them however is about a story in the realm of mathematics so arguably isn’t much of a science article at all.
Older computer users will probably remember the ELIZA chat bot written in the 1960s. It was created only as a early demonstration of natural language processing but many people took it seriously as a virtual psychotherapist. This article from the BBC talks about a real attempt at creating a virtual therapist to help real humans. As such it goes much farther than just printing text output onto a screen. It has an onscreen avatar which it can control, can verbalize its responses, can listen to patients’ voices and observer their body language and so forth.
So many claims of success at achieving cold fusion have been refuted over the years that many people now think it is impossible. This article from ExtremeTech covers the latest such claim and given the secrecy involved, it seems likely that it is just another scam. In this case however, a number of scientists from reputable European universities have been allowed to study the device, though they are still being kept in the dark about how exactly it works, and their preliminary, non-peer reviewed, finding is that it works as advertised. Given the potency of cold fusion as a power source, which would allow it to completely supplant our currently fossil-fuel based energy economy, this is something that deserves a lot of attention.
The next article from the so-called smart rifle. It comes with a color graphics display that allows the user to lock on to a target. The rifle then uses its own suite of sensors to determine when exactly to open fire, taking into account factors such as wind and distance, to ensure a hit. It even comes with Wi-Fi so the data for every shot can be shared online.
Finally here’s an extensive article describing the excitement in the mathematical community surrounding the release of a series of papers by a Japanese mathematician in August 2012. The mathematician in question Shinichi Mochizuki posted the papers onto the Internet claiming that it was a proof of the ABC Conjecture, a number theory problem that has important ramifications for mathematicians, and for all intents and purposes simply walked away, refusing all media interviews and requests to field questions. Other mathematicians of course delved eagerly into the papers, but there are 512 pages in total, and those were filled with new mathematical concepts and constructs that Mochizuki had seemingly invented himself. This meant that no other mathematician has so far been able to verify the correctness of his proof and so many months later the entire community is still waiting with bated breath to see if the proof is correct.
Part 2 of Princeton’s Algorithms course on the Coursera platform officially ended last week but I haven’t had time to write about it until today. I gushed about the first part of the course in a post last year and this is just the follow-up. It was originally meant to be offered at the end of 2012 (and the lecture videos have dates indicating that they were indeed filmed last year) but due to apparent problems with finalizing the exercises and assignments, the course was only opened at the end of March of this year. My thoughts:
I thought that Algorithms 1 was the best educational experience I’ve ever had and part 2 only goes on to confirm it. It is amazing in all respects. Based on discussions on the online forums with other students, it is my understanding that no other course available on the Coursera platform currently can match its wealth of teaching materials. It has lecture videos, PDF copies of the slides used in the lectures, extensive exercises, very interesting programming assignments, the code on the booksite provides a vast amount of examples to look at and its fiendish difficult interview questions will challenge you for weeks after the end of the course.
As to be expected part 2 is markedly more difficult than part 1. I’m pretty sure that the graph structures we learn in the first weeks of the course are more complex than anything else seen in part 1. I note that the official forums were pretty much devoid of dumb posts asking for help with elementary topics quite early on. There were some posts talking about the difficulty of the material early on and then everything was down to business. I surmise that the initial difficulty curve scared off plenty of people.
Part 2 only has four programming assignments whereas part 1 had five but they are all much trickier. Many students had high praises for the week 2 assignment which involved implementing the seam carving algorithm invented in 2007 and now used in many graphics editing software. It’s not just a toy program either since our own implementations can be used as elementary clients to perform seam carving on our own photos. I also liked the WordNet project which uses graphs to capture the relationships between words in the English language and is apparently used as part of the core of many programs that need to understand English, including IBM’s Watson computer. The other two projects were implementing the Burrows-Wheeler data compression algorithm and using graphs in an unconventional way to figure out whether or not teams have been effectively eliminated in baseball (or any other sports) leagues.
The latter stages of the course become increasingly theoretical. For example the professors provide an implementation of a linear programming solver but admit that it is only a toy solver and stress that serious users should look towards industrial-strength versions. This contrasts with previous work in the course in which the textbook implementations we study are suited for general use. The course ends with a discussion on reductions and a look at the classic P vs. NP problem in computer science.
The exercises used share the same format as part 1 of the course. The main purpose seems to determine if a student can manually trace how an algorithm works on pencil and paper. But the final exam in part 2 is a surprise in that it has a strong focus on the theoretical. Several students posted in the forums to note that the final exam feels like it wouldn’t be out of place in Tim Roughgarden’s Design and Analysis of Algorithms course. This is great of course since theoretical questions like this really test whether or not you understand the algorithms in question in a deep and intuitive way but tests like this are much harder for me.
While I could complete all of the assignments and exercises with a 100% score and I scored over 90% in the final exam, I feel that the difficulty of this course comes close to hitting my limits of what I can do. For example, for some algorithms, such as the fiendish Kosaraju-Sharir, while I can certainly follow the implementation and trace its results, I find that I can’t quite wrap my head around why it works and hence a deep, intuitive understanding of it. Similarly, for the data compression assignment, I can implement it and get it to work perfectly and why it works the way it does feels like magic to me.
So yeah, I greatly enjoyed the time and effort I spent on this course. My next Coursera course won’t be starting until mid-June or so which means I have some free time on hand until then.
Late this month due to an extended stay in Kuala Lumpur for the Malaysian general elections. Here are the three articles I’ve managed to glean from around the web in April.
This article from the Pacific Standard magazine covers a paper whose authors examined the obituaries of over 1,000 famous people published in The New York Times to determine if there are any patterns in them. They found that the famous people who died earliest were athletes, performers and non-performers who worked in creative fields. The famous people who died later were politicians, businessmen and military officers. The tentative conclusion is that people who work in sports and the performing arts incur psychological and physical costs that curtail their lifespan.
Here’s a link to a paper claiming that vervet monkeys were able to solve a multiplayer coordination “game” in which a captive monkey was trained to open a container holding a large amount of food, but only if the dominant monkeys of a wild troop stayed outside of an imaginary circle away from the food. The wild monkeys were able to infer the correct behavior by observing the trained monkey and receiving feedback from the trained monkey without the intervention of humans.
The Economist has an article talking about the tells that give players away when playing say a game of poker. Most people instinctively believe that the key to not giving away information about the hand you’re holding to other players is in keeping a straight face. As it turns out, experiments show that observers achieve a much higher success rate at correctly predicting the quality of a hand of cards held by another person not by looking at the player’s face but by looking at the player’s hands. This is sure to be a result that will revolutionize poker playing strategies.
One of the MOOC courses I’m taking on Coursera, Learn to Program: Crafting Quality Code, has effectively ended. It was taught by professors Jennifer Campbell and Paul Gries of the University of Toronto. It is by a large margin, the most disappointing of the online courses I’ve taken so far. What follows is a long post that I wrote on the official forums explaining why I was disappointed with the quality of the course:
This post will probably be unpopular. Judging from the tone of the posts in this forum, this course seems to be well-liked by many students. Yet I note that for many of these posters the first Learn to Program, which I did not take, appears to have been their very first programming course. I disagree with that assessment and have a low opinion of the quality and usefulness of this course. I started this thread to set out my reasons for holding this opinion.
I’ve been busy with programming stuff since Algorithms II just started up. The first assignment involved processing WordNet graphs. The second assignment involved implementing seam carving, also known as content-aware resizing. Anyway that’s why I’ve been browsing less lately and so have fewer articles. Here we go:
IBM’s Watson supercomputer made the news in 2011 when it won a special Jeopardy! tournament against human champions. This article covers some of the first commercial applications it is being used for, helping doctors to diagnose diseases in various hospitals in the United States. It also talks about how its size has since shrunk from that of a bedroom to that of a bathroom and how it could eventually be a handheld device. Cool note: Watson uses Princeton’s WordNet to help it parse and understand the English language.
In other computer news, the next big thing in computing is supposed to be quantum computers, and it has been for a while now but actual implementations have proved as elusive as nuclear fusion. This article talks about just such an implementation. It will be used by Lockheed Martin to “create and test complex radar, space and aircraft systems“, i.e. make weapons and works at temperatures close to absolute zero.
The next article is about Russian scientists discovering completely unknown forms of bacterial life deep under the Antarctic ice. The samples come from the underground Lake Vostok, a body of water that lies 3,700 meters under the ice and is thought to have been isolated from the rest of the planet for millions of years. Needless to say the Internet is waiting to see if they have awakened Cthulhu or dug up The Thing.
Finally we have an article about research into whether or not smiling before matches in the Ultimate Fighting Championships affects the chances of the martial artists’ success. Pre-match photographs of the two combatants were analyzed for the presence and intensity of smiles and matched with the results of each fight. The researches found that as expected, fighters who smile and smile more intensely, lose more often. There are various plausible explanations though none are proven. It could be that smiles are an involuntary sign of submission, or that smiling fighters simply aren’t as aggressive.
So the Programming Languages course is just about over. There’s actually one peer assessment yet to go and it will probably be a few weeks until the course staff tabulates and releases the results, but all of the serious work is over. I wanted to get this post written before I start on two new courses next week. It’s run by Dan Grossman of the University of Washington and ran for 10 weeks. This course used a slightly different schedule than other Coursera courses which usually had weekly sections. For this course, the eight course sections were spread out over the ten weeks so we didn’t have a predictable schedule of when each section would be released.
The course used three different languages, starting with SML for the first four sections, Racket for the next two sections and ending with Ruby for the final two sections. There was a programming assignment for each section, with exception of section 4 when we had a mid-term exam. At the end of it all, we have the final exam which is still ongoing as of the time of writing. There were no other quizzes though each section’s programming assignment had to be submitted twice. Once to the autograding system to test for correctness and once again to the peer assessment system to check for style issues and correctness isssues for which it was difficult to get the autograder to test for. One cool thing is that most of the programming assignments have optional challenge questions which award a small number of bonus points for a comparatively huge amount of extra work, just for those who want to run the extra mile.
Due to the Chinese New Year festivities in February, I had less time to browse around for science articles so only three of them this month:
This article in The New York Times covers some very preliminary work on how brain signals can be transferred from one laboratory rat to stimulate another rat to perform an action intended by the original rat. It’s as if the original rat were remote-controlling the other rat, in this case made even more impressive by the fact that the signals were encoded and transmitted over the Internet from one rat to another. As the article goes on to note it is very simplistic and the responses were correct only slightly more often than random chance, but it’s still a step in an intriguing direction.
The next one is from Wired which discusses how dolphins may have personal names of their own, called signature whistles, and may address each other by these names. This suggests that dolphins are able to learn specific signals, as opposed to intuitive ones, and use them to communicate, all without the intervention or guidance of human handlers.
Finally this last one from The Atlantic comes with a video must be seen to be believed. It pulls the wraps off DARPA’s 1.8 Gigapixel video camera that can cover pretty much the entire area of a medium-sized city with enough resolution to spot a person waving their arms on the ground. That’s some serious Big Brother surveillance capability there.