A large part of the talk was used to explain the differences between a CPU and a GPU. In both cases, the architecture of the silicon is crafted in such a way to allow for processing of what's known as "atomic" statements. As the name might imply, atomic statements are instructions that the CPU is able to process naturally, without any need for breaking the instruction down into more basic instructions. Atomic statements are therefore "basic" instructions.
A general purpose CPU like the Intel Core i7 is crafted so that it can, with a little extra work, compute just about any problem that a piece of software or hardware requires. A GPU on the other hand, like the NVidia GTX 295, is crafted so that it can quickly work on mathematical problems like matrix math. This is due to the fact that the graphics being displayed on a monitor, whether it be the Windows desktop or a rendered scene from a video game, is the result of numerous mathematical calculations on geometric objects. So instead of the problem being broken down into basic parts on the Core i7 (which adds considerable computational overhead), the problem can be immediately processed as is on the GTX 295.
For parallel computing tasks, such a math-oriented environment is ideal. Many parallel computing problems require the hardware to spend enormous amounts of time crunching floating-point numbers. Adding to the computational power is the amount of cores on a modern GPU; the current NVidia Tesla GPU contains 240 cores dedicated to processing. Naturally, the Tesla GPU outperforms the Core i7 almost threefold in computationally intensive tasks.
Of course, the disadvantage to the GPU implementation is its utterly abysmal performance on crunching general computational. The CPU is designed for a general case in mind, after all. An additional note that wasn't touched upon in the seminar is the misconception regarding a GPU's performance compared to a CPU. It seems that the amazing threefold advantage can only be achieved by problems termed "embarrassingly" parallel, or problems that are (laughably) easy to break down into parallel computing components.
I guess number-crunching fits that bill quite well.
In an era where a sizable portion of the world's literature is stored in some digital form or medium, the relative shelf life of such mediums leave a lot left to be desired. The irony is that while efforts in "archiving" the printed medium of centuries long past were intended to preserve such works, the printed medium will more than likely outlast the digital collection it has been copied into. This is due to theoretical limitations in utilizing semi-conductors for such tasks.
Researchers have discovered an alternative to the traditional semi-conductor approach by using some modern techniques in the field of nanotechnology. Digital information is usually stored in the medium as a machine-readable set of 1's and 0's. By storing an iron nanoparticle inside a carbon nanotube in one of two positional states, one can induce the iron to move between the states in the presence of electricity. This can effectively represent the machine-readable 1's and 0's required by the system.
Greater storage space can be achieved by packing components of the digital medium into dense clusters. This relationship is directly proportional: the greater the density of the medium given a physical space, the greater the storage offered. Unfortunately, semi-conductor placement also have an inversely proportional relationship to its shelf-life: the greater the density of the medium, the less the shelf-life will be. The nanotube system is believed to be relatively stable in this regard, as nanotubes can be packed as densely as needed while yielding the same shelf-life: over a billion years.
Nanoscale Reversible Mass Transport for Archival Memory [Nano Letters, ACS Publishing]
Electrical engineers at UC San Diego are working on their music identification system. What adds a new twist to the formula is their efforts at taking a general, genre based approach at music selection. For example, when a user searches for "easy listening", the system will attempt to identify every song in its database that it determines to be "easy listening" and return the results as a recommended list to the user. The key point, of course, is to understand that the software (and to some extent, hardware) will be able to determine the genre by analyzing the digital information that represents the music.
Apparently, the system had trouble with Queen's Bohemian Rhapsody. I can't blame it.
The system will need initial parameters in order to determine what types of music fit into what types of genres. The researchers originally paid students to listen to songs and label them manually, but switched to a new model that involved members of the social-networking site Facebook playing a series of games which accomplishes the same objective. Named Herd-It, the game involves users listening to music, identifying instruments, and finally labeling the songs in order to earn points in a high score table. The closer a player's submission matches the normative answers among all players of the game, the higher the score.
The researchers are also saving money. Hiding a research effort in the guise of a game allows them to utilize a human computer farm at the cost of a few hours of programming.
From A Queen Song To A Better Music Search Engine [Science Daily]
The new USB standard, published at the end of last year and expected to first hit markets at the end of 2009 or early 2010, is a significant improvement over the previous and current version, 2.0. USB 3.0, dubbed “Super-Speed” (as opposed to the previous “High-Speed”), has several important features that will allow it to be adapted into the marked. Probably one of the most important features is backwards compatibility: All current 2.0 devices are usable with the new system. This is accomplished by making two sub systems: one is the 3.0, and one is the 2.0. The two systems are pin compatible, so they share the same USB port. Any 2.0 will be usable with 3.0, and vice versa.
Cyber-physical systems (CPS) encompasses all instances of interaction between a physical object and some computational portion of that object in the modern world. A specific example of a CPS would be the interaction between an airplane and its collision detection system.As with any piece of computer software, a given design with an apparent flaw will ideally be fixed to working order in while still in the design stage. If the flaws proceed into the manufacturing state, then the only way to find those flaws would be to employ trial-and-error. In the case of an airplane collision detection system, such trials would be expensive, time-consuming and ultimately impractical.
A research team at Carnegie Mellon have developed a piece of software that, provided with some initial parameters, will attempt to find a counterexample which illustrates a flaw in a given design. For example, an airplane on a collision course with another airplane will recieve evasion instructions from its collision detection system in order to avoid a universally fatal plane crash. The software, after recieving the parameters of the collision detection system, will attempt to find a scenario where the two planes will crash into each other.
Employing typical "brute-force" methods to this problem would be a computational nightmare, if not outright impossible. This is due to the infinitely many variables present in the real world that can affect the outcome of the system. The method employed in the software attempts to bypass this difficulty by extrapolating "differential invariants", or basic pieces of the problem that never change regardless of the variables. Using the many differential invariants in the collision problem, the software will attempt to piece together a counterexample or prove that the design is sound when no such counterexample can exist.
The latter is obviously the difficult part.
Method For Verifying Safety Of Computer-controlled Devices Developed [Science Daily]
IEEE Spectrum: This magazine is the IEEE general interest magazine. It lacks the significant depths of the more specific journals, but it generally has interesting articles of a tech forum category.
Scientific American: This magazine is a bastion of the science magazines that I read. It’s a little too Pop-Sci to be truly reputable, but it has intermediate level articles that with some dedication can be well to read.
Make Magazine: This magazine (really almost a soft cover book) is a hobbyist technology magazine. It’s almost what Popular Mechanics was back in the day, before it was corrupted by bad articles. Make has many unique and interesting projects that, while I’ll never build most of them, are inspiring to do something of my own. Add in the culture that they nourish, and it’s an enjoyable read. Think Portland.
Continuing along the quantum computing tour path, researchers have been able to harness a glaring shortfall of quantum computing methodology and morph it into something practical. The premise is so simple that you would have to wonder why the research didn't take place much sooner.As with conventional computers, quantum computers are vulnerable to random noise in their quest to process data. The popular design is to use and alter the fundamental unit of quantum mechanics, the atom, to represent data in a meaningful and ultimately useful way. Unfortunately, random noise for a quantum computer can be anything from the heat of the sun to the movement of electrons in the air that the quantum computer is sitting in.
A immediately practical application for such levels of sensitivity is to use as a core for an extremely tiny, atom-sized sensor. Such "quantum sensors" would have the ability to detect natural occurrences several orders of magnitude smaller than what would currently be considered "undetectable". The article mentions, as an example, tiny magnetic waves emanating from the ocean floor that may indicate untapped oil reserves.
The Oxford researchers named their system the "quantum cat", after Schrödinger's thought experiment involving a box, a cat and a vial full of lethal poison. Perhaps the most interesting (and ironic) part of the story is the paradigm shift required in the manufacture of quantum computers.
"Many researchers try to make quantum states that are robust against their environment," said team member Dr Simon Benjamin of Oxford University’s Department of Materials, "but we went the other way and deliberately created the most fragile states possible."
Quantum Cat’s 'Whiskers' Offer Advanced Sensors [Science Daily]
