kmeme

Physical Bits

noreply@blogger.com (Philip) — Sat, 11 Feb 2012 18:38:00 +0000

Suppose you were to sneak into a datacenter run by Google or Facebook or Twitter and take a pickaxe to a random server. What would happen? Firstly, you would be arrested.

But what data would be lost? In all likelihood nothing. You might have rent RAM chips and hard drive platters asunder, but the bits that were previously inside the machine would be safe and sound. How is this possible? Careful and deliberate redundancy. At large internet companies the exact same information lives in several geographically diverse datacenters. Even a catastrophe at one location should not result in lost data.

It’s important to realize that this extreme resiliency and lack of locality is not an inherent property digital information. Quite the contrary. Run of the mill digital information, say the unbacked-up hard drive in your laptop or a DVD of family movies in a shoebox, is highly location specific and incredibly fragile. Trauma or simple degradation can completely and permanently wipe out these islands of isolated data. Consider yourself warned.

Digital ones and zeros are an abstraction. The abstraction is that they are pure and absolute, noiseless, lossless, and permanent. The reality is that bits are physical and fragile, subject to noise, degradation and loss. Real bits are stored in “everyday” physical constructs like electric charges, electromagnetic waves, or magnetically charged particles. All of these are subject to corruption and noise and will degrade over time. Only by the overwhelming application of infrastructure do we approach the abstraction, do we make bits seem omnipresent and indestructible.

Both views of bits are worthwhile. At times think of them as weightless and otherworldly, out there in the ether with no real physical location. Other times keep in mind that every copy of your digital information has some real-world physical representation.

Consider composing an email. Deep inside the computer a unique physical structure is coming into existence. Key press by key press you mold and shape this tiny little physical thing. And when you hit send? A physical representation of your object is transmitted vast distances through wires and over the airwaves, culminating in the recreation of a perfect copy of that same physical structure in some remote machine.

Not dramatic enough? Consider Twitter. Lady Gaga has close to twenty million followers. When she, or her publicist, hits send on a tweet the little physical gem of wisdom is replicated in literally millions of phones and computers around the world. Picture a stadium hundreds of times larger than the Super Bowl filled with people. Simultaneously this identical physical little nugget is constructed in all of their phones, tablets and computers. The power of that one little send button to effect physical changes in the world is amazing.

This physical nature of data relates to virtual worlds as well. In computer games and graphics we’ve gone from Pong to Modern Warfare 3 in a few decades. Soon computer simulations will rival reality in graphical detail. What would it mean to interact with a virtual world whose complexity and nuance compares to the real world?

The virtual world like all digital information is really manifest as a real-world physical structure. The structure exists in the memories of real computers. The structure came into being by the physical interaction of real people. A programmer typing software code or a digital artist modeling objects, both are physically changing structure of the real world.

The brain of course is also a physical system. Your warm and fuzzy memory of an early Christmas is represented somewhere in the physical connections of neurons in your brain. The thought you had a second ago about what to have for dinner is represented by firing patterns of neurons, firing patterns which are entirely physical involving voltage-gated ion channels and the diffusion of molecules across small physical gaps.

So our future human sits down to interact with a virtual world. What’s really going on? On the one side we have this messy physical brain and on the other side we have this complex physical system, and the two are interacting. What is virtual about that?

There isn't separate plane of existence where digital information lives. When you write software or email or facebook posts you are modifying the actual physical world we all live in. So carry on changing the world.

Firework

noreply@blogger.com (Philip) — Wed, 13 Jul 2011 02:38:00 +0000

I will kick things up a notch on this blog with a salacious admission from my personal life: I love the Katy Perry song Firework.

Or I should say I used to love it. From the moment I first heard the song I would crank it up on the radio. Eventually I half-mastered the lyrics and would sing what I knew at the top of my lungs. I got my 6 year old daughter hooked on the YouTube video, really so I could hear the song some more.

We even bought the album, an actual physical CD. The Parental Advisory sticker wasn’t lying so we couldn’t play most of it for my daughter, and I never really got into it either. But the Firework song was golden for a while.

Do you ever feel like a plastic bag / Drifting through the wind / Wanting to start again.

Then it all stopped. Well there was this cooling down period. Then finally song lost all its power over me. It become kind of annoying really. The actual texture of the sound seems different now, it seems extremely flat, no sparkle no hook.

Pandora describes Firework as having nu-disco qualities, electronica influences, a subtle use of vocal harmony, extensive vamping, major key tonality, melodic string accompaniment, prominent percussion, and romantic lyrics.

What was going on in my brain when I first heard the song, or when I had it cranked up for the 20th time. How about now when it seems dull and faded?

Sebastian Seung is an MIT professor of Computational Neuroscience. In a TED Talk from 2010 he presents a wonderful metaphor for your brain. The connections between neurons are a stream bed, and the neural activity is the stream. The water can shape the stream bed. Or to put it a less poetic way neurons that fire together wire together.

I normally don't like songs like Firework but when I first heard it something tickled my brain in just the right places that I paid attention. Attention is the most valuable resource we have because everything which impinges on our brain is gated by it.

So Firework was deemed worthy of my attention. Then I immediately began to learn the song, from the beat to the opening notes to the lyrics. Learning manifests itself in physical changes in your brain, in new connections between neurons. These physical changes were be followed by dopamine and other reward chemicals, to encourage more learning.

As I grew to know the song better the experience of listening to it would change, I would know some of the lyrics so I could follow along in my mind, and get rewarded again as I learned additional snippets. Similarly the instrumentation, the breaks, the expected and unexpected rhythms. Eventually though saturation kicks in. One more listening produces no physical changes in my brain, no reward.

Now a true Katy Perry fan would not stop at one song. They would continue learning new songs, new facts about the singer’s life, the band, go to a concert, etc. Firework wouldn’t fade because it would be at the center of a much bigger growing web of information and experience.

For me though, it went dark.

Which reminds me, have you heard the Mumford & Sons song The Cave? Seriously, it's really good. Go learn it.

The Games People Play

noreply@blogger.com (Philip) — Wed, 19 Jan 2011 21:51:00 +0000

One morning my 3½ year old son was still rubbing the sleep out of his eyes when he asked to play a game. He meant a board game, probably something we had been playing the night before like Chutes and Ladders.

I tried to steer him back to the normal morning routine, “Okay how about we play breakfast?”. Instead of objecting he adopted a “hey wait a minute” look and asked quizzically “What is a game?” I dodged the question and threw some waffles into the toaster.

Wikipedia says a game is a “structured activity” undertaken for enjoyment or sometimes education. Rules of Play is great book about games, it takes a chapter to come up with “a system in which players engage in an artificial conflict, defined by rules, that results in a quantifiable outcome”.

With both definitions we see a game is not an artifact in the world. It’s not the Chutes and Ladders board, pieces and spinner. Instead it's the structured activity of playing Chutes and Ladders, it’s the experience you have while playing. The experience of being close the finish, but falling down that long chute and having start all over. One could recreate the essence of the Chutes and Ladders experience with alternate equipment, a home-made board, buttons for the pieces, a die instead of a spinner.

I first started really thinking about games in 2003 when I joined Harmonix. I was not really a gamer nor had I ever worked for a game company, so it was entirely new territory for me. I worked on two games for them, Antigrav and Guitar Hero. I grew curious about what made game development different from “normal” software development.

When a video game company creates a new game they are really doing two things simultaneously. Firstly they are inventing a structured activity, inventing an experience they want the player to have. Secondly they are creating an artifact in the world, a piece of software, which will produce that experience, or as close to it as they can get.

For example consider Tetris, it was created by Alexey Pajitnov and released in the Soviet Union in 1984. Since then it has been re-implemented hundreds of times. Legal issues notwithstanding a teenage programmer can write a solid version of Tetris in weekend. The experience of Tetris was invented once, but the artifacts in the world that produce that experience continue to multiply. If a Tetris clone is faithful enough to the original it will produce essentially the same experience in the player as the 1984 version, even though it was implemented some other way in some other time by some other person.

In the future software development might advance enough that a teenager could create their own versions of current blockbuster games like Call of Duty or Rock Band. Future players of these clones will have an experience nearly identical to today’s players, despite the radically different origin of the artifact.

This bifurcation between experience and implementation roughly corresponds to the division between design and engineering. Designers specify how something should work, what the player should experience, and engineers or programmers code up something which produces the desired effect.

Design comes from the word designate which means to indicate how something should be. Before you can designate you have to make a decision, so the design of a product roughly corresponds to all the decisions that were made during development. If you really like the design of the iPad that means you really like the decisions that were made during its development.

Non-game software generally cares about user experience as well. Recall Windows XP stood for Windows eXPerience! The difference with games is the user’s experience is everything, it’s the end goal. With regular software the end goal is to bring about some change in the world, to create, transform or communicate. The experience is merely one attribute of the software.

It is no small irony that some gameplay highly resembles “productivity” or “communication” software. A World of Warcarft player juggling windows, typing in multiple instant messengers, and defining and using shortcut keys is not completely different from a stockbroker on the job. If you watch what someone is physically doing while playing a game there is nothing obviously fun about it, the experience is entirely in their head. This maps on to the real world as well.

Consider two people picking up litter on the side of the road. One is doing community service and is entirely unhappy about the task. Another is a Cub Scout volunteer who is playing a game with his pack mates: 10 points for every soda can, 100 points for an item of clothing, etc. Each person is physically going through identical motions: reach out and stab some trash, place it in the bag. In their heads however they are having completely different experiences.

Since our Cub Scout is really picking up trash, does this violate the idea that games should only be about experience? Here we have layered two activities on top of each other, so both are going on simultaneously. The Cub Scout is really picking up trash, bringing about a change in the world, but he is simultaneously playing a game. Because games are entirely about experience, the experience can not only be reimplemented in another form, it can be mapped on to another activity.

40 years ago there was no video game industry while today it earns tens of billions of dollars. Recent Game Developers Conferences have had more than 17,000 attendees, which is only a fraction of all game developers out there. Thousands of games are produced every year. We have a critical mass of talent and interest in crafting entertaining user experiences, how can we apply this resource?

Imagine a future Cub Scout with a heads-up display playing “TrashPickr 2025”. At a glance he can see his score, his friend’s scores, and the scores of the pack across town. Some trash items are recognized by the game as powerups, which change the scoring or unlock items he can collect and trade. It’s not the most exciting game in the world, he’d rather be home playing DeathBlow 2030, but it makes a few hours on a Saturday a lot more fun. It motivates him and the exercise he’s getting is real, trash is really getting picked up.

In the future we will spend more and more time in software mediated environments. The physical actions we will be performing: scanning displays and touching screens, clicking, entering text, will be nearly identical whether we are working or playing a game. With luck the line between these will be ever more blurry.

The Simulation Hypothesis

noreply@blogger.com (Philip) — Sat, 01 Jan 2011 01:53:00 +0000

There’s a Philosophy professor at Oxford who maintains an online shrine dedicated to a single question. Nick Bostrom asks “Are You Living In a Computer Simulation?”. The centerpiece is his technical paper from the Philosophical Quarterly (2003) with that question as the title, but then he also has several “popularizations” which explain his argument more simply. He has links to New York Times articles and videos and all manor of commentary about this intriguing question.

Before we get into the question, let’s ask a simpler one. Is everything just a dream? How do you know the people you interact with in your daily life aren’t just characters of your own creation in some ultra-realistic dream?

This is a late-night “whoa dude” kind of philosophical question. It’s worth thinking about, but it doesn’t really go anywhere. If the world is a dream, it’s one where we’d get locked up for running around shouting “life is just a dream!” That is, there’s no ramifications to everything being a perfectly realistic dream, it doesn’t change anything. Plus although it's logically possible there’s just no reason to doubt other people’s existence. Occam’s Razor says pick the simpler explanation, the world is real, people are real, everything in the world is not about you, end of story.

Nick Bostrom’s Simulation Agument is similar but insidiously different from the “is life a dream” question. Bostrom actually gives us a real statistical reason to suspect the world could very well be a simulation.

The argument is as follows. Look at how technology is developing, computers are getting more powerful every year. We use them to simulate the weather or earthquakes or the stock market. In the fullness of time as computers becomes monstrously more powerful, might not there be a day where we can simulate a whole universe? Maybe not in a hundred years, maybe not in a million years, but some day, some ultimate point it time, might that be possible?

Well once we can simulate one universe, it’s only a short extrapolation away to imagine simulating many universes. We might try various initial conditions or physical constants, run experiments. Eventually even home computers could do these simulations, maybe cell phones or screen savers. There would be millions then billions of these simulations going all the time.

In some of these simulated universes life would likely evolve, then intelligent life, then advanced societies with their own technology. And some of these simulated civilizations might eventually invent their own simulation technologies. That is, inside our simulation these little life forms would create another nested simulation.

So we’ve set the stage now we can make the argument. Consider these millions and billions of simulations, and nested within them millions and billions more. What are the odds that our own little universe is one of these simulated worlds as opposed to be the one true root universe which started it all? Doesn’t it seem more likely that we are living in one of the simulations?

Whoa dude. Unlike “is life a dream” here is a real argument for the non-real. You can picture those trillions of simulations. Suggesting we are not one of those would be the height of hubris.

Now there are some ways out. Firstly maybe we won't ever make a simulation that detailed, it just won't happen. We go extinct or never get around to it or the technology isn't good enough. Or maybe we can, but for whatever reason decide to never simulate whole universes. In either case the chain is broken, if we aren't going to create simulated worlds, the it's more plausible that we are the root universe.

But suppose it is true: we are in a simulation. What then? This part is the same as the "is life a dream" question. It doesn't really make that much of a difference. Try running around saying "it's all a simulation" see how that works out for you. Life would go on exactly as before, simulation or not. Human life would have the same value, whether that value is in "reality" or in the "simulation" does not really matter.

The new Tron: Legacy and the old The Thirteenth Floor each show simulated entities finding their way out of the simulation, into the outer world. If this were possible, if the Large Hadron Collider opens up a door out of our universe, then of course it does matter that we are living in a simulation. Or if the creators of the simulation intervened and messed around with things, if we saw "glitches in the matrix". But so long as we are an isolated simulation running on our own it's really the same as a reality, it is our reality.

The only real difference is the simulation story gives us a "creation myth" for the universe. It explains what's outside the universe: another universe running our simulation on some kind of computer. But what about the root universe? There's still no explanation for how it came into existence, or what is outside of it. So we've added a lot of simulations and nesting but the net amount of mystery is unchanged.

So does Occam's Razor again suggest the simpler explanation is to just assume our world is real? I think it does. Reality is simpler than piles of nested simulations. But the nested simulation scenario has a certain elegance. And I for one will keep my eye out for a portal or a glitch in the simulation, just in case.

Why You Should Vote

noreply@blogger.com (Philip) — Thu, 28 Oct 2010 11:28:00 +0000

If you vote in the midterm election on November 2nd it is virtually certain that your vote will not change the outcome. Researchers studied 16,000 Congressional elections over a 100 year period and determined that only once was the winner decided by a single vote, which is 0.006% of the time.

Whether you vote or not the same politician will be smiling at you from the morning paper. So in that sense your vote does not matter. And yet, the winner of the election will in fact be determined by voters. Is this a paradox? Does your vote matter or not?

Elections are an example of emergence, where the behavior of individual entities sum up to produce a result at a larger scale. A physical example of emergence happens when you blow up a balloon. Trillions upon trillions of air molecules bounce around inside the balloon. Each has a seemingly random speed and direction and will indiscriminately smash into other molecules and into the inside surface of the balloon. However the sum of all this chaotic behavior is pressure, a macroscopic phenomenon which smoothly expands the balloon’s surface.

In an election we are like the air molecules. The statistics of our combined behavior determine the outcome, but our own direct contribution has no traceable impact. This type of emergence appears elsewhere in our daily activities. Very often our individual choices don’t have a measurable impact, and yet out our combined contributions do make a big difference.

Suppose you are in a hurry one day and buy a McDonald’s drive-thru meal. As you finish up the last french fry do you immediately chuck the packaging out the car window? Judging by the scene along many highways, some segment of the population does exactly this. I don’t know who these people are, but they are out there.

I myself prefer to “pitch in” and throw my trash away like a civilized person. But why make the extra effort? How much damage would one little hamburger wrapper really do? Given the amount of trash already out there, one extra wrapper makes no measurable difference. And yet the sum of all these little decisions adds up to something significant, since there is a very visible amount of trash.

This decision about whether or not to litter is an ongoing election of sorts, a referendum on how grungy we want the country to be. The midterm election, like most elections, is seemingly about the candidates, Fiengold vs. Johnson or Boxer vs. Fiorina. But first and foremost an election is a referendum about Democracy, the issue at stake is whether we want the system as a whole to continue functioning. And our stance on this issue is derived from whether or not we choose to vote.

Mid-term elections have been averaging around 37% voter turnout. The winner will take the spotlight, but voter turnout is fundamentally the most important metric for any election. If the voter turnout is small relative to the whole population, the election’s outcome will drift arbitrarily far away from what the people actually want. With a small enough turnout any fringe group could rally a few voters and install a candidate that essentially no one wanted. With too few voters the whole system becomes unstable and subject to manipulation and corruption.

Absolute numbers are really important to emergent processes, you need enough participants to assure stable and meaningful results. The fact that your vote essentially never picks the winner seems like a drawback at first, but it is absolutely crucial to the whole process. No individual and no small group should be able to manipulate the results. Your vote won’t decide the winner, but by casting a vote, for anyone, you will directly contribute to making the election a viable and legitimate process by increasing turnout.

So please vote, and pretty please don’t toss that bag out the window.

The Future of Google's Street View

noreply@blogger.com (Philip) — Wed, 06 Oct 2010 02:13:00 +0000

The Need for Speed racing game franchise has been around for 16 years, sold over 100 million copies and generated more than $2.7B in revenue. People love to drive fast and they love to race, and it’s much safer to do this in a video game than in the carpool lane.

To make a racing game you need detailed road models and attractive models of the nearby scenery. Some games use imaginary roads and locations created from scratch by designers. Other games painstakingly and expensively recreate real roads and real environments. The verisimilitude of seeing real landmarks whiz by adds to the fun.

In either case the cost of creating good quality content is one of the major costs of creating the game. In the future there will be third option. A game company will be able to license the entirety of the real world’s roads and all the surrounding areas from Google. The data will be streamed over the network from Google’s servers, where it is displayed seamlessly with the actual game elements.

The physical road surface will be mapped at centimeter resolution, every bump and rise accurately depicted. The visuals along the route will be in 3D and super high resolution, you could see the spots on a ladybug if you slowed down enough. And playback during the game will be silky smooth thanks to copious bandwidth and massive amounts of local hardware to warp the imagery in response to each subtle change in eye position.

Street View

Today Google’s Street View is an early preview of this virtual world technology. Google’s Street View was launched in May of 2007 and for the last 3.5 years Google’s cars and bicycles have been scouring the planet extending coverage, which is now comprehensive in many cities. Google recently announced that all 7 continents have at least some imagery: they added a few token panoramas in Antartica.

I have used Street View periodically. When I’m driving to a new location I sometimes use Street View to get my bearings and visualize landmarks. I’ve used Street View to see if a hotel was in a good neighborhood. I’ve used Street View just for nostalgia, to sneak a peek at places I used to live.

Despite my felicitous experiences with the feature, the sheer number of minutes I’ve used Street View is pretty small. I rely on Google Maps itself much more heavily, and use other Google products way more than Maps. So I have wondered, is it really worth the effort and expense of driving every mile of road of the planet just for this ability to peek at remote locations?

The investment starts to make sense when you think long term. Mostly what Google is doing is developing technology. From the camera clusters and range finders on the vehicles to server-side processing to client rendering. As the technology improves the quality of the virtual world will improve. Eventually the world will be useful for more than just “peeking at a remote location.” It will become a platform on which games and other graphically rich applications can be built. Google will likely license the content to 3rd parties, who will actually build the games and applications.

To make this all work Google will need to improve upon today’s Street View in 3 significant ways.

Removing all people and vehicles

Google blurs faces and license plates today. This is a processing step done after acquiring the imagery but before making it available for viewing. In the future Google will need to do even more sophisticated processing to remove all people and vehicles completely. Most games and applications will want to populate the world with their own computer generated vehicles and people, the static ones which happened to be there when the Google car drove by would be a distraction.

Removing these objects will require combining the data from multiple Street View runs, leveraging the fact that many cars and people will move between runs. However there will still be areas which are obscured in all runs. Google will employ techniques similar to Photoshop’s Content Aware Fill. This amazing feature infers what might have been under a deleted object, see for example this video.

Smooth movement

Today panning and zooming in Street View are smooth, but moving is more like slideshow, a blurry step from one location to another. By collecting tons more data and improving the client display algorithms it will eventually be possible to have completely smooth visuals during movement whether walking, driving or zooming along at cruise missile speeds.

Increased Resolution

Street View today is fairly high resolution in places but degrades quickly as you look far from the main path. Camera hardware is always improving, so that alone will improve quality over time.

The other way to improve resolution is to collect imagery from more locations, including those well away from any street. Google has reportedly been experimenting with quadcopters as a possible way to get additional vantage points. A swarm of Google copters could collect imagery far more comprensively than the Street View cars and bicycles.

Google may also be looking to acquiring imagery inside stores and buildings. Building interiors would certainly enhance the virtual world.

Once these technologies are in place and a smooth and believable virtual world exists, what sorts of games and applications will be developed? The already mentioned driving game is an obvious choice since Street View is focused on streets today, but as content for the wider world comes online, more possibilities open up.

Shooting Games

First person shooters are one the of the most popular game genres. You run around on foot or in specialized vehicles trying to shoot other game characters (AI or human controlled). As with driving games players crave new spaces to play in. The real world would provide an almost infinite variety of terrain and obstacles.

These games could benefit from the ability to inflict realistic damage on buildings and terrain. The idea of destroying and fighting around real world landmarks will be controversial, possibly even banned for some locations, but where permitted the target audience will happily embrace the real world as yet another play field.

Sim Games

Will Wright’s style of simulation games like Sim City and The Sims could be rewritten for Google’s virtual world. Initially games might keep the bulk of the world unchanged, just add it game entities on top. A more technically difficult version would allow changing the world itself, adding or subtracting buildings and even terrain, changing rivers and mountains. In either case you could always drop down into a first-person view, ride along with a delivery vehicle to see first hand how your new traffic plan is working out, for example.

Training and Monitoring

So called “serious games” will also be popular on Google’s platform. Bus drivers can learn their new routes, fire and police can practice cutting through traffic, disaster preparedness drills can be run with many online participants. In kind of a reverse simulation, the the position and status of real world objects, say all the police cars in a city, could be displayed in the virtual world for city officials to monitor. This is David Galernter’s Mirror Worlds idea from the early 1990’s.

Exercise Machines

It’s not a new idea to have a treadmill or exercise bike with a screen which shows some kind of virtual world, to motivate and entertain the exerciser. Hooking them up to Google’s virtual world would just be a way to vastly increase the amount and quality of available content. People could bike or run along scenic trails in Hawaii or Malibu or anywhere they choose.

Shopping and Social

Non-game activities like shopping or socializing in famous locations would be a draw for many people. In a virtual world you can interact seamlessly with people from all over the real world, wherever they are physically located. Similarly you can travel instantly, visit LA and Tokyo and Milan all in the same evening. This type of casual usage is reminiscent of the “metaverse” concept from Neil Stephenson’s Snow Crash.

Not all games will take place in Google’s virtual world. In fact most games will continue to build their own custom worlds, with fantasy or historical elements that just don’t exist in the real world. But a healthy subset of games and applications will, at least some of the time, reside in this new virtual world.

Google is not the only player in this game. Microsoft’s Streetside feature is a nearly identical to Street View. It is likely in the future there will always be handful of companies with the resources and motivation to model the entire world. These different companies will compete and will repeatedly re-image the planet as technology improves. The real world is a valuable source of content and from here on out it will be heavily and repeatedly mined for our benefit.

Addendum: This post is speculation on my part. I have no inside information regarding Google's plans. I do not work for Google.

Clock Speed Wall

noreply@blogger.com (Philip) — Mon, 13 Sep 2010 21:26:00 +0000

Intel 8088

In 1981 the original IBM PC was released. The CPU which powered the PC was an Intel 8088 running at 4.77MHz. That means the internal clock ticked 4.77 million times each second. Over the next 24 years the clock speeds of PCs rose consistently and rapidly.

For PC enthusiasts the old speeds are a trip down memory lane: 12MHz, 25MHz, 33MHz, 66MHz, 90MHz, 133MHz, 300MHz, 600MHz, 1.2GHz, 2.4GHz.

About 5 years ago this 24 year trend came to an end. In 2005 the fastest CPUs ran around 3GHz, 3 billion ticks per second. Today clocks are no faster, still around 3GHz. For the foreseeable future clock speeds will be around this speed, they are no longer progressing.

Clock Speeds peaked in 2005

What caused speeds to hit a wall? The answer in one word is heat. As integrated circuits were made smaller and smaller, the clock speeds could run faster. But faster clock speeds generated more heat. And as the nail in the coffin, smaller sized chips had less surface area to dissipate the heat. So manufacturers created faster and faster chips until one day they simply could not. Overheated chips have random errors, they don't work.

For those 24 years the increasing clock speeds were an exponential trend. People were predicting 10GHz speeds just prior to discovering 3GHz was actually the limit. This is a good example of how exponential trends can and do sometimes end suddenly. More on that later.

Clock speed isn’t a primary feature of a CPU. It’s similar to engine size in cars. For example the 2010 Ford Mustang has a 5.0L engine. All else being equal a bigger engine will yield better performance, but all is not always equal. A more advanced engine can outperform an older one of the same size. Same with MHz and CPUs, with ongoing effort newer CPUs continue to perform better despite having the same clock speed as older ones.

Interestingly Moore’s Law says nothing about clock speed or performance. Moore’s Law only says the number of transistors on a chip will double every two years, that’s it. Moore’s Law gives chip designers ever larger “transistor budgets” which they can spend however they think will best benefit performance.

When clock speeds were rising designers spent their transistor budget on ever more sophisticated ways to “keep up” with the new clock speeds. Each passing generation did not just have a faster clock, it was more clever and did more work with each tick. Once clock speeds stopped rising, chip designers no longer needed spend transistors on clock speed increases.

Intel Core i7 CPU

In order to find productive uses for all the new transistors designers began adding multiple cores. A core is portion of silicon which implements the “core functions” of the CPU. So a multi-core system is one with multiple mini-CPU’s inside the full CPU. Each core can be executing a separate program or a separate thread of a program at the same time. The chip can do several things at once.

The big catch is that software must be specifically written or modified to take advantage of multiple cores. With 4 and 6 core systems out there software is just starting to use multiple cores. As the number of cores in chips increases, the penalty for not have multi-core software will increase. Suppose you had a 1000 core chip, a naive program could only use 1/1000th of the chip, while more advanced software could use the whole thing.

Let’s take a step back and consider a hypothetical world where a form of Moore’s Law existed for constructing buildings. The law would state that every two years the size building you could “inexpensively” build would double. Now suppose for the first 24 years the industry spends their construction budget building higher and higher buildings. At one point pundits would predict we’ll soon have buildings 10 miles high.

Instead everyone finds that out due to air resistance and the properties of concrete and steel no building can in fact be higher than 300 stories. So now what would the construction industry do? The only thing they can do: build out. They would create clusters of 2 then 4 then 6 office towers, all about 300 stories tall.

This is where we are today, we have just started to build out. We have just turned the corner from faster clocks speeds to more cores. What lies down this street? Certainly more cores, but how many? Where will it end?

The "clock speed wall" has actually been a good lesson for us all that real Physical limits are out there, that technology trends can be long running yet finite. Moore's Law itself is next on the chopping block. Some predict it will end by 2020 while other see possible extensions though increasingly clever means. Eventually Physics will win. When engineers can no longer double the number of transistors on a chip, designers will have to make due with fixed transistor budgets just like today they make due with fixed clock speeds.

If we are lucky at that point some futuristic technology will be ready for the spotlight. The new technology will not have the same limits which integrated circuits have. Maybe we will we will once again get to build up as well as out. I look forward to finally getting those 10 mile high buildings.

The Singularity Is Always Steep

noreply@blogger.com (Philip) — Thu, 29 Jul 2010 17:45:00 +0000

Moore’s Law predicts that the number of transistors we can pack into an integrated circuit will double every two years. This has held true since integrated circuits first appeared in 1958 and is expected to continue until around 2020 when physics will prohibit making conventional circuits any smaller.

But the party won’t end there. Ray Kurzweil explains that integrated circuits are not the first “information technology” to follow this consistent doubling of capability. Instead they are the fifth such paradigm. Earlier paradigms included mechanical relays and vacuum tubes. Each paradigm reigns supreme for a while, but when a physical limit is reached the paradigm steps aside and a newer more advanced technology carries the ball.

Kurzweil’s prediction is that a new technology will take over around 2020 to continue the exponential growth. I believe this will happen. The economic incentive is too great and absolute physical limits of computation are too remote.¹

So if exponential growth continues past 2020 where does it take us? For all these graphs I show a doubling every single year, not every two years as Moore’s Law predicts. Kurwzeil says if you consider price as well as performance doubling every year is more accurate. And as we'll see, the exact numbers do not matter as much as you might think.

Here is the 30 year projection:

This shows performance relative to today. So this graph predicts we will see a 10⁹ (one billion) fold improvement by the year 2040.

As you can see the line shoots skyward as we approach 2040. Kurzweil predicts The Singularity will happen around 2040. This graph makes 2040 look special, it generates excitement around that year. Let’s look out a little further:

Something funny is going on. The manic peak around 2040 is basically flat in this new graph. The huge increases in the final 5 years dwarf all progress from previous years. Now 2050 looks like the place to be.

Let’s look further out:

Again the same pattern. The whole first half of the century through 2050 now looks an absolute bore. But around 2055 things start rocking.

One more projection through the end of the century:

Basically nothing going on until around 2095 when things take off. It’s very much like finding your keys in the last place you look. Wherever you choose to stop drawing the curve, that looks like the exciting peak. This is the nature of exponential curves. The "big event" is always right at the end of your graph.

The solution of course is to use a log scale:

Now all the excitement has been replaced with a boring line. But wait. Which is more representative of how we perceive progress? Is it building to an exciting peak? Or just chugging along incrementally?

We in fact are making steady progress and not building to anything in particular. Think about your own experience. A computer twice as fast is always just a barely noticeable improvement. This will hold in 2050 the same as it does today. Doubling a 2050 computer is adding a mind-blogging amount of power, but it will only count as one small tick of progress. So the log plot is the right one.

Raw computing capacity is increasing exponentially, but computational utility, what computers can do for us, is increasing only linearly.

There is no peak at 2040 or any other year. So what does this say about The Singularity? The Singularity is not about graphs. It's about reaching a threshold where machine intelligence is common place. Where machines create new generations of machines ad infinitum.

Nothing about these graphs suggests The Singularity won't happen. If we reach the threshold where we can build machines more intelligent than ourselves that will be a strange and exciting day to be sure.

What these graphs remind us is that no magic upward surge in technology is coming in 2040 or any other year. It's steady consistent progress, one hard earned doubling at a time.

What is the ultimate physical limit of computation?

Bremermann’s limit describes how much computation is possible in matter, for example if you could use every molecule in water towards your computation. There are a lot of molecules in matter so the limit is very high. You could also try making the computer very big.

Wikipedia says a computer the size of the earth could perform 10⁷⁵ operations per second. Today a supercomputer can do 10¹⁵ operations per second. Our 2100 prediction is computers would be 10²⁷ times faster than today, so the 2100 era supercomputer would run 10⁴² operations per second. This is amazingly fast, but it's nowhere near 10⁷⁵, so we won't be near the absolute physical limit even in 2100.

See also:

The Singularity Is Always Near

Inception Game

noreply@blogger.com (Philip) — Mon, 26 Jul 2010 13:18:00 +0000

Spoiler Warning: I discuss the movie Inception below.

Inception is an ambitious and brilliant film. But it starts badly. A very stiff and cliched setup motivates our hero to “do just one last job”. Please. Thankfully the final 2/3 is so spectacular and so well executed it completely erases your memory of the first 1/3.

The director Christopher Nolan is ridiculously ambitious. A dream within a dream would have been something. But like nesting dolls he keeps packing one timeline inside another and pulls it off with tremendous visual style.

Inception is the latest “virtual reality” movie. In these films the characters spend a significant amount of screen time completely immobilized, lying in a bed or chair with beeping equipment nearby. Meanwhile their minds are engaged piloting avatars in a virtual world of some kind.

In The Matrix (1999) the avatars and the world were computer generated. In the disappointing Surrogates (2009) the avatars were mechanical and the world was real. In Avatar (2009) the avatars were biological and the world was alien. In Inception (2010) it’s a shared dream.

Although the purported technology of each of these is a different, they are in effect all illustrations of what a conventional computer-based Virtual Reality system might be like. With a sufficiently advanced brain-computer interface the computer could synthesize any world, realistic or surreal.

What types of virtual realities will people want to experience? Prurient activities are a given. But the games and movies of today are probably not a bad indication of our interests. We'll see similar plots and activities extended into fully realized 3D worlds. There will be lots of guns.

Inception depicts what could be a pretty good structure for a virtual reality game. Forget the chore of performing inception itself and consider just navigating the multiple timelines, getting from one level to the next.

Suppose this happens on a Playstation 9 console. Your friends all connect to the machine and enter the first virtual world. Say you start in a hotel. This level is simple. You just need to rendezvous with everyone on your team without creating a scene. You find a quiet room and setup your virtual Playstation. If you didn’t generate any suspicion no one should come looking for you, and you should have time on the next level. You descend into a completely different scenario below.

Each level gets progressively harder. As in the movie 5 minutes of undisturbed time on one level buys you 60 minutes on the next. You can't leave anyone behind, because there is no way to "speed up" our brains as in the movie. But you can have arbitrarily many levels going at once with your characters suspended in each previous level. The previous worlds will still be running. If someone does discover your immobilized avatars in the game, it instantly brings you back to that level, you have to evade them and start over from there.

So the challenge is to have your avatars safely plugged-in on every level at once enabling you to reach the final level. There is no reason to stop at 3 or 4 as in the movie. A hard game might be 12 levels or 30 levels or something extreme.

It would be both a puzzle and an action game. On some levels there is a trick to discover, a way to make a distraction or a secret location to find. In other cases maybe you have fight it out. The clock is always ticking, if you run out of time on the level above, the game is over.

Unlike the movie there is no need for existential crisis. You know you are in a virtual world because at any time you can bring up an in-game menu, and chose ‘exit’ if you wish. And perhaps for fun the way to bring up the menu is by spinning a totem in the game world.

Robot Cars: Getting from here to there

noreply@blogger.com (Philip) — Wed, 16 Jun 2010 13:08:00 +0000

When will you first ride in a robot car?

It might be on a business trip. You wearily trudge out of the terminal and into the line at the taxi stand. About 1/3 of the taxis are robot cars. They look just like regular vehicles but with dark windshields and some extra gear on the roof and the front.

And no drivers.

Initially the robot cars might only be certified to go to certain destinations such as major hotels in the city center. This is fine, you are going to the downtown Hilton. A human dispatcher loads your bags in the trunk and asks for your destination. He speaks into his wrist and the taxi's roof light goes off to indicate it has taken a fare. You tentatively hop in.

There is no driver to talk to, but there are large screens in the back seat showing your progress along the route to your hotel. And a big red call button if you need assistance. The ride is uneventful and smooth. You arrive quickly and safely. You pay electronically from your mobile phone. By recently established convention you are given no option to tip.

Years later you might purchase a robot car of your own. You go to the Honda dealer and select a mini-van. Not radically different from today's models except again for some extra external gear.

The driver's seat swivels because you can face any direction while the robot is driving. The robovan still has manual controls so you can steer on the rare "unsupported" road. During a typical drive however no intervention is required. You enter the address and then read, play with the kids, watch a movie, or even snooze. In the latter case the car can be programmed to wake you a few minutes before arrival.

When will all this happen? The technology required is actively being developed. In 2004 DARPA ran a "Grand Challenge" contest for robot cars. No vehicle completed more than 7 miles out of the 150 mile course. That's like holding a marathon and no one manages to run more than a mile! The very next year however 5 cars finished the completed the entire 132 mile course. That's incredible progress for one year. And that was 5 years ago.

Today robot cars are hitting speeds of 130 mph, drifting around corners and performing stunt moves like sliding into parking spots. It will not be long before robot cars have the technical ability to drive in normal traffic on any road.

The question is, will we let them? This is a big social and legal question. How will people feel about sharing the road with robots? How will people respond to the crashes involving robot cars? Where does liability fall? What happens to people who make their living driving today?

Robot cars today seem "futuristic" but by the time they become common the world will be a different place. Advances in computers, robotics, medicine, and manufacturing will dazzle us. Robots cars will be impressive, but will pale in comparison to whatever is brewing next.

Consider if it is 20 years before you personally own a robot car. That's 7,300 issues of the New York Times you'll have to plow through, 350,000 rotations of CNN Headline News. You will see and hear every angle of this story unfold in excruciating detail. Every technical milestone and every legal battle, every setback. You'll be sick to death of robot cars before you set foot in one.

Can this all really happen in 20 years? I really don't know how long it will take. It's a social and legal question mostly, how quickly do we want it to happen? What I can predict is there are inevitable milestones we'll have to pass along the way.

So these are my meta-predictions about robot cars:

1) 1 billion hours of simulated driving

Robot cars will come of age in a era where have copious computer time to run all manor of simulation. One focus for simulations will be crash or near-crash situations. Yes robot cars can crash, if there is simply no way to avoid it. But robot cars will minimize injury or damage, instantly evaluating every possible evasive maneuver and executing the best one.

2) 100 million hours of actual test driving

A fleet 10,000 test cars could hit 100M hours of use in just a few years. Starting on isolated test tracks with fake traffic and eventually winning legal approval to drive on real roads with real traffic. At first a human will be sitting behind the wheel, ready to override the machine. Later humans will be on waiting in a centralized location to remotely pilot vehicles only if needed.

3) Robot car drives from New York to LA

This is more of a stunt than an important milestone, but it will get a lot of coverage. Small crowds will gather in each city as the car drives though. It will be a slow meandering course, not a robot Canonball Run. Security will be a concern as auto-Luddites make threaten to destroy the car.

4) 100 million hours of real-world driving

Pilot programs would exist for private vehicle owners as well as commercial vehicles like taxis, delivery vehicles, long haul trucks. The commercial world has the biggest motive to adopt robot cars, for cost saving and safety, but also the biggest resistance because of the job loss. Vehicles would be monitored closely with remote humans standing by to take over if needed.

5) Crash statistics demonstrate safety

Even if robot cars were only moderately better drivers than humans, they would be vastly safer. Today we have more than 30,000 vehicle fatalities per year, and 10,000 of them involve alcohol. Robot cars will never drink, never be tired, never be distracted. Crash statistics will show robots are 100X safer than human drivers.

6) Robot car beats top Indy 500 drivers

They won't let a robot car in the real Indianapolis 500, but robots will compete in exhibitions against professional drivers. The competitions will be repeat until the humans no longer stand a chance at winning. Normal human-only competitions will continue.

7) Staged rollout for private use

Unrestricted private use will begin slowly, with early adopters lining up for short runs of a few thousand cars. Over the years production runs will ramp until tens of millions of cars are sold every year. Finally it's your turn. Stroll down to the dealership and pick up your brand new robot car!

Intelligent Designer: Gaia

noreply@blogger.com (Philip) — Sat, 15 May 2010 15:45:00 +0000

Intelligent Design is a clever modern reformulation of Creationism, stripped of all overt references to God or Religion in order to sanitize it for inclusion in public school curricula.

The Intelligent Design argument is that the diversity and complexity of life on this planet simply looks designed. They assert that chance could not possibly have created the human eye, for example, or the complex ecosystem of the rainforest. Just look at it, they say, there must be an intelligent designer behind the scenes.

Intelligent Design is not taken seriously. It's seen a just a way for fundamentalist Christians to get religious views taught in schools.

But what if they are on to something? What if Intelligent Design is in fact exactly right? What if there is an intelligent cause? Not an intelligent cause which replaces Natural Selection, but an intelligent cause which produces Natural Selection.

That is, what if Natural Selection and Intelligent Design are both right?

Natural Selection takes place in the union of all organisms and their environments on Earth. This superorganism is called Gaia. Suppose Gaia is intelligent. Not just kinda-sorta intelligent but literally intelligent in the same way we humans are intelligent. Then Gaia is the intelligent cause. And the way she works is Natural Selection.

The elephant in the room here is what do we mean by intelligence? There are a grab bag of attributes which we associate with intelligence: abstract thought, reasoning, planning, problem solving, communication and learning. The key question is, are all of these required for something to be intelligent? Could Gaia have kind of a dry mechanical intelligence, devoid of a human-like consciousness?

Maybe the human eye looks designed because it is designed, not by a supernatural being, but designed by Gaia using a process similar to how a human would do it. Consciously we don't try millions of alternatives in order to see which works the best, but maybe we do in our subconscious. Maybe the spark of inspiration, that "aha moment", is just the final product after many mechanical iterations of evolution, an evolution of neural firing patterns rather than genes.

This points to the big difference between Gaia and ourselves. We have a nervous system. A nervous system is immensely useful for an organism. It allows rapid communication across distances at speeds which are decoupled from locomotion or the undirected diffusion of chemicals. A nervous system lets you think without any externally visible action.

In Avatar James Cameron depicts a planet-wide intelligence which the natives call Eywa. Eywa does have a central nervous system, a very big-screen-friendly bioluminescent energy that plants and animals sensuously plug into.

Gaia did not evolve a biological nervous system like Eywa. But using her plodding conventional intelligence she has created organisms who are dutifully wiring up the planet with fiber optic cables and wireless communication channels. Could these communication links layed for our own small purposes be used by Gaia as well? How will Gaia change once she has a nervous system?

The Healthcare Diff Part 2

noreply@blogger.com (Philip) — Fri, 19 Mar 2010 03:46:00 +0000

I received some good comments on my post The Healthcare Diff including some on Hacker News. One reader pointed to this excellent article in The New Yorker by Atul Gawande. Gawande says the Senate healthcare bill is chock-full of experiments, it's just they are called pilot programs. The article includes an extended analogy between healthcare today and the state of agriculture in the early 20th century. He asserts the transition to a modern efficient system of food production was largely due to government sponsored pilot programs.

This is very encouraging.

One reason information about the healthcare bill's contents is not widely known is "the process story".

NPR reported that the media loves explaining the process by which a bill is making its way through Congress, but they generally minimize discussion of the bills themselves. For example here is the top hit in a search for "healthcare" on Google News for each of the last 14 days:

March 17th: Nuns in US back healthcare bill despite Catholic bishops' opposition

March 16th: Democrats Gain Momentum for Healthcare vote, but Process Draws GOP Attacks

March 15th: House Democrats Begin Final Push on Healthcare as Pressures Intensify

March 14th: Democrats lack votes to pass healthcare bill, but White House confident

March 13th: Obama delays Asia trip to push healthcare overhaul

March 12th: Middle East peace efforts: lessons from healthcare reform

March 11th: Democrats Edge Nearer Agreement on Final Healthcare Package, but Obstacles Remain

March 10th: Panel on Compromises to Pass Health Care; What's Next in Afghanistan?

March 9th: Abortion could stymie healthcare reform legislation

March 8th: Obama adds fire to healthcare debate

March 7th: Anthem a boon to Obama's healthcare efforts

March 6th: Obama: Congress 'Must Act Now' On Health Care Bill

March 5th: Obama's Health Care Push: The Race is On

March 4th: Pressure mounts for/against healthcare bill

It's clear that a casual observation of the daily news will not teach you much about the bill itself, only about its status. As the NPR reporter quipped "news is what's new". A good reason to continue to support long-form journalism like The New Yorker!

Two other good links from the comments

Impressive RAND study conducted from 1971 and 1982 about "cost sharing" and health care effectiveness
Paul Fromer talk on "Charter Cities", a radical call for experimental governments.

The Healthcare Diff

noreply@blogger.com (Philip) — Sat, 13 Mar 2010 03:53:00 +0000

On November 7, 2009 the House narrowly passed its health care bill. On December 24th the Senate passed their own version. Today several months later a "reconciliation" between the two is in progress, the final version is expected "soon".

It is a wonder of the modern age that you can pull up the full text of either bill with a single click. Each is a relatively small file, much smaller than a YouTube video. Here is the 1990 page House bill or the 2074 page Senate bill.

Of course having access to the text is one thing, understanding it is another matter. Aside from the flowery and confusing prose in general certain passages stand out as particularly opaque. Consider this gem on page 876 of the Senate bill:

The reason this is hard to read is that it doesn't stand alone. It is a change to existing legislation, in this case the Social Security Act. You need to look up the passage is question and manually apply the changes.

A software person immediately recognizes this as a "diff". A diff in software is a "difference report", a listing of differences between two versions of a program. Programmers live and breathe diffs, they are both routine and essential.

The style of diffs in these bills manages to be both verbose and cryptic. It is a surgical description of the exact change with zero context. Compare this with a software-style diff of the same change:

Here we plainly see two lines are going away and are being replaced by 4 new ones. This already is much clearer than the bill. Clearer still is a graphical diff which shows the changes side-by-side:

Click above image to enlarge.

Even if the bills replaced their cryptic diffs with beautiful graphical ones, there would still be a big problem. The changes represented by these bills are completely untested.

Programmers "branch" when making big changes. They create two whole copies of the software. Changes can be applied to the branch without harming the main version. If the branch proves to be successful, after extensive testing, the branched version can be "rolled back" into the main version.

Or consider Google. Google handles hundreds of millions of searches every day. But not everyone is treated equally. With every search there is a small probability you will be part of an experiment. Google runs hundreds of experiments simultaneously. Some are minor variations like a different font or color, while some are major and return completely different search results.

Google never bugs you with a survey for feedback, they simply collect data about what results you click on or the timing and pattern of your follow-on queries. The results of the experiments are used to improve the service, which generates every more people to use in future experiments.

Our approach to legislation is that we don't branch and we don't experiment. We fight and argue and cook up a giant legislation-bomb where half the people think it will work and half think it will be a complete disaster, then we just try it out.

Instead of arguing over the one true solution on the basis of hunches and intuition, we should implement more than one approach. Pick the best 2 or best 5 solutions and try them all. We have 307M people in this country, they could be generating and enormous stream of empirical data about what works and what doesn't. Statisticians could be poring over patient satisfaction surveys, hospital expense reports, bankruptcy rates. Find out what settings on the knobs actually solve the problem instead of just winging it.

The early internet developers shunned hammering out agreements in detail, instead they valued "rough consensus and running code". Ship early, iterate often.

Pancake Stacking

noreply@blogger.com (Philip) — Thu, 11 Feb 2010 22:04:00 +0000

The blog Singularity Hub has a post about robots in factories. The post includes 5 pretty cool videos (plus one commercial, which is fictional). I thought the best video was this one:

There's something so mundane about the task here. In a lot of ways it's hard to believe it's not just easier to have people do the stacking. While the robot is super fast, it's only doing a single pancake at a time. I imagine a person would scoop up a few and plop them down already stacked together.

At any rate they must have done all the math and figured this is better or cheaper or both. Whenever I open a toy for the kids I wonder what steps in the assembly were performed by robot vs. human. Did a person twist tie all these wretched wires immobilizing poor Dora, or was that automated? We the consumer are generally oblivious. Maybe some day every product will be required to come with a "making of" video, so we can evaluate the production ourselves.

For example here's a video of a Chinese factory where they are stamping some kind of metal part:

Talk about poor working conditions and a mindless repetitive task. So who decides when it makes sense to drop in some fancy industrial robot to take their place? It's clearly an economic decision and the cost of labor is probably the major factor.

Ultimately as wages rise and robotics get cheaper there will be more and more automation and less factory workers. But how long does that switch-over take, given that factories are being built all the time, many in low-wage countries? It seems a complex dance between economics and technology that will play out in the coming decades.

Ultimately robots will "win" because they will just keep getting cheaper while human wages should go up over time. Historically when automation takes over an industry (like farming in the US) the lost jobs are more than made up for by entirely new jobs which didn't exist before (like computer programming). Hopefully that will continue to happen.

At any rate the transition is going to take time. Even if the robots were perfect and free today there would be some transition, as new factories were planned and constructed. And those robotic pancake pickers can't be cheap. There are no prices on the FlexPicker website, but feel free to contact them with your intended application and they will be in touch.

iPad: Bring It On

noreply@blogger.com (Philip) — Mon, 08 Feb 2010 04:38:00 +0000

On January 27th Apple announced their new iPad tablet. It's a nearly featureless rectangular slab with a thick black frame surrounding a 9.7" screen. It has a single button on the front. The back is shiny metal with the Apple logo. It looks exactly like a big iPhone.

Interaction is through the "Multi-Touch" screen, same as with an iPhone. It was designed to fill the gap in their product line between an iPhone and a MacBook. That is, it's meant to be more computer-like and capable than an iPhone, with a bigger screen, but simpler and more compact than a laptop.

There has been a lot of discussion about the iPad announcement. Some of it quite negative. The Free Software Foundation says it is "a huge step backwards in the history of computing". And there are the inevitable enumerations like 8 Things That Suck About the iPad or 16 Reason the iPad Sucks. Other bloggers call the iPad's future frightening and dystopian.

Why all the fuss? There are multiple reasons, but much of the anger stems from the "closed" nature of the device. Apple chose not to create a computer-like tablet: a stripped down MacBook without a keyboard. Instead they created a scaled-up iPhone and they are adopting the "locked down" model which they've successfully wielded over the iPhone.

What does "locked down" mean? Physically the iPad doesn't have standard computer connectors like USB or DVI or HDMI. You cannot connect the iPad to the hundreds of millions of USB devices out there, or hook it up to your monitor or TV. Perhaps Apple will make special connectors available, but this is entirely at their discretion, they are in control.

Similarly the software model is "closed." Anyone writing software for the iPad will need pay Apple for the developer's kit and then get their software approved by Apple. Apple can choose to deny any software it doesn't want on the device. And if they approve your software then you have the privilege of forking over 30% of your revenue: they get a piece of every sale.

This is entirely unlike how the computer business works. Both Macs and PCs today are "open" in the sense that developers are free to write and distribute software without getting any approval and without any revenue sharing. For example Microsoft has no direct control over Windows developers.

So where does this "locked down" model come from and why is Apple pushing it? Apple's first successful product was the Apple II computer released in 1977. By 1984 they had sold 2 million Apple II's and had released the Macintosh, the common ancestor of all current Apple computers. Over the years Apple created desktops, laptops, the all-in-one iMac, even servers.

In 2001 Apple came out with their first major non-computer product, the iPod. The same year they opened their first retail store. In 2007 they announced another non-computer product, the iPhone, and shortened their name from "Apple Computer, Inc." to just "Apple Inc.". Both the iPod and iPhone were run-away hits, racking up huge sales while becoming cultural icons. Today Apple has 294 retail locations worldwide and has sold a total of 240M iPods and 40M iPhones. During Apple's most recent quarter alone they sold 21M iPods, 8.7M iPhones and 3.4M computers.

So for the last 10 years Apple has been selling more and more "devices" rather than traditional computers. The iPod was basically a fixed function device, like a digital camera or a GPS. But the iPhone and iPad are more computer-like, they can run 3rd party applications. There are over 100,000 "apps" available in Apple's "App Store". Apple has essentially evolved a new product category, something in between a traditional consumer electronic device and a computer.

So is this a good thing? I believe that innovation is a good thing. For better or worse they are choosing to innovate not just in software and hardware but in business models. All else being equal I personally would rather have a device that was open. I enjoy the freedom and flexibility of normal computers and I think the model has been successful.

However, I also believe Apple has the right to establish any terms for their product so long as they are legal. Apple is beholden to their shareholders, if they believe a closed model will lead to greater revenue, they should pursue that. Apple is choosing this route not capriciously but because they feel the extra control will provide the best possible user experience. The traditional computer experience is often quite painful.

This "locked down" arrangement is normal in the world of video game consoles. Sony, Microsoft and Nintendo all maintain very tight control over their platforms, they approve every release and receive revenue from every game sale. These closed platforms are very successful, they generate billions of dollars in game sales and millions of people enjoy them. This is not to say computers should follow suit, but it demonstrates highly commercial closed systems can deliver satisfying experiences for wide audiences.

Apple spends over a billion dollars a year on R&D. They are taking a huge risk with the tablet, the form factor is unproven and even possibly even jinxed by past failures. The company expects a return on their investment and their risk taking. Again if I could get the iPad hardware without the restrictive platform, I would take it. But that's not an option at this point. The platform is closed, but we are free to choose for ourselves whether to buy one or wait for something different. This is how it works. The marketplace will sort it out in the end. Open, closed or something in between.

Good ideas will flow to all platforms eventually. If the iPad is successful, we'll see open alternatives, the tablet business is unlikely to evolve exactly like the console business. I appreciate and value open platforms, but they cannot be mandated. Competition is even more crucial than openness.

Green Screen Scenes

noreply@blogger.com (Philip) — Sat, 23 Jan 2010 02:37:00 +0000

This is a great "behind the scenes" video by the production company Stargate Studios. It shows how popular television shows like 24, Ugly Betty, and Heroes use green screens in places where you would not think that special effects are involved.

My favorite scene starts about 30 seconds in. It's from Ugly Betty and she walks out of a building and smacks into a bus stop shelter. I figured they would at least built a real bus shelter for her to walk into and just insert a new background. But even the bus shelter is fake! In the green screen version she just walks into a small piece of plastic. Essentially everything is added "in post".

This insight into current TV production is a great complement to the Avatar behind the scenes video. Both illustrate different forms of virtual productions. In Avatar everything is virtual, the main characters and the whole environment. With TV shows we see they are mainly "faking" only the sets. For now.

The big question is, how many years until Avatar-style "performance capture" is used to generate an ordinary human character? Maybe it will first be done in a movie, but costs are always dropping so soon after they will start popping up on TV. This will be a big milestone: a virtual character on a mainstream TV show. Not a sci-fi show, but a drama or comedy, just another member of the cast.

For example we might see an older accomplished actor "driving" a younger version of themselves, or someone else entirely. A completely convincing virtual human performance is not that far away. And after the first one, many will follow. What will be the implications for actors? What about the the poor out-of-work plastic surgeons? It will be a strange and interesting future.

Computational Footprint

noreply@blogger.com (Philip) — Tue, 19 Jan 2010 02:47:00 +0000

A statistic you won't see on Box Office Mojo is how many CPU cycles were spent making a given film. An exact count would be impossible, but no one gives even a rough estimate!

I saw Avatar in 3D the other week and I would guess it is the most computationally expensive movie ever. The complexity of the scenes and the sheer number of screen minutes which are 100% virtual is stunning.

Avatar's main characters are synthetic but so is their whole world, with rich and detailed plant and animal life and stunning landscapes. If you watch the "making of" footage you see them "filming" in an empty sound stage: essentially everything is 3D graphics added in later. But its all believable, it's so well done that it's not hard to suspend disbelief and just sit back and enjoy the movie.

Most of the effects for Avatar were created by Weta Digital, and this article about their server farm says they have 40,000 processors. But for a full accounting of CPU cycles let's go all the way back to James Cameron typing up the screenplay. Even an average laptop crunches away at a few billion cycles per second. If he spends 8 hours writing he alone has burnt through trillions of cycles.

When pre-production started probably thousands of computer hours went into proving out the technology and developing the characters. And once in production an army of 3d modelers, animators and lighters cranked away for years. Not to mention less powerful laptops and desktops of everyone involved, from the producers to the schedulers. And let's throw in everyone's iPhones and iPods and anything else used on set or in support of the production. All these are ticking away millions or billions of times per second.

Now let's cast the net even wider. What about cycles spent long before production started. All these high tech products were designed on computers. All the software used in the production was written somewhere, slaved over line by line for months or years. Software like Windows or Photoshop has been around for decades, so the total computer time spent developing them is staggering. These CPU cycles were spent to benefit many users, not just Avatar's staff, but they provided the foundation, the context in which this highly advanced movie could be made.

Now you can see what it takes to become the most computationally expensive movie to date. Any big modern production can vie for the title, because with each passing year the computational mountain we all stand on is built up a little higher. In Avatar's case you take all that legacy, a very big production team, and a 40,000 processor server farm, and I bet they have set a new record. For now.

Cancun Beach Reconstruction

noreply@blogger.com (Philip) — Thu, 07 Jan 2010 03:18:00 +0000

We went to Cancun for a family wedding on New Year's Eve. We stayed at the Fiesta Americana Condesa resort which was pretty nice but not without faults. One problem is our trip happened exactly when the resort's beach was being reconstructed!

The government is marching down the long strip of hotels in Cancun rebuilding the beach. They move pretty fast, over the 5 days we were there they passed completely over the resort, from one side to the other. The hotel's beach access was closed the whole time. This newly rebuilt area was not open when we left but probably will be shortly. Further up the beach it was open, so we visited the beach there. The new beach is nice although there is a 6 foot drop off near the water line, they must expect it to blend in over time.

The process was kind of interesting to watch. They are using a single pipe which makes landfall in the distance, then they extend the pipe for each new section. The whole reconstruction is just sand, they don't add any concrete or other supports, they are just creating a big pile of sand. I wonder how long it will last?

Here is our first day and they've sculpted a new section to be filled with sand:

The sand they pump is very watery but the water just runs off. Bulldozers and backhoes push and dig up the sand as it comes in. Here they've just starting pumping:

Here is a closer section, ready to be filled in:

This is very far into the process of filling in a section, towards the end. In the back is the ship which is actually doing the pumping. I'm not sure where the sand is coming from, but I think its pretty far away.

The backhoe digs deep into the water to scoop out the sand:

The most interesting part is extending the pipe. Here a guy is bring in a new section of pipe:

The surprising thing is they add a new section while pumping sand. So the backhoe holds the end of the pipe with a chain, and just kind of pulls it against the current end where the sand is coming out:

He does this several times and keeps dropping it. Each time more sand gets added at the end of the new pipe. Here the backhoe is holding it and the bulldozer is moving the end around:

Eventually using the new sand they secure the new pipe section. The backhoe is still holding the new pipe though. They must run in and bolt it in place, but I never saw that part.

Finally I saw them tearing up the pipes. I thought they were going to open the new beach but they turned around and rebuilt the pipeline pretty much as before. They must have been adjusting the direction of the line or creating more secure connections.

Finally here is one of the workers. These guys were out there day and night in hot weather working pretty hard.

The Singularity

noreply@blogger.com (Philip) — Wed, 09 Dec 2009 20:37:00 +0000

Ray Kurzweil is an accomplished inventor and entrepreneur. He's had an amazingly productive career with multiple significant inventions and ventures in deeply technical fields. He started young. In 1963 at the age of 15 he wrote his first computer program.

In his recent book Outliers Malcolm Gladwell writes at great length about how unusual it was for Bill Gates to have access to a computer in the late 1960's, and how formative it was for him. So for Kurzweil to be writing software in 1963 was exceptional and it is understandable how this head start, combined with prodigious natural talent, could lead to a remarkable career. At the age of 17 he appeared on the game show "I've Got a Secret" demonstrating a machine he invented which could compose music:

He went on to enjoy several technical and commercial successes in the related but different fields of optical character recognition, speech synthesis, and speech recognition. He even returned to computer music in the early 1980's. This time he created line of advanced synthesizers which for the first time could fool listeners into thinking it was the real instrument. As with his other ventures he eventually sold the company, but the Kurzweil synthesizer brand remains and its products are still popular and respected today.

I really admire the depth and breadth of Kurzweil's technical achievements, and his success as a serial entrepreneur. So I was surprised when I learned about one of his more recent businesses: he sells vitamins online. Not just regular keep-your-skin-looking-healthy vitamins, but an extensive line of "longevity products." For example $65.06/month will get you the Anti-Aging MultiPack with ingredients like Resveratrol and Ubiquinol.

The vitamin selling is really an offshoot of yet another one of Kurzweil's careers, possibly the one is best known for today: he is a futurist. A futurist is anyone who makes predictions about the future. There are no particular qualifications to be a futurist, but to make a career of it you have to have an angle. Kurzweil's predictions are intertwined with a concept called The Singularity. The Singularity is a point in the future when technological progress accelerates so much that it outstrips anyone's ability to comprehend it. It is impossible to "see beyond" The Singularity, because humanity will be so radically transformed by it.

Sometime during The Singularity it will become possible to capture the complete state of a living person and "download" them into a computer. That is, the person will still exist, but not as a biological entity only as a pattern of bits in a computer. This is where the vitamins come in. Kurzweil believes The Singularity is close, that it will start around 2045. He also believes medical technology will advance rapidly, such that even a modest extension of your life span might be enough, depending on your age today, to keep you alive until The Singularity. And if you make it that far, you can in theory live forever.

This is only a thumbnail sketch of his ideas. Kurzweil goes into great depth about the reasons he believes this all will happen, as well details about what exactly will happen leading up to and even after The Singularity. He explains it all in several books as well as long articles on KurzweilAI.net.

So this is a pretty radical prediction, what do I think about it? For one, predicting that The Singularity will arrive about 35 years from now seems awfully suspicious. This puts immortality just within the reach of his core book and vitamin buying audiences. It's too neat of a pitch: an amazing transformation of all humanity is coming, act now and you too can be a part of it!

Even if he his completely genuine, his aggressive timetable is still the most unbelievable aspect of his predictions. If he were talking about thousands of years in the future, anything seems possible, but 2045 is just not that far away. Kurzweil calls this the "argument from incredulity" or the "I just don't believe it" reaction. He doesn't consider this to be a very good objection.

I guess I can refine my incredulity a bit. I just don't believe the guy making the most radical predictions is likely to be right. Kurzweil is saying 2045 but a lot of people don't think the Singularity will ever happen, they don't think computers will ever be smarter than humans for example. So what is the safest bet? I have to go with something in the middle. Some portion of his predictions will come true, but it will take a lot longer than he expects. The wisdom of crowds applied to futurism: the average is going to be more accurate than any single prediction.

Kurzweil's biggest contribution isn't even a prediction but a historical observation: The Law of Accelerating Returns. This is like Moore's Law on steroids. His claim is that all "information technology" improves on a smooth exponential curve. The doubling of disk drive capacity or CPU performance, the halving of various costs, the doubling of bandwidth. He is saying that all these phenomena are due to the same basic mechanism. And it isn't new. He shows the same trend lines going back to vacuum tubes, even the invention of fire in one case.

All of Kurzweil's technology predictions are based on applying The Law of Accelerating Returns, extrapolating trends which have already begun. Kurzweil makes many specific predictions most of the about what another author calls the GRIN technologies: genetics, robotics, information technology, and nanotechnology. He predicts these technologies will all flourish and start to converge, they will all become information technologies to some degree. He believes Artificial Intelligence will finally succeed, that computers will exhibit human level intelligence and soon after they will become far more intelligent than humans.

Even though I don't buy Kurzweil's timetable (or his vitamins!) I think there are a lot of good ideas in his predictions. This is a smart guy and he's given this a lot of thought. I consider his writing to be essentially Science Fiction. It is a form of Science Fiction which is heavy on predictions and light on story. We don't reject a Science Fiction author because some of his ideas seem crazy, we still consider them a worthy source of insight and entertainment.

I particularly like Kurzweil's Law of Accelerating Returns. I would like to write more about this in future, it is of great interest to me not whether this is happening, he hammers it home that it is happening, but exactly why it is happening. I think many of Kurzweil's other predictions make a great jumping off point for philosophical discussions: what exactly would it mean to be downloaded into a computer? I will more have to say about this also in a future post.

Abstractions

noreply@blogger.com (Philip) — Fri, 27 Nov 2009 21:30:00 +0000

In a previous post I referenced the short film Powers of Ten. The film visually zooms from the known universe down to sub-atomic particles. While the pace was quick it was glacial compared to how fast we can jump around with our thoughts. We can consider an atom, a ham sandwich, and a galaxy within two seconds. How can we move around so effortlessly?

Abstraction is a tool of thought. Abstraction is the process of ignoring some details in order to focus on others. We do this all the time. It is our main weapon against the complexity of the world. We jump from one abstraction to another depending on our immediate needs.

When I turn the key of my car in the morning, I am not thinking about the chemical reactions in the battery or the details of the combustion cycle. I just turn the key and listen for the sound of the car starting. This is my mental model of the car: you turn the key and the engine makes a starting sound. Now suppose the car doesn't start, I hear only a click. Then I adopt new abstractions. I start thinking about the electrical system, the starter, the battery.

One of the most persistent and convincing illusions which we enjoy is that of a continuous high resolution visual field. In reality we only see "high resolution" over an area about the size of a quarter held at arms length. This is our fovea and it is only 1% of the retinal area but maps to 50% of the visual cortex. That is, your brain is overwhelmingly occupied with this tiny portion of your view.

Suppose you are at a scenic overlook. You feel like you are seeing the full splendor and detail of the scene all at once. Instead your eyes are darting around from one detail to the next, and you assemble the full panorama only in your mind. You can witness this phenomenon by staring at a single word in this text. Now without moving your eyes try to see a word a few words away. If you can hold your eyes still you should find it very difficult to resolve details in these nearby words.

I believe abstraction works much like the visual system. We can think about complex objects like cars, but the details are mostly abstracted away at any point in time. Research has found that we can usually hold only 5-9 items in short term memory. I think abstractions tend to have about that many concrete details. When we need to consider something else, we quickly shift to a new abstraction to bring the necessary details to the forefront.

Writing software is all about creating and managing hierarchies of abstractions. In the real world as you dive down into subparts of objects you eventually hit bottom, quarks or something we do not know how to subdivide. In software there is no such limit. In theory you can pile abstractions to the moon. In practice however there are limits. Large software systems are notoriously hard to write and prone to errors.

Joel Spolsky a software developer and blogger has written about The Law of Leaky Abstractions. His claim is that all non-trivial abstractions "leak". The abstraction advertises that certain details are private, that you can ignore them, but those details actually do matter and cause problems.

One of his examples is the "remote filesystem". Your filesystem is a software layer that deals with files on disk, like those sitting on your hard drive. However sometimes files are not local, they are located across internet. Generally the approach is to hide this "detail" from the programmer. If you want to open a file it works identically wherever the file actually is. This approach can backfire. As Joel says the abstraction leaks because the location of the file really does matter.

Joel suggests there is no real solution to deal with these leaky abstractions, that dealing with them is just what makes programming hard work. I think he is mostly right, but I think there is a general way we can sometimes combat the problem. Remember abstraction happens in the mind of the observer. A software layer might suggest a particular abstraction, but it does not impose one. Instead, the person writing code is free to invent their own more productive abstractions.

Most filesystem interfaces are agnostic about how long operations should take. Any details about timing are abstracted away. But the code using the filesystem could time each operation itself and draw conclusions about the underlying physical setup. For example, if operations are taking only milliseconds it behaves one way but if they are taking multiple seconds it swaps in a completely abstraction, one where it considers every filesystem operation to be expensive and minimizes them.

In general I think the solution to leaky abstractions is more abstractions. We should not rigidly buy into whatever abstraction is implied by lower level code. We have to invent more and different abstractions of our own and fluidly switch among them, just like we do when thinking.

Seeking What's New

noreply@blogger.com (Philip) — Sat, 21 Nov 2009 00:57:00 +0000

I started taking the web seriously in the summer of 1994. Before that I had used the web, but I wasn't quite sure what to make of it. That summer I had a job with Unisys in Kendall Square, Cambridge. They put me in a cold machine room and told me to learn about the web. They didn't even have an internet connection! I had to use my roommate's university dial-up account. I downloaded whole copies of websites so I would have something to demo. I learned about browsers and HTML and web servers. At the end of the summer I gave a big presentation and said yes, you should continue to look into this thing.

Back at college I paid much closer attention to the web, and spent a lot more time online. I distinctly remember launching Netscape 0.9 for the first time. Netscape was free, but it was commercial software. It was incredibly slick. My first impression was watching it download 4 images at the same time. It was like magic, much faster and smoother than previous browsers. It felt like the beginning of something big.

It's amazing the web made such an impression because the content at the time was minimal. A primary activity then, as now, was simply searching for new sites. There were various indexes and "what's new" pages, but critical mass built around an index called Yahoo. It wasn't a company yet, it was a side-project by two Stanford grad students. I once sheepishly sent them some kudos and questions through their "feedback form" and received a gracious reply from Jerry Yang, the co-founder and later CEO of Yahoo. It was a simpler time.

For a while it was possible to scan through everything on Yahoo's "what's new" page every day. By the end of 1994 you could no longer keep up. There were just too many sites. Fast forward to today and we find a staggering amount of new information coming online. YouTube says they receive 20 new hours of video every minute. I've heard no similar figure on the number of new web pages, but no one can check out even a small fraction of them.

Two modern technologies for finding new information are RSS readers like Google Reader and collaborative filtering sites like Digg, reddit, or my favorite Hacker News. These two approaches work very differently. An RSS reader lets you "subscribe" to dozens or hundreds of websites and follow their ongoing updates from a single location. While collaborative filtering sites require no input ahead of time. Instead a community of users collectively generates an endless stream of new and interesting links you would never have found on your own.

So why all the emphasis on finding what's new? Its seems obvious that we'd prefer new information to old, but it's worth thinking about the reasons why. New information, entertainment or otherwise, is full of potential. You never know if that next article or YouTube video is going to be the coolest thing you've ever seen and the anticipation keeps you going. But why do we need to see what's new right now? After all if we waited a while it would still be new to us. The reason is that new information has social value, and this value is time-critical.

It is far more exciting to see a movie on opening night than towards the end of its run or much later on DVD. This applies whether the movie is great or a horrible bomb. Part of this is the excitement of the crowd, the actual experience of watching, but the difference lasts long after the movie ends. As soon as you leave the theater you hold information which has significant value. People who haven't seen the movie want to know if it was good, and people that have seen it want to talk about how great or horrible it was. Some of the excitement around the movie rubs off on you. This value fades over the coming weeks as new movies enter into our collective awareness.

This same social value applies to any media as well as all content on the internet. With the internet it's even more pronounced because you can send interesting content directly to your friends. So not only do you hold valuable information, you can share it as well. Sharing high quality information is great for your credibility while forwarding stale sites will lower your stock.

So does social value alone explain our obsession about what's new on the internet? The final piece comes to us from the world of neuroscience. They have found we have a hardwired "seeking" behavior, that our brains get into a state of "crazed excitement" and are continuously induced to seek more and more. Web surfers trolling for new sites is just one type of seeking. Bargain shoppers flipping doggedly through rack after rack of clothes are under the same spell. Gamblers at the slot machine or blackjack table are desperately seeking a payout, even if the amount of money at stake is small.

So are we prone to seek anything? Could we get stuck in our front yard, seeking blades of grass? This doesn't happen. The things we end up seeking do have genuine value. The new information, the discounted shirt, the money. Plus all three have social value. Bragging about your latest bargain or how much you won at the tables is a big part of the fun. What the neuroscience tells us is the brain magnifies this intrinsic value. It locks us into a feedback cycle where we seek with an intensity which is not really justified by what we are looking for.

So what is the future of "what's new" on the internet? We know the desire for new information is very real and very deep seated. Each individual contributes some to the demand. With more and more users online the relative advantage of having newer or better information becomes larger over time. There is intense pressure to develop new technologies to better satisfy the demand. I don't know what the technologies will be, but however good they are one thing is sure... we will still want more.

The Human Scale

noreply@blogger.com (Philip) — Sun, 15 Nov 2009 13:53:00 +0000

This great interactive flash animation from the University of Utah lets you grab a slider and zoom from a coffee bean down to a carbon atom. This is clearly inspired by the classic "powers of ten" film by Charles and Ray Eames which also uses a carbon atom towards the small end of the scale.

The film's own website says it was made in 1977, but wikipedia says 1968. Whenever it was made the whole 9 minute film is on YouTube:

I would love for someone to do a flash version of the larger half of the scale, the solar system and galaxies. In both directions the animations are powerful because they reveal worlds which are simply not evident in the human-scale in which we live every day.

The spatial scale of the film and flash animation are one thing. The other axis worth considering is time. Just like space there are scales of time which are completely out of our unaided perception. We can no more experience events which are microseconds apart than we can directly see microbes. Similarly we might have GPS readings proving mountains are growing taller, but we can't directly perceive it even if we watched for a lifetime. We are locked into a certain human-scale relying on instruments or recordings for anything beyond.

All of this is why I love these time-lapse videos by Keith Loutit:

Beached from Keith Loutit on Vimeo.

Bathtub III from Keith Loutit on Vimeo.

Go to his channel on vimeo for more and to watch in HD. These are not just time-lapse, they are also tilt-shift, or at least an effect which resembles tilt-shift photography. Possibly the technique here is smallgantics, which is specific to video. However it was done, the end result is that it looks like little models rather than real people or objects.

Explanations of tilt-shift usually emphasize how the images are produced, but explain little about why we perceive them to be tiny models. I believe it comes down to the physics of lenses and relative scales. You might think viewing a detailed model of a house would be the same as viewing the full size house, if viewed from the same scaled distance. It sounds exactly equivalent. But you are not viewing from some mathematical point, our eyes and camera lenses have a certain physical size, and a related physical focal length. How these sizes compare to what we are viewing greatly affects the resulting images. Our brain knows the depth of field artifacts here could "only" come from viewing small objects. Except here we were tricked.

So tilt-shift and time-lapse complement each other, they mess with our mind's sense of both space and time. Both tell us this isn't normal human-scale, these look like tiny objects and they move like tiny objects. It trivializes everything, these are toys, they are funny. Of course the rub is they are completely real. That's why I think this succeeds as art: the videos make you see normal life very differently.

And to extrapolate further, what strikes me with these videos is how arbitrary the human-scale really is. Our eyes and cameras are a certain size, and our attention spans are a certain length, but are these universals? Perhaps some alien or artificial entity would perceive us more like these videos, and less like we subjectively experience life. And what would be the consequences of that?