On Saturday, September 3rd, I participated in the 2011 Maui Channel Swim - a 10 mile open water race between Lanai and Maui: 

This is an annual organized event, swum by 20 solos (including myself) and ~70 relay teams of 6 people each. Each team (or solo swimmer) has a dedicated escort boat to guide them across the channel and to be there in case something goes wrong. My boat was captained by “Norm” and Caroline rode with him to support me. The water is in the high 70s, so no wet-suits are allowed. On a good day, the winning times are < 4 hours, and most teams finish within the cutoff of 8 hours. I was well-trained, well-focused, and pumped for the event.

But on race day, not everything went according to plan! Winds of 15-20 knots made for choppy conditions, even by 8am for the start of the race. Rough water makes the swim a bit harder, but the real concern is safety, primarily because boat drivers have a hard time seeing swimmers in choppy water. My own escort boat didn’t find me until I was a mile off shore. Here’s a video of me less than an hour into the swim, to give you an idea of what it was like.

Tragedy struck this year when a swimmer was hit by an escort boat near the finish line of the race, resulting in life-threatening injuries. This was really a shame, and I hope that either some policy can be put into place to prevent this type of accident from happening again, or that the event is just called off in the future. It makes the whole event not worth while. I hope his rehabilitation is as quick and smooth as possible, given the circumstances. In a completely separate incident, another one of the escort boats reportedly sank near the start of the race. Swimmers and supporters on board had to abandon ship, and thankfully other escort boats in the area saw what happened and took them on. According to the race organizers, there has only been one other safety incident (also involving a boat) in the race’s history. I didn’t witness any of these incidents first-hand, so I can’t speak to the causes directly. 

As you can see in the video above, Norm’s small (~15 ft) boat wasn’t able to ride in a straight line next to me as planned, and it was almost impossible to communicate with him over the sound of the wind, the white-caps, and the engine of the boat. Instead of navigating by staying near the boat, I was navigating by the sun, which was the only thing I could reliably see over the waves! ;-) Nevertheless, 3 miles into the race I was on course and on pace to finish in around 4 hours. 

And then I saw something in the water. At first I thought the faint circular shape underneath me was a jellyfish, so I took an abrupt stroke to the side to avoid it. It didn’t seem to move at all and I realized it was much farther away and much larger than I initially thought. Maybe it was a rock on the ocean floor - after all, visibility in the area was ~100 ft. As it came into focus I saw that it was a tiger shark, and it was heading right at me! The shark continued to swim right up at me and probably came within 10-20 ft from me before turning to the side. I could literally see his teeth! I sprinted for the escort boat, which at this point was ~150 ft away, yelling “shark!” with every breath. These were the most terrifying 30 seconds of my life!

Tiger sharks are fierce creatures. Adults are over 10 ft long and they weight as much as a car. According to the International Shark Attack File, tiger sharks are responsible for more attacks than any other species other than white sharks. And they’ll eat anything: car tires, license plates, even other sharks. 

I had so much adrenaline pumping by the time I got to the boat that I scratched both my knees and my elbow on the non-slip surface just trying to climb on - and I didn’t even notice, nor did I care. I was just thankful to be out of that water! It might as well have been fire as far as I was concerned! We radioed the other teams to let them know that we had seen a tiger shark and the teams nearby us all popped out of the water as well. I don’t know whether any of the other teams called off their swim, or if anyone else had a close encounter, but my day was done. There’s no way I was going back in shark-infested waters while bleeding! Norm advised against it as well, saying “you’re lucky you didn’t get tagged! Trust me, don’t get back in that water with those cuts.” Honestly though, even if I wasn’t bleeding, I don’t think I would have gotten back in. Here’s me on the escort boat back to Maui:

I was super let-down that I didn’t finish the swim, but looking back on it, it could have been a lot worse! Maybe I’ve become a bit too reckless. This experience makes me realize that crazy and fun challenges are great, but not worth dying for. Next time I’ll pick a challenge that’s a lot safer…and far away from any sharks.

While the war on terror is not over, recent events make this an appropriate time to look back at the sacrifices that have been made. Yes I support our troops, and yes I realize the plot below does not capture all of the sacrifices (coalition and civilian casualties, collateral damage, economic costs, etc). My point here is not to add my commentary, but rather for the numbers to speak for themselves:

On the drive home from work today I thought about the problems I face which can or cannot be solved with more money. In most situations, there is a clear and nearly continuous trade off. A little more money can buy me better food, a faster computer, or nicer clothes. Not all problems have this property though.

Most of the “monetarily inflexible” problems pertain to health. For example, no amount of money will keep me from ultimately dying. No amount of money can cure the flu any faster once it has been contracted. One particularly relevant exception to this trend is the commuting problem. No matter how much I am willing to pay, I cannot get from Pasadena to Burbank in less than 30 minutes during rush hour. Spending more money on a car could improve nearly any quality of that car except, paradoxically, the single most important one: how fast it will get me to where I am going! I suppose if I were absurdly wealthy, I could use a helicopter…but you must admit this problem is poorly-suited for a brute-force approach! 

I would like to propose that we fundamentally change the nature of the problem by introducing something I would like to call “speed lanes.” The idea is similar to a toll road, except what I envision is a lane (like the carpool lane) on most major freeways which is dedicated for a special use. Commuters who have preregistered their vehicles and received some identifying sensor will be able to use the “speed lane” for a price. That price during off-hours will be $0. However, once the speed lane reaches full occupancy, and traffic is about to cause a slow down, the fee for entering the speed lane will increase. This will cause certain drivers to leave the lane and it will remain jam-free. Using sensors in the road for feedback about occupancy in the speed lane, the prices can be adjusted in real-time and an equilibrium can be reached such that the speed lane maintains up-to-maximal occupancy and full speed 24/7.

This would be fantastic for a number of reasons! When you are late for an important event, no matter what time of day, you still have an option to get there quickly no matter what the traffic conditions are. How much would it be worth it to you to not miss an interview, or a presentation, or the start of a class, or a test, or a sporting event? Probably a lot more than the equilibrium price of the speed lane. This is an opportunity for value to be added! The speed lane would also provide a source of revenue for whatever government entity controls the roads. I don’t know what the equilibrium prices and occupancies would be, but intuitively I expect that it would be a significant fraction of the revenue brought in from gas taxes. Most importantly, people would feel more free. Although the speed lane necessarily won’t always be the right choice for you, it would be there for you if you need it. In your most urgent moment, you will not have to be a victim.

Do you ever hear someone say “is” twice in a row in a sentence that seems like it technically would have been fine with only one “is,” while paradoxically, it somehow also seems correct with both “is”s in there? e.g. “What the problem is is the lighting.” Today I was bothered by this so much that I decided to try to figure out why. I may not be 100% correct, but the following interpretation makes sense to me.

In all examples I’ve thought of so far, the “double is” phenomenon only occurs in use with a “noun clause.” That is, a group of several words that together play the role of a noun. In the example above, the bold words (I think) are a noun clause:

“What the problem is is the lighting.” - Together those 4 words act like a single noun (the subject in this case). 

Of course, the speaker could have just said:

“The problem is the lighting.” - This is also obviously correct.

Since the second option is so much more direct, and obviously correct, why does anyone ever choose the former? I don’t know - perhaps being less direct is comforting in a way because it makes it easier for the speaker to hide his or her true feelings about the matter - perhaps the extra syllables allow the speaker to indicate emotion more clearly - perhaps the meaning is different in a subtle way.

I want to point out one more example of the “double is” which bothers me even more, but for a completely different reason, e.g.:

“What we’re finding is is that certain elements have a different effect.”

In addition to the “double is” problem, this also bothers me from a logical stand-point - have you found it or have you found nothing!? At what stage of the scientific process could you possibly be for this statement to hold!? Whenever I hear someone say this, I feel like he or she is trying to imply something that they cannot prove and would like the listeners to take on faith without realizing the assumption they are making. It’s a cowardly technique to confuse the listeners into believing something they otherwise wouldn’t. It’s a statement made with the intent to deceive, and therefore, is the definition of a lie.

Anyway, the next time you say “is is,” this is what I will be thinking! ;-)

2 weeks ago I competed in the Superseal Olympic distance triathlon in Coronado, CA. I had a great experience and even surprised myself a little bit about how fast I was give that I hadn’t raced [tris] in almost 3 years. Results, including split times, were posted online a few days after the race, and this evening I took a moment to conduct some simple analysis. In all the plots that follow, red points are the “elite males,” green points are “elite females,” I am the one yellow point, and everyone else (excluding relays) is in blue.

The first thing I looked at was the correlation between each of the sports (swimming, biking, running) and the overall times for each person:

What you’ll notice in each graph is a strong correlation between the individual sport-time and the finish time. This should come as no surprise! Nevertheless, the correlation is not perfect. People who fall below the line (or towards the bottom of the distribution) in a given plot are disproportionately good at that sport. People who end up at the top of the distribution are disproportionately bad. By looking at the yellow dot above (me!) I can confirm what I already suspected - that I’m a baller at swimming and I suck on the bike! I guess 7 years of competitive swimming (compared to 0 for biking) makes a difference! In the title of each plot, I gave the correlation coefficients of each of the aggregate data sets (0.78, 0.91, 0.92 respectively). What I’d like you to notice here is that swimming is way less important to your overall time compared to the other two sports! This is the nature of international triathlons: A 20-minute swim is followed by an hour-plus bike - it’s anti-swimmer! ;-)

The next question I asked myself was, “what do the relative split times look like for fast people compared to slow people, overall?” The following 3 plots answer that question. They show the fraction of the overall time spent in each of the 3 sports for each athlete. Something jumps out here! Naively one might expect that slower athletes have proportionately slower times for each of the sports - but this is not the case! Examine the fraction of time spent biking compared to the overall time:

It appears that slower athletes have disproportionately *fast* biking times. So where does that extra time come from then? The answer is that those athletes are going much slower on the run:

…and their swim times are hardly affected:

In other words, slower athletes are disproportionately fast on the bike and slow on the run. Perhaps these athletes “hit the wall” or ran out of energy before the race was over and suffered during the run portion. Perhaps running is just a good indicator of overall physical conditioning. Perhaps the “fitness” required to run a little bit faster is super-linear with your current running speed, while biking is not. I admit I don’t fully understand this phenomenon yet.

One final Easter-egg to leave you is this plot of transition time (T1+T2) shown against finish time for each athlete:

Wow! This actually surprised me! It appears that the transition time correlates with overall time almost as well as the swim time does! ;-) Evidently, faster athletes have either more of a sense of urgency during transition, or invest time (or money) in making this part faster one way or another.

The following is the complete set of rules governing a system I created around the time I first learned to read, which determines a certain “flavor” for each word or phrase. The spelling of the word(s) in the phrase often dramatically changes the flavor. In that way, this system helped me to learn how to spell so well when I was young. That wasn’t the reason why I started doing this though, it’s just a side-effect. I do this every day even now, though not obsessively anymore.

According to this scheme, every word or phrase has exactly one natural break-point, determined only by its spelling (special characters, numbers, and spaces are ignored). The break point always occurs near the letter-wise half-way point of the phrase, but never at exactly the halfway point. For phrases with an odd number of letters, the break point will either be 1/2 letter before or 1/2 letter after the midpoint. For phrases with an even number of letters, the break point will either be 1 letter before or after the midpoint. For example, 5-letter phrases are either 2|3 (meaning the break comes after the second letter) breaks, or 3|2 breaks (meaning the break comes after the third letter). 8-letter phrases are either 3|5 breaks or 5|3 breaks. Breaks that occur before the halfway point are called “early breaks,” and breaks that occur after the half way point are “late breaks.”

To give a simple example, the word: “everyday” has 8 letters and must either be an early break: “eve|ryday” or a late break: “every|day.” The way to determine the outcome is to compare the letter that immediately precedes each of the possible breakpoints - in this case “e” and “y.” The following set of comparisons are applied to determine which letter will be chosen. If the first comparison yields a tie, then the second must be considered as a tie breaker and so on until the tie is eventually broken:

• Choose the letter than comes first in the alphabet (“e” comes before “y”).
• If the letters are the same and vowels, choose the one which in this instance is pronounced as the “hard” (a as in late) vowel sound over one that uses the “soft” (a as in apple) vowel sound and choose the soft vowel sound over any other use of that letter (a as in hard or board or alumni) and choose any sound made by that letter over one that is silent (e as in late).
• If the letters are consonants, or the vowels make similar sounds in their respective instances, choose the one that comes first in its word. For example, if the phrase is: “content neutral” which breaks either after the last “n” in “content” or after the “n” in “neutral”, this would be a late break (“content n|eutral”) because that “n” is the 1st letter in its word, while the other “n” is the 6th letter in its word.
• If the letters are both in the same position within their respective words, choose the early break.
• If the phrase is only one or two letters, there is no break.

At this point, you would be justified to say that this is pointless. Even so, I share this so that you will know that I am probably doing this in my head when you talk to me! 

Like a lot of people I spend most of my day in front of a computer, and the following UI “features” bother me so much I am compelled to write about them here. If you suffer from these same problems, I hope you can take solace in the fact that you do not suffer alone!

Scroll Bars in Windows

Open up a window with a scroll bar, but do not maximize it. Starting with the scroll bar at the top, hold and drag the scroll bar below the bottom of the window and then to the right (bringing the cursor outside of the window entirely). There! What the hell was that!?! The scroll bar *jumps* back up to wherever you started! If you release the mouse now, no command is interpreted by the scroll bar. This is really dumb because the most frequent time when this happens to me is when I accidentally drag the scroll bar a little too far when I wanted to scroll all the way to the bottom. In that case, the exact worst thing happens, and I find myself exclaiming, “How could you possibly think that’s what I wanted to do!?!”!

Cursor blinking

In Word and Outlook, the undisturbed cursor blinks on and off with a period of 1 second. That seems reasonable to me because without waiting more than 0.5 seconds, your eyes will detect motion and be redirected to where the cursor currently is. Could be better, but I’ll let that slide! However, 0.5 seconds of delay for no reason would be an atrocity! And that is exactly what happens *all the time* when you are typing in Word or Outlook. Open one of those programs and start typing some text. The cursor disappears (no idea who thought that would be a useful feature in the first place). What’s worse is that when you stop typing, the cursor starts blinking again….beginning from the OFF part of the cursor cycle instead of from the ON part of the cycle!! That means you have to wait 0.5 seconds just to see where your cursor is!! Why wait 0.5 seconds when it could be 0 milliseconds?? This problem is most cruel when you are trying to add or eliminate spaces from the end of a line of text.

Double-clicking

Finally, I would like to convince you that the invention of the double-click is one of the most costly mistakes in the history of UI designs. Clicking a single mouse button is nearly instantaneous. I would consider it a negligible fraction of human reaction time. However, performing a double-click is a much more difficult task.

Consider the algorithm for performing a single-click:

• Push down with index finger by some amount and then release. If you overshoot on either the pushing down or the releasing, no big deal, the click still works. Undershooting on the push down would result in failure, so you might as well just give it a little extra gas and guarantee that it gets done right.

In comparison, consider the algorithm for double-clicking:

• Push down, lift, push down, lift. The reason why you have to do all of these in one step is that human reaction time is so long that you can’t afford to wait for confirmation of each step (human reaction time x 4 = 0.8 seconds).
• If the mouse has been both pushed and released twice, you are done.
• However, if the mouse was not fully released between the two clicks, you are screwed. This is extremely common, since you need to have a very well-calibrated finger that knows how far down and up to go each time. Since either undershooting OR overshooting both result in failure (either by missing one of the clicks or failing to lift between the 2 clicks), you cannot employ the shortcut used in the single-click algorithm of “overshooting a little just to be safe.” 
• If you did only achieve one click, try another single-click as fast as you can and hope that the double-click operation has not timed out! 
• If you failed, try again.

Consequently, most of my single clicks sound like:

__—

And most of my double clicks sound like:

______—__—

…where the long line of underscores represents a push down, a failed lift, my reaction time to recognize the problem, and then the initial release.

Now you may be sitting over there acting smug saying that double-clicks still have their places where it is essential to make sure that the user deliberately intended that command - and you would have a point if that’s how double clicks were used. Unfortunately, that is not the case. For example, confirming a purchase on-line is an important action, and yet it is most frequently accomplished with a single click (famously in the case of Amazon). Meanwhile, playing a song in iTunes or opening any file in Windows Explorer is trivial and harmless, yet requires a double-click operation. This makes no sense to me. For these reasons I urge my fellow software developers to *swap* the usage of right clicks and double clicks in your applications compared to their traditional roles!

If you work at one of the companies responsible for these features, you owe it to the world to do what you can to resolve them (looking at you, Microsofties)!

My desktop computer has 8GB of RAM. That’s 68.7 billion bits of data that can be recalled at any instant for use in a calculation or for the synthesis of a new construct. By contrast, how much RAM does your brain have? Of course, there is no perfect analogue to RAM in the human brain – but the concept of storing data such that it can be immediately recalled and used to synthesize new results does have such an analogue.  To illustrate, imagine you are asked to perform the following calculation in your head:

31x9 = ?

After thinking for a moment you will probably find yourself among the majority population that does not have a cached result in long-term memory, so you’ll likely do something like the following:

31x9 = 30x9 + 1x9 = 10*3*9 +9 = 270 + 9 = 279. Yay!

The last two steps are made possible because most of us have previously memorized multiplication tables through 10. Overall, that was kind of hard actually, and required a decent amount of concentration! At one point, I had to store the numbers (10, 3, 9, 9), and the corresponding operators (*,*,+) simultaneously in my mind in order to synthesize the desired result. How much data is that really though? If each value takes 1 byte (assuming horrible efficiency) then this calculation would take a grand total of 7 bytes (a few billion times less than my computer can handle). Since you could do that in your head, you at least have that many bytes of “RAM” available. How about a slightly harder one:

4658*845 = (4000 + 600 + 50 + 8)*(800 + 40 + 5) = 4000*800 + 4000*40 + 4000*5 + 600*800 + 600*40 + 600*5 + 50*800 + 50*40 + 50*5 + 8*800 + 8*40 + 8*5 =

4*8*10^5 + 4*4*10^4 + 4*5*10^3 + … = 3936010.

To be honest, I cheated on this one and asked a computer to finish it for me. Nevertheless, at the worst point it would have only required ~20 bytes of simultaneous recall to finish this problem.

…

That’s pathetic! What about this claim that our brains are far more powerful than any supercomputer ever created!? This is a remarkable shortcoming if you stop and think about it for a moment. Our brains have apparently not been designed for high-level reasoning. Although we’re better than all the other animals as far as I can tell…we still suck at it. I am not exactly claiming that our brains only have access to <100 bytes of RAM at any time though. Obviously tasks like recognizing faces, or reading never-before-seen fonts, or interpreting facial cues and body language, or understanding what behavior is acceptable in social situations, or empathizing with a fellow human requires many orders of magnitude more. My claim though is that we cannot intentionally allocate any significant amount of memory for high-level tasks. It appears the memory allocated to our “consciousness thread” (to borrow another computer science analogy) is dismal! And this is the pertinent quantity for any problem to be solved via logical deduction.

But with this realization comes a work-around. Instead of attempting to do a few parallel syntheses of large amounts of data to solve high-level problems, we must instead re-cast our problems in order to take many syntheses of very small amounts of data. Although you may not have thought of it this way before, you use this principle all the time. For example, if someone asks you to add two very large numbers, you might struggle to do it in your head, but you can easily do it with a piece of paper. Using the technique of “carrying the numbers” you have transformed this synthesis of a large amount of data into a sequential set of syntheses of relatively small amounts of data. Now the maximally complex step in the process does not scale at all with the size of the numbers in the initial problem! This is a fantastic algorithm that addresses the exact shortcoming I pointed out above.

But implementing this technique is not possible without writing. When transforming a complex problem into a sequential set of simple problems, intermediate (and useless) data is inevitably generated. Storing this data in “RAM” defeats the purpose of most such algorithms and therefore it must be stored externally somehow: traditionally by writing it down (since there is no good way of quickly committing it to long-term memory or any other kind of memory for that matter).

The examples I’ve given above are trivial of course, though the meaning I intend to convey is that we should use analogues of the techniques I cited wherever possible if we intend to make strides towards achieving progress on the high-level problems we face. In other words, when you are stuck on a problem, write about it! The process of writing down the logical steps you have already taken will allow you to synthesize their corresponding results and continue to make progress in a way that is effortless compared to the struggle of holding it all in your “RAM” simultaneously. I hope you find this as insightful as I have, and try to use the written word to advance your thought processes as efficiently as possible. Let me know how it goes!

Memories are priceless. If not for memories, many of my life experiences would no longer exist. I could always keep a record in writing if I need to preserve the essential facts, but words can only capture one slice of an experience, so any such attempt is an ineffective substitute for a memory. Something as simple as a smell or a sound can instantly recreate a compelling array of feelings and images. That’s the power of memories.

But memories also have a dark side that troubles me deeply. I don’t expect you to offer a strategy for dealing with this problem, but I’m writing about my concern anyway for two reasons: I hope that I can alert you to this problem so that you can protect yourself, and I hope that articulating my problem will minimize its impact on me.

Memories have some undesirable properties. They fade over time, as you know - but more concerning is that their integrity can be accidentally jeopardized. In fact, I assert that anytime you “read” (i.e. think about) a memory, you necessarily corrupt and obscure it. These are 2 distinct and equally tragic problems.

Memory corruption occurs when the information within a memory is accidentally altered. One of my earliest memories is of a day when I ran (on foot) into a parked car and cut my lip. I’m aware of a great amount of detail about that day, though at this point, it’s very difficult to distinguish fact from fallacy, and what details I remembered myself, and which were pieced together later. Let’s imagine though, momentarily, that we could make these distinctions. Let’s also imagine that at one point I recalled that:

• The car that I hit was an old, white pickup truck.
• It was on a neighborhood road in Texas.
• I needed stitches later that day.

Please take a moment to visualize the scene before reading on.

…

…

In your vision, what time of day was it? Was it sunny or cloudy? What was behind the truck? Were there any trees, or grass, or people nearby? In my vision, it is a sunny day around noon, and the pickup truck is in front of a one-story house with a grassy lawn. My brain has filled in certain details, because it needs to - the scene makes no sense without them - certain values cannot be null! Then several months or years pass and something calls my attention to the fateful day in Texas, and I recall that:

• The car that I hit was a pickup truck in front of a one-story house.
• It was on a neighborhood road in Texas, on a sunny day around noon.
• I needed stitches later that day.

After a series of successive reads and writes, this memory will soon fail to resemble the original event at all! I will no longer be able to distinguish between the true events of the day, and the events that I have filled in to complete the story.

The other tragic shortcoming of memory is that it is obscured when it is read. The taste of Red Bull used to remind me of Fall and Winter of 2004 - my first few months at Caltech. This was the first time I ever had Red Bull, and I drank a lot of it to get through the 60 units (=20 semester-hours) I was taking at the time. For the next few years, anytime I drank a Red Bull, I would be reminded of that era. Now when I taste Red Bull, I’m reminded of drinking Red Bull mixed with Vodka in 2009-2010. The previous association is gone! The pathway to feelings of despair, exhaustion, and confusion that Red Bull used to provide for me is obscured forever! 

In the Spring of 2005 (which was a particularly happy time for me) I first experienced the smell of that small, white flower that blooms on trees here in Pasadena. I spent most of 2006-2007 in Hawaii, and when I returned to Pasadena in the Spring of 2008, those flowers were once again in bloom. They were like a time machine, taking me back 3 years in a second. But days go by! And they bloomed again in 2009 and again in 2010. Now, those flowers remind me of last March, when I sat sleepless in bed, sick with the flu. The pathway to my original (and very sweet) memory has been obscured! As time goes by, it’s increasingly difficult to experience the rush of emotions and images that used to be associated with those flowers.

I have been living with this realization for some time now, and it is depressing. I don’t mean to make you depressed, but I mean to alert you to the fact that when you read your memories, you are spending them! I cannot indulge in the smell of those flowers an infinite number of times and experience the powerful and realistic memories that they carry. It wears out over time. If you’re not careful, your memories will also die, and then they won’t exist anymore. So please, try your best to preserve the integrity of your memories when you read them, and be mindful of the cost involved.