Humans suck

Tags: other, randomthought, webdevelopment, real life.
By lucb1e on 2012-01-17 23:55:21 +0100

Lately I've been working on two 'private' projects. Both failed due to human imprecision.

Firstly, I thought of something which might identify which song is playing in your head. You can't play it back (Shazam won't work), and most people don't want to sing it in front of the microphone on websites like Midomi. My system would recognise the tap pattern from a song. You'd drum your fingers on the desk or whereever, and it was supposed to search in all records for one that matches your drumming pattern.
Before I really got far with it, I discovered the drumming pattern was too variable to ever work. I was already worried that a monotone (it would only measure time between taps, not tone heights) would not be enough to identify millions of songs in the end, but this really failed it.

Last week I've been busy with a script I called Traindoor. I was standing on the station in front of the train next to hundreds of other passengers who wanted to get in the train first. If I could just know where the doors would stop... But hey, that's calculable! I realised that if I would create an app on my phone by which I could measure the time between the first and second door of the train (as it drives into the station), and the second and third door, I could measure the speed and deceleration. A train is always 26.4 meters, here at least, so off I went. After just over an hour I had it fully functioning, only now I was home and didn't have any trains to practice on.

The next morning at the train station I tested it. And wow it actually got close to about 2-3 meters! Not accurate, but much better than random guessing yourself. And this was a first test.
Later that day, coming back from school, I tried again. It hardly worked this time. It pointed me in the right direction (it can tell you to go x meters towards the front of the train, or x meters to the back), but that was about it. It might as well have been plain luck. As that morning might have been.

I continued testing, but didn't really get anything conclusive. It seemed to work about half the time. Getting a GPS plot of the train's speed as it drives into the station until it stands still, it turns out that about a quarter of the time, the train decelerates in a stage-like manner. It would go from 30KM/h to 20, then go 20-19 for a while, then drop another 10KM/s, go on that speed for a while, until eventually it gets to a standstill.

Darnit, I can't measure it when the train decelerates randomly! But this was only about a quarter of the time, the accuracy was about 50%. The other 25% it turned out was just my fault. I can't press accurately enough.
I made a simulation on the computer to scale, using actual speeds, deceleration, space between doors... Everything was accurate and the train decelerated linearly like it would 75% of the time. Measuring about 10 times, it turns out the human error factor is huge. I got results ranging from spot on to 4 meters off. And that were the good cases. If you are paying a little less attention, the calculation is easily off by over 10 meters.

A good tip from Bart Erven was to include the 4th door for accuracy, but I'm not sure if I broke it or if it confirmed my inaccuracy. The result would vary from saying "move 2 meters to the front of the train" at the third door to "move 5 meters to the front of the train" at the fourth door. I gave up anyway, 3 doors is nearly 80 meters already. A train isn't infinitely long, and you aren't always on the back side of the platform.
Also, when the train goes still 40KM/h at first, you tend to press as accurately as you can. At the fourth door the train goes like half the speed as before, and you actually start doubting when you exactly pressed at the first door. Did I press right at the start of the door, in the middle or when it had nearly passed? Usually you tend to press too early. But I guess this is also a matter of practice.


To get back to the subject, humans are imprecise. Maybe it's time we get replaced. Quote: "Within thirty years, we will have the technological means to create superhuman intelligence. Shortly after, the human era will be ended." Vernor Vinge, 1993.

(No, actually I don't honestly think we should be replaced. But that's human instinct, we want to preserve ourselves and our race.)
lucb1e.com
Another post tagged 'webdevelopment': What is XSS and how to protect your website

Look for more posts tagged other, randomthought, real life or webdevelopment.

Previous post - Next post