I sometimes wonder what it would be like to learn your own life by heart. Let me explain what I mean with this weird expression. I've been keeping a diary for some time (since January) in which I just write a quick summary of what I did every day. In practice, I use this diary as a way to be able to find back, a few years from now, what I did on a specific day (it might prove useful in a somewhat bizarre set of circumstances), or as a way to re-read what took place quite some time ago (whenever I feel nostalgic, or just curious). But in fact, my deeper motive in writing this diary is as follows (it's quite simple, in fact): I find it really sad when I realize that I haven't the faintest idea of what took place in my day-to-day life a few years ago and that I have no way at all to find out. Of course, it's impossible to write down everything that took place, but just having something to start with (which often conjures more memories) is already a lot.

If you start to think about this kind of diary using computer science terminology, you could see such a diary as a way to sort memories using time as an index. Indeed, if you start searching for memories in your brain, you realize that you can easily find them thematically, but that it's nigh impossible to find them chronologically (except for the very recent past). My diary could serve as a way to circumvent this problem. (An external way, which I keep on my computer rather than in my brain.)

Which leads to my question: what would it be like to learn such a diary by heart, as it is written, and to be able to recite it like a poem? I'm not thinking about learning the exact wording of entries (although it could be a way to start things off), but to learn the succession of events, and to be able to remember one day after another, in order, over a long period (ideally, the whole of one's life).

The thing I find seductive with this idea is that it is deliciously meta. How natural and canonical to learn your own existence by heart! And how weird to remember not only real events, but also the act of memorizing real events, and the act of memorizing the act of memorizing, and so on...

If you follow this objective to its logical conclusion, you would spend most of your time thinking and trying to remember the sequence of your thoughts, and trying to remember that as well...

Even in a less perfect way, having some period of your life which you learnt by heart and can remember chronologically much later would be quite cool. I don't think I'll ever take the time to do it, but I wonder if someone already did something like that...

Related: hyperthymesia, the condition of having unusually good autobiographical memory.

Very related: Lembransation, a blog describing a project to remember every day in the blogger's life, spanning across a period of 10 years (last updated in 2022).

Just a quick rant about the fact that many users (including people who should know better) seem to have some sort of belief that computers (especially laptops) are not meant to stay powered on for extended periods of time, and that doing so isn't normal use and is likely to damage the hardware.

Needless to say, I don't agree at all with this view. In my opinion, non-faulty hardware should be able to withstand maximal load for arbitrarily long periods of time. To say things differently, if a computer fails under such circumstances, it isn't normal: it means that the computer isn't working properly.

It strikes me as odd that most people seem to consider the radio adjustment of watches as a neat and useful feature, and that no one seems to think about the security implications. What I mean is this: radio adjustment gives to an (untrusted) third party the power to adjust your watch at will. This third party could be malicious, but it may simply be incompetent: if the adjustment service broadcasts the wrong time, your watch will start displaying garbage. Worse: since this adjustment system cannot usually be deactivated, your watch will keep reverting to the wrong time until the broken adjustment system dies off for good.

I don't know of any real-world situation in which watches became useless because of something of this kind, but I don't find it that unlikely. In the meantime, I'll rather stick to watches I can control and adjust by myself and avoid unneeded interference from outside services...

I started a blog a while ago to report weird things I found in TOSs, EULAs and the like. I don't post on it that often, because most of these documents are just plain boring with nothing amusing in them whatsoever. It's nanoblogger, because fugitive didn't exist yet and because I just wanted to test it.

I never advertised its address anywhere, and I don't know where to do so, so here goes: Weird legalese.

Imagine a high school level maths exam in which students are requested to solve the following simple equation: x² - 3x + 2 = 0. How would you grade the two following answers:

1. We proved in class that such an equation has at most two solutions. [A few lines of maths in which the student computed them.] Therefore, the solutions to this equation are x = 1 and x = 2.
2. We proved in class that such an equation has at most two solutions. We check that 1² - 3×1 + 2 = 0 and 2² - 3×2 + 2 = 0. Therefore, the solutions to this equation are x = 1 and x = 2.

Many teachers dislike the second solution, because they think that the student must have cheated. Indeed, the solution seems highly suspicious, because we cannot guess where the solutions come from. (Alright, those were easy to find, but you get the idea.) Notice however that, from a logical standpoint, the second solution is perfectly valid: it justifies the fact that there exists at most two solutions, and makes them explicit in a perfectly rigorous fashion.

[Interestingly, I had at least one maths teacher who preferred solutions of the second type, because he only wanted us to prove that our answers were correct, and believed that the process leading to the solution was none of his business.]

The point I would like to draw attention to is the following: whereas both solutions prove what they state, only the first one explains where it comes from. Of course, there are more complex examples of this distinction: while some math papers, books and teachers make a great job of explaining things, others just prove theorems mechanically without any effort to show what's going on behind the scenes.

This distinction isn't limited to numerical solutions to equations either. Even with no equations involved, there are proofs in which you can see that things work out as planned but can't understand why they do; proofs in which you build ridiculously complicated concepts with which you unexpectedly manage to prove simpler claims; proofs in which, to put it simply, the author writes things but does not explain what gave him the idea to do things this way. Proofs which, in a way, look like a program with no comments.

The important fundamental difference between explaining and proving is, in my opinion, the following. As far as proving is concerned, you can, theoretically, write proofs which are undoubtedly correct, using only the axioms and core deduction rules. Of course, you never do that in practice, but it means that you could reach perfection if you wanted to. However, when you explain things, you have no reason to believe that it is possible to write something which is undoubtedly understandable. To do this, you would need to make each step seem natural, that is, not only do things, but also explain, at each step, what gave you the idea to do so, and why it gave you such an idea, and so on, and so forth. [Philosophically, the idea is that you cannot understand well enough what's going on in your head when you think to explain why you are thinking in this way...]

Keep in mind that this issue of writing proofs which explain things in addition to proving them should not be confused with the well-known problem of writing proofs with the right level of detail, ie. the tantalizing fact that when you prove something, justifying every step is infeasible in practice but just writing "Trivial from the axioms." isn't acceptable either so you have to find a compromise between these two extremes. The issue I am dealing with has nothing to do with this: the second answer in the example above could formally justify every logical deduction (even going back to the axioms) and still manage to cook up arbitrary values for x without explaining how they were found.

It is quite sad that many math books and courses seem to be written for machines rather than humans in that they prove things but don't really try to explain what they do. In my opinion, things would be better if the two were present and clearly distinguished, with a machine-readable formal proof, and an informal discussion explaining how the proof is built and why it is built that way.