> 2. Distractions galore - Social media and trillions poured into the distraction economy ensures the ADHD-prone builders have less hours and are less productive in that precious 5PM-10PM.
Not enough is said about this. It's almost comical when you think about it. As technologists we are both complicit and victims. I've spent half a decade in one of these 'attention economy' companies and let me tell you the amount of money, talent and resources that our industry deploys to forcefully grab and monetize users' attention is staggering.
Recently I've shifted to using single-use, fit-for-purpose devices (Kobo ereader hacked with KOReader, KingJim Pomera DM250 digital memo) for my day-to-day and it was like a weight that I never knew was there was magically lifted away. If capitalism could find a way to produce such devices at scale, not only would it be a public health win, it would be a massive boost to the economy long-term.
But with most corporation's incessant focus on short term metrics I'm not holding my breath that this will ever be a reality.
Indeed, Lasch was a prescient thinker and reading through Culture of Narcissism I am impressed at how accurate his predictions were. Chapter 3 in particular is almost painful to read.
> It’s okay to be the one who initiates. It seems like not everyone can, somehow the level of social anxiety has gone up in the world. In the end, I get my socialization full and so do they. So I’d recommend to whoever feels a bit lonely — reach out to your past friends and current acquaintances.
Agreed. Before Covid I used to have a pretty vibrant social life but I was the initiator and back then I could easily set up physical events. Covid obviously added a lot of friction to that. Now that we are half-way through the 2020s I have enough perspective to say the bad habits that a lot of folks developed during Covid have stuck and it's a shame.
> The only issue with always being the initiator is that no reciprocation is a bit of an… issue to our social brains. Validation, trust, confidence, and friendship itself forms better when there is reciprocation. It’s best not to overthink it, the world is different and what it means to have friends has changed. This is the new normal. It’s better to be the one who leads all the friend groups and activities all the time than to be lonely.
How do you set boundaries?
It would be great if I could go back to how things were, but unfortunately I've changed. I was a lot more naive back then, and usually leaned into giving people the benefit of the doubt. It didn't help that the friendships / acquantainces I'd developed weren't exactly high quality.
In retrospect none of it was sustainable. All this happened when (1) I had lots of free time (2) Could physically meet with friends and (3) Hadn't suffered through betrayals from people I thought I was close to.
As much as it pains me to admit, I just don't have the emotional reserves to deal with one-sided interactions anymore. I would really love to hear from folks who have been in such a situation and have gotten over the hump.
I don’t, I just naturally hang out with the people who are most reciprocating and being the most good vibes.
Ultimately, it’s about me first. I function better and am happier when I have recently socialized. I’m not doing this for others, so it’s all good and I don’t feel like anyone particularly pushes me in a way where I’d need to maintain boundaries for myself in this particular context.
Regarding your three points, I hear them. But you probably have enough free time to text a friend, and sometimes physically meet up with them. I’d say, if you really don’t have time for that (and I was there myself), then probably you’re neglecting your social needs (basic socialising, safety net) for something else. Consider if that something else is worth it. Regarding point 3, I used to expect something reciprocal from friendships, but now I just expect to spend time with/among people. Yeah, most won’t help you in a time of need, and some will speak ill of you behind your back. But you’ll also meet many great people that legitimately will be great friends. So I’d say don’t worry about it — “trust the process”.
Overall, I hear some social anxiety in the 3rd point. Social interactions aren’t always ideal but don’t catastrophize them. Just do what’s good for you yourself first. Make sure your social needs are met. Let all other things and friendships develop or not as they would.
> It would be great if I could go back to how things were, but unfortunately I've changed.
I'm in agreement. I do think it would be a nice thing for friendship to work out, but I've been burned one too many times and the motivation just isn't there any longer.
I sometimes view this as a positive. I used to very much be a people-pleaser and thought that I was going to suffer and die if I didn't come out of one of my social outings with an acquaintance at some point. I was always told humans are tribal, we have a need to feel listened to, and not having friends leads to premature death. In practice I was just forcing myself to socialize based on that doomerism and that rubbed off onto the people I met, so it wouldn't have helped anyone.
When I turned inward and chose to put my own needs in front of those of others, I did become more comfortable with being myself. That's a prerequisite to having healthy relationships anyway (though I still wouldn't say I have any). Since then I've had lots of great conversations with people I've met at outings and large gatherings. I sometimes have conversations that go on for hours about all sorts of topics I may or may not know about and they're satisfying in hindsight.
...But I don't feel like being friends with any of those people anymore. I just let them pass and cherish the moments we did have together. I decided that the only person I have the capacity to fight for is myself from now on.
Nobody said you had to have friends in order to have a source of socialization to stave off bad health outcomes. "Having friends" and "being a bit social occasionally" are two different beasts.
> Before Covid I used to have a pretty vibrant social life but I was the initiator and back then I could easily set up physical events. Covid obviously added a lot of friction to that. Now that we are half-way through the 2020s I have enough perspective to say the bad habits that a lot of folks developed during Covid have stuck and it's a shame.
Another thing that maybe people don't like to talk about so much: COVID outed a lot of really mentally unwell people who had prior to it managed to keep that part of their personality to themselves. COVID and stay-at-home brought out so much anti-social nastiness: Selfishness, anger, belligerence, rudeness, conspiracy theories, defiance, and just this sense of contrariness for the sake of contrariness. I got a text during the height of the pandemic from a (now former) friend saying "Dude, did you know there's a bar downtown that's ignoring stay-at-home and letting people in the back door? We should go!" Like, what the fuck, man, don't you see there's a deadly disease going around? What the hell is wrong with you? I think during that time we really saw a lot of "true selves" that we now can't unsee.
> To me, that feels like a failure of the deeper social system. I want to be loyal to the people I work for/with, not treat our relationship like a transaction that is socially acceptable to end at any minute.
Great comment. The confounding variable here is culture.
American cultural norms devalue stable relationships in favor of personal fulfillment and self-actualization.
It isn't like this everywhere. There's a reason why business culture is different in Asia. The underlying attitudes there regarding social norms and how people can relate to each other i.e. what's acceptable and not acceptable, are very different. As a result, commerce there is conducted differently as well. Richard Nisbett wrote a book that goes into detail on this topic [1]
I will not make a judgement on which approach is better, or tie it into economic metrics but the bottom line is that attitudes towards work such as this one are highly influenced by the underlying behavioral norms. Without acknowledging this I don't think you can have a productive conversation on the topic.
[1] The Geography of Thought: How Asians and Westerners Think Differently...and Why - Richard Nisbett
I would push back on the notion that this is purely a geographical phenomenon though, because in the very recent past, it was totally normal for Americans to work for the same company for decades. You could say that recent developments are just an inevitable consequence of underlying social attitudes, but I’m not so sure it’s that simple.
I'm exactly in the category of AI majors who are not familiar with numerical methods. Can you broadly explain where the gap in AI pedagogy is and how students can fill it?
The series of articles posted here are interesting and I plan to review them in more detail. But I'm concerned about what the "unknown-unknowns" are.
Sure. The original normalizing flows used a fixed number of layers. Someone at UToronto recognized that, as the number of layers gets very deep, this is essentially an ordinary differential equation (ODE). Why?
Suppose you have n residual layers that look like:
x_0 = input
x_{i+1} = x_i + f(x_i)
x_n = output
If you replace them with an infinite number of layers, and use "time" t instead of "layer" i, you get
x(t+dt) = x(t) + dt f(x(t)) <=> x'(t) = f(x, t)
so to find the output, you just need to solve an ODE. It gets better! The goal of normalizing flows is to "flow" your probability distribution from a normal distribution to some other (e.g. image) distribution. This is usually done by trying to maximize the probability the training images should show up, according to your model, i.e.
loss(model) = product model^{-1}(training image)
Notice how you need the model to be reversible, which is pretty annoying to implement in the finite-layer case, but with some pretty lenient assumptions is guaranteed to be true for an ODE. Also, when you're inverting the model, the probabilities will change according to the derivative; since you have more than one dimension, this means you need to calculate the determinant of the Jacobian for every layer, which is decently costly in the finite-layer case. There are some tricks that can bring this down to O(layer size^2) (Hutchinson++), but the ODE case is trivial to compute (just exp(trace)).
So, turning the model into an ODE makes it blazing fast, and since you can use any ODE solver, you can train at different levels of precision based on the learning rate (i.e. the real log canonical threshold from singular learning theory). I haven't seen any papers that do this exactly, but it's common to use rougher approximations at the beginning of training. Probably the best example of this is the company Liquid AI.
Finally, this all turns out to be very similar to diffusion models. Someone realized this, and combined the two ideas into flow-matching.
-----
This is one place it's super useful to know numerical methods, but here are a couple others:
1. Weight initialization --> need to know stability analysis
2. Convolutions --> the Winograd algorithm, which is similar to ideas in the FFT and quadrature
> Can you broadly explain where the gap in AI pedagogy is and how students can fill it?
Machine learning can be made much more effective and efficient, by first modeling the problem, and then optimizing that tailored representation. This is an alternative to throwing a bunch of layers of neurons, or copy pasting an architecture, and hoping something works out.
One of the most successful applications of ML is convolutional neural networks and is based on this principle. Image processing algorithms come from an optical theory which can be modeled with convolution - what if we used optimization to find those convolution kernels.
With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.
Also you need to know when a problem is NOT optimization - for example solving equations via the bisection method.
I wish I could upvote this twice. A strong product intuition with middling technical knowledge will take you a lot further than razor sharp technical skills.
Unfortunately technical knowledge is all the school system and the cottage industry of interview prep companies focus on. If you can't get an MBA later, 4 courses I would recommend every CS undergrad take are, accounting, corporate finance, economics and maybe theatre.
Right on the money. After musing on this topic for a while and drawing from a lifetime's worth of experiences I've come to the conclusion that the social environment in America is very hostile and does not encourage developing close relationships. The prevalence of these articles really surprises me and makes me wonder if the people who write them have ever experienced living in a different (warmer) culture where the default mode of interaction isn't transactional.
Sadly I haven't. Travel is expensive, the US is a big wide country, and other nations don't really want me hanging around unless I marry someone or drop a lot of money
As a non-American, I can confirm that I've noticed that Americans tend to develop lots of superficial friendships, but aren't very willing to open up, which keeps the relationship at arm's length. I remember I was really surprised when a coworker was telling me about how she had some condition, and then immediately apologized for telling me. I didn't understand why someone would apologize for opening up, but in the US that's considered oversharing, and is frowned upon.
I've installed KOReader in about 3 devices but I haven't yet cracked how to keep them all in sync - frankly, I don't know how the syncing process works.
I used to think that it syncs up both the books and the reading progress / metadata but when I tried doing it, it didn't seem to work that way. I would love to hear how folks have done this. It's about the only thing left that's keeping me from using KOReader more frequently on all my devices.
The progress sync is separate from how books get onto the various devices (at least as far as I know). Something OPDS capable works well to get the actual files across a network (or calibre via usb). For progress you can use their provided sync server or run your own (the original one gave me enough trouble to write my own implementation).
For those interested in a simple to use command line tool that accomplishes the same I've had success with percollate - https://github.com/danburzo/percollate
Not enough is said about this. It's almost comical when you think about it. As technologists we are both complicit and victims. I've spent half a decade in one of these 'attention economy' companies and let me tell you the amount of money, talent and resources that our industry deploys to forcefully grab and monetize users' attention is staggering.
Recently I've shifted to using single-use, fit-for-purpose devices (Kobo ereader hacked with KOReader, KingJim Pomera DM250 digital memo) for my day-to-day and it was like a weight that I never knew was there was magically lifted away. If capitalism could find a way to produce such devices at scale, not only would it be a public health win, it would be a massive boost to the economy long-term.
But with most corporation's incessant focus on short term metrics I'm not holding my breath that this will ever be a reality.
reply