It’s a simplified example of a broader principle. This structure shows how oscillation can emerge from a very simple system and how it can be remedied by adjustments of the delays. Oscillating systems are very much a thing and this explains why.
By the way, what you said about stock levels is correct, of course, but doesn’t necessarily avoid oscillations. See, all of these metrics are feedback-based as well. With a delay. How much history do we take into account? Are account lead times influenced by delivery times. This is a classic problem in system dynamics: where do we draw the system boundary? Since everything is connected to everything, the right thing to do is to model the entire universe. Since this is not practical, we draw a line _somewhere_. _Where_ exactly does not matter much as long as the resulting model is useful.
Mac user for 20+ years now. Helped my mother with two Windows laptops. One would not allow her to reset her pin. Nothing allowed me to do a safe boot, the option to enter the password was just not there. The other wouldn’t connect to her wifi. Switched to an admin account and, after login, am greeted with a black screen, a cursor and nothing else. Yesterday OneDrive just randomly would not sync a specific file. Just wouldn’t.
I could go on. That’s the network effect for you. The only reason I ever touch MS (or Meta or whatever) software is because someone I’m connected to cannot switch. And they can’t switch because others are in the same situation. At some point you don’t have to care any more. There’s no money in working software compared to igniting a massive feedback loop that just can’t stop regardless of if and what you do.
We had a lady down here in Estonia working for police. She used their extensive databases for non-work related reasons and got, very publicly, fired. She decided to appeal and lost. Through all levels of court up to the EU one. So for years, there would be a newspaper headline “remember that person who invaded your privacy? Still guilty”. Could not wish for a more impactful public execution.
It’s a sign of a badly designed system. In a well-designed one (have seen one, have helped to improve several) the drivers license database only has a personal identifier and the class of license identifier. Which is private information for sure but it would hardly be an end of the world if the feds learned I have the right to operate a motorcycle.
Perversely, this is the consequence of misguided privacy initiatives. Government agencies are discouraged or even forbidden from sharing data. Nobody has a reliable identifier. As a consequence, every single database out there will contain your name, dob, address and probably a photo. Oh, and because there’s no identity, they need to store a bunch more data to make sure you are not a fraudster. Also, in absence of a sensible electronic identity (because there’s no identity at all in the system), all of this data is used as a shared secret. Which makes it immediately commercially useful for identity theft, once obtained. And, of course, any of these giant piles of personal data can and will be breached in some way or another by both local and remote state actors.
It’s all because “identifiers are bad” and “no sharing information”. Are you guys satisfied with the privacy situation yet?
There’s one more aspect to this. Look at Claude self-reported availability figures. It’s two nines at best, not enough to build something serious on top of. Historically, adding another 9 is an order of magnitude cost shift. Even in best case scenario, where a LLM scales much differently, you are looking at significant cost increases just to achieve something people would be willing to have their business depend on.
You just need to go north. Most of Finland requires massive amounts of heating for most of the year for industrial and residential purposes. If the math doesn’t add up there, I really can’t see how math for orbital data centers would add up.
Why would you want to earn a billion dollars? You must have the maturity to handle the money. Hopefully you’ll mature sufficiently through the process of making that money but if the money’s fast, that might not happen. I’ve seen people get 9-figure rich overnight. Didn’t change them at all, they are still the wonderful folks they were. But I’ve also seen people drink themselves to death as it turns out a few mill in the bank does not answer the question “wtf do I do when I don’t have to do anything?”
Thus is the crime of the communist Russia: forcing millions into hard labor to die for progress yet squandering innovation for ideological reasons. But the same mechanism is there in, say, Microsoft. To get the attention of leadership, your idea must have 9 zeros at the very least. If it doesn’t, you either leave M$ or stay there and abandon your idea. But a 7-zero idea is a pretty expensive one to be abandoned.
True. But in the physical world, ideas (or memes, if you will) are bound to people and can only survive, if the carrier survives. The idea of curing headaches with striking your head with a hammer does not spread, because the carrier dies. On the interweb, that connection is no longer there. An idea is independent and can spread without direct human involvement. We are like a tundra species finding itself in a rainforest: completely unprepared, lacking both immunity and the ability to keep up with the environment as it changes.
Exactly. So is that level of obvious hygiene where the bar is or is it somewhere else. What ticks me off is the audacity of blanket claims without an attempt to even remotely state why it’s said this is a list of successful patterns and what does success mean. We’re just supposed to eat it up, because, you know, Claude.
How very interesting. In an industry, where things shift around in months if not weeks, there’s been not only enough time for clear patterns to emerge but also these patterns have proven successful on large codebases. What’s the success criteria? Didn’t delete production database? Team velocity has increased? Codebase TTL has increased? Operations guys are happier?
There used to be such a thing as profit. A return on investment. If your exit strategy is to get sold to Google, focusing on revenue is a perfectly fine strategy. If you _are_ the Google, however, the money poured in should eventually be made back. We seem to have forgotten that. The current level of commercialization just means the US is burning investments faster, than anyone else. Eventually this might change and the bet might pay off. But every minute this goes on, the expected payoff must be larger to pay for the loss made this minute as well as interest for the previous minutes. I’m not entirely sure this is what “winning” looks like. Tic-toc.
It’s a general principle. Complexity breeds complexity via a multitude of mechanisms thus growing exponentially. As complexity is also related to cost, this means, that there are two limits approached exponentially: financial (we can no longer afford complexity) and cognitive (ee can no longer understand). In an ideal world, financial barrier arrives first, as it is necessary to understand something to make it constructively simple. If it doesn’t, the only solution is destructive simplification by simply breaking the system into pieces forcefully. This is what Musk did to X and tried to do to the US Government.
Unlikely. There’s no change in operating profit per employee trends for major software companies like Alphabet since GenAI became a thing. But MS employees are now making 3 times more profit, than they were before Nadella took over. Clearly leadership can make a difference but there is no visible impact after several years of the technology being available. I can’t imagine a technology, that shows no economic impact at all while we figure it out. There ought to be _something_. Yes, big companies have inertia, but Nadella showed clear results in a year.
My kids walk/bike to school, take public transport and are pretty free-range. But that’s absolutely nothing compared to my childhood. I was born in 1975. We dug up literal explosives from WWII and made our own small bombs (some matches, two large bolts…). Access to all sorts of chemicals was easy and we would set things on fire or just mix stuff to see what happens. Playing around on construction sites. Taking someones boat out to go fishing. Making bows and arrows, that would go straight through plywood. All _that_ is definitely gone. Some of it for good - kids loosing their fingers or worse was a common occurrence. So there definitely is a trend in case of me and my friends.
Soviets did not have two things the West did. Concern for quality and market forces to direct development focus. This means Soviet stuff varies amazingly between specimen and can sometimes be over-engineered in particular ways. Soviet optics had a specific visual style, but everyone ditched them as soon as alternatives became available as hunting for the ones not made on a Monday was just too tedious.
An excellent analogy. Everyone is an expert in taking the photo. But this does not make them a photographer. Even that expert claim is actually not fully true, the phone camera is woefully inadequate in many ways. But the main difference between a photographer and a layman like myself is the ability to produce output strongly linked with clear artistic intent.
Writing code is not the hard part and never has been. The hard part is having a clear understanding of how to solve a specific complex problem and being able to express that intent in code. Getting a decently exposed image was never the hard part.
Finally, there’s no scaling issues with cameras. You just make them better until it stops making economic sense. This is not true with code. To make llms better, good human-made code is needed for training. Better llms lead to less human-made code being available. This means there’s not an exponential growth in quality but a S-curve with a balance point. I’d say we are already there: innovation is shifting from the models to the ways of harnessing the models.
By the way, what you said about stock levels is correct, of course, but doesn’t necessarily avoid oscillations. See, all of these metrics are feedback-based as well. With a delay. How much history do we take into account? Are account lead times influenced by delivery times. This is a classic problem in system dynamics: where do we draw the system boundary? Since everything is connected to everything, the right thing to do is to model the entire universe. Since this is not practical, we draw a line _somewhere_. _Where_ exactly does not matter much as long as the resulting model is useful.