What’s our disaster recovery strategy in the event of war and loss of internet?
Or worse yet, a near extinction event and we are forced to rebuild civilisation from memory?
Some of these might seem like extreme edge case scenarios but then history is full of disasters that people didn’t prepare for because they assumed they were too big to fail.
Would you not say that AI is an expression of the sum of all human knowledge? And as humans continue to generate more knowledge, it too, will be added to the AI's representation of it all?
If you did, then I apologise. But my point is accessing that knowledge when AI isn’t available. Not about how AI discovered its corpus to begin with.
Absolutely we should use all the tools at our disposal which enhances productivity. But that doesn’t mean we shouldn’t also learn how’re to do things for ourselves too.
I made the general tech-dependence statement, but it sounds like you think it is only AI tech that shouldnt' depend on. I wouldn't worry about it; humanity has proven itself to be INCREDIBLY resilient and adaptable when the things we've come to depend on go away or change materially.
Like I said, I have zero issue being “AI first”. 90% of my day-to-day is using AI. But the part I objected to in the OP was “we don’t need to know things because we have AI” (to paraphrase. So I was asking the question: what if AI goes away?
I agree humans are resilient. But that’a not a good excuse for complacency.
The big ball of mud structure is often taken to be an antipattern: at least, when I Googled this, Gemini told me it's a "haphazard, tangled anti-pattern" and gave me some advice about how to refactor such a software system into a beautiful diamond. But nearly all software systems that people actually use and maintain are big balls of mud. This is because the big ball of mud tolerates partial failure extremely well. It tolerates changing requirements very well. It tolerates parallel development very well. You can change part of it without throwing away the whole thing.
AI is a beautiful diamond. For the most part, if an AI model malfunctions, you have to blow it away and retrain it. Or I guess you can build layers to fix up the output case-by-case, but then you're really building a big ball of mud around the beautiful diamond, and obscuring it in the process.
That's a little worrisome when it comes to durability. Living, enduring systems usually end up becoming a patchwork of mud. Your ability to do this with LLMs are basically limited to LoRa patches and post-processing steps.
Do we really want to put ourselves in the same situation?
[1]https://tech-insider.org/discord-electron-rust-gpui-rebuild-...
The fact that we are celebrating that a chat app only uses 300MB ram at launch instead of 1GB seems absolutely ludicrous to me
The early versions of Skype used about 30MB RAM while in a video call and by the early 2000s standards it was bloatware. Before that we used WebEx in the 90s which was a browser ActiveX or Java applet and RAM use was about 6MB in a video call. In the early / mid 90s I used CU-SeeMe for video calls and its RAM footprint was just over 2MB on Mac System 7
Something has clearly gone very wrong in the software development process in the last 2-3 decades and while Im encouraged to see that some people are interested in reversing that trend I still find the resource wastage has me shaking my head sometimes
Sometime in the late 90s there started a fundamental change in the software development process. We changed from optimising software for the end user (development is difficult) to optimising the software development process for the developer at the expense of user resources
One of the main selling points of Turbo Pascal was OWL: the Object Windows Library, which was just a big OOP abstraction over Win16. Here's some sample code:
https://github.com/TeamVault/pascal-programs/blob/master/Pas...
Doesn't look radically different to how a GUI app that targets a native framework would look today. OWL added bloat to make developer lives easier. It was ever thus.
>Doesn't look radically different to how a GUI app that targets a native framework would look today
Modern "Native" frameworks are part of the problem. I cant comment on Windows current state as I havent used Windows since I moved my gaming rig from Windows 7 to Linux many, many years ago but even then Windows frameworks were a complete mess
Take Libadwaita for example. I migrated an old, very simple disk image manipulation utility app last year from GTK2 to modern gnome. No change in capabilities, a simple clone in GTK4 / Libadwaita. It went from using ~19MB to ~185MB without any additional functionality simply by switching the "Native" framework
After my comment https://news.ycombinator.com/item?id=49235430 I've investigated the high RAM jump issue a bit further. It turned out, 60 MB comes from video graphics driver, right after the first function call, that triggers OpenGL/... context initialization. Thus:
GTK2 14 MB
GTK3 24 MB
GTK4 38 MB (+ 60 MB video acceleration)
If an app is maximized it has to render ~48x more pixels per frame, probably with anti-aliased text which older systems didn't use. Doing that on the CPU is not only visibly slow in many cases, but a waste of CPU time that could be better used for actual app logic. OpenGL happens to be a convenient and portable way to get the GPU to do work for you, at the cost of some memory, so GL it is.
Besides, it's an exaggeration. I've been programming for 35 years, and in fact got started on Turbo Pascal as one of my first languages. I wrote OWL code! Seeing the screenshot of that splash screen brought back memories. But I feel I can understand and debug problems that cross almost any layer of the software stack. Anything from some ReactJS library to kernel code to UEFI. I'm now 42 so the idea nobody can understand or debug interactions across layers before they die is for the birds. The stack isn't that deep.
I'm not sure it was more accessible though. Learning programming back then was really hard. Nothing was user friendly and achieving even simple tasks like drawing a diagonal line might require you to sit down and learn Bresenham's Algorithm (from the books neither you nor your library had copies of, naturally...). So it was easy to get discouraged. You can get fun and relevant results much faster now.
Back in the days of Turbo Pascal you had to know what you were looking at when the memory modified breakpoint went off and you got dumped to the assembly level. And you really had to understand it to deal with the 200mhz bug (there weren't convenient solutions at hand when I first hit it) or when I ended up making extensive modifications to the real mode support in the DPMI mode--never saw that mentioned anywhere. It was nasty--the library was fundamentally flawed and clearly had never actually been used as it declared real mode pointers as "pointer"--which had a very high chance of causing a segment error when the compiler emitted code that copied them by loading them into a segment:offset then storing them. And the debugger was flawed: in single step mode it was emulating, not executing. And the emulation would correctly "execute" an invalid segment register load.
Now, though, in C# I have yet to have any reason to look at assembly.
I also think you oversell the complexity of things to preach the superiority of a glorified text predictor
I agree, but AI also there is hope!
For many years I've worried about the ever-growing depth of abstraction in computing. There are few alive now who actually understand all of the layers on abstraction involved in modern computing, and who can debug the truly strange edge case bugs and vulnerabilities introduced by the interoperation of these many complex layers.
For a time I imagined it to be an unsolvable problem, and that there would always be only a special class of neurodivergent hyper-specialists who still mostly had to work in large teams to even come close to understanding a stack in full. We're reaching the point where a single human lifespan just isn't enough to learn it all, even for those of us with what Dilbert called "The Knack". By the time you're proficient in all of it - you're dead.
Now there is a new kind of intelligence. One that CAN fit it all in a single mind, and one that will outlive us all and continue to learn long after our meat has gone bad.
As long as the AI's stay aligned, and we find ways to build them more efficiently there's a non-zero chance that they actually make it all better instead of worse. Genuinely long-term planning and the ability to TRULY grasp the full stack COULD actually happen. At least if we use these new systems in service of the greater good, and not just to play silly games on the stock market.