"Preventing the Collapse of Civilisation" is an excellent talk by Jonathan Blow (of Braid fame) at DevGAMM 2019. It's inspiring and I recommend you make the time to see it. My notes below:

Executive Summary

• It is a myth that technology only advances. We've often seen significant backward slides in tech. (Lots of examples to illustrate this point)
• Software is in decline. Both software robustness and programmer productivity is declining. (Lots of examples to illustrate this point)
• We need to fight complexity and strive for simplicity in every step if we want to battle degradation and loss of capability.

Technology degrades over time

• Unless people make intentional effort to avoid this.
  • We have this idea of technology marching forward, but it goes backward all the time.
  • We've forgotten lots of tech over time when empires have fallen. (excellent examples in the talk)
  • Entire inter-connected thriving global system collapsed in 1177 BC. (end of the Bronze Age)
• It takes a lot of energy to communicate from generation to generation and there are losses inevitably.
  • This has happened to us with Microprocessors already. (Interview with Bob Colwell, Chief Microprocessor Architect, Intel illustrating this point)

Thesis: Software Robustness is in decline right now

• It's hard to recognize a slow collapse, the fall of the Roman Empire took 300 years.
• Software has been free riding on hardware, so it looks like we are doing okay.
• Software is having small local technological improvements (like machine learning) while suffering from degradation is other areas (like creating efficient apps)
• We don't expect software to work anymore, and this has been happening for so long that it's become the new normal.
  • "Just restart it!"
  • Jon: "I'm just going to note down / take a screenshot every time software I use has an obvious bug or unintuitive / incorrect behavior." ~ collected loads of examples in just a few days.
  • Try counting for yourself and you will be surprised by how many bugs there are, where the system just doesn't do the thing it's supposed to.
• If your laptop itself is less than 3 nines uptime, nothing running on it will be 5 nines. 5 nines was a standard that we were selling in the 90s, but it's completely infeasible today.
• If you haven't seen an entire industry produce robust software in a decade, what makes you think they can do it at all?

Thesis: Programmer Productivity is in decline right now

• Programmers: "Working on a higher level of abstraction lets me be smart! Save Effort!"
• You start out by being right and then you extrapolate it too far into the wrong territory.
• The flip side of higher lever of abstraction is the loss of capability altogether.
• Productivity per programmer is dropping, not increasing. This is easy to look at in terms of innovation in Products across multiple companies vs number of engineers they employ.
  • Interview with Ken Thompson talking about writing Unix in 3 weeks to illustrate this point.

If Robustness and Productivity are both declining, Software cannot be advancing

• "You cannot just"
  • Copy a program from one computer to another
    • The installer is for getting around incompatibilities added at the OS layer. The OS removes capabilities like compatibility.
  • Statically link
  • Draw a pixel on a screen
  • Write a shader
  • Compile a program without a manifest
  • Run an unsigned program
• All of these things add friction, bugs, engineering time, headspace. They keep us from thinking about and doing interesting things.

Complexity accelerates knowledge loss

• An example of complication: Language Server Protocol
  • A complicated, slow way to do libraries.
  • It's a hassle to write plugins (made up problem). So instead, just talk to a language server over a socket.
  • There is nothing special in looking up the location of X. It's just an API, and what you are saying is we should architect APIs like this. Which is crazy.
  • If you have to author something for this system, you now have to author and debug for a distributed system, where state is not located in any central place.
  • You are basically converting your fast system into a slow distributed debugging hell.
• Is there a point of reasonable complexity, and if so are we past it? How would we know?
  • Example of how complexity will impact the gaming industry: First there won't be people to make competing Game engines, then there won't be people good enough to maintain the existing systems.
  • In the bronze age, reading and writing were not widespread, so they were fragile.
    • Today, almost nobody knows what is happening on a CPU/GPU, that skill is fragile.
• Complication accelerates Knowledge Loss
  • More to know, so you only end up knowing a small percentage.
  • Deep knowledge replaced by trivia
  • Good information drowned by noise.
    • The harder it is to understand something, the more noise / bad answers around it.

Resources:

• Videos recommended by Jonathan Blow:
  • Casey Muratori: The Thirty Million Line Problem
  • Samo Burja: Civilization: Institutions, Knowledge and the Future
  • Eric Cline: 1177 BC: The Year Civilization Collapsed

Published On: Sun, 12 Apr 2020.