Hacker News Reader

Comment by marshfram | Hacker News Reader

Analog is next. Software first, then build the machines. No more models, reductions, loss. Direct perception through measurement and differences.

View on HN

Analog was before, though. General computing was never realized using those architectures; granted, they were mechanical in nature, so that is a big ask, both figuratively and literally.

Maybe we could create continuous-valued electrical computers, but at least state, stability and error detection are going to be giant hurdles. Also, programming GUIs from Gaussian splats sounds like fun in the negative sense.

You've just described vacuum tube computers as well as all the early transistorized computers. Digital computing is a relatively late concept

Of course there were analog 'computers' but vacuum tubes were also used to realize digital computers in the early days.

https://en.wikipedia.org/wiki/Vacuum-tube_computer

The important difference is that all those early analog computers were either bespoke or suited for a very narrow subset of tasks, like fire control computers. They were very far from general purpose computers, and that is the reason the von Neumann architecture dominates today: we are free to change the operation of the computer without literally changing gears or re-wiring all logic paths. Before, the hardware was the software.

You have to withdraw from the binary in all senses to begin to imagine what an analog spatial differences measurement could function as.

Again, think software first. The brain is always a byproduct of the processes, though it is discerned as a materialist operation.

Think big, binary computers are toys in the gran scheme of things.

We'd need a real breakthrough in physics to have such a technology that works at a scale even remotely comparable to what a low end digital CPU can do today. The thing is, there's not even any real evidence (at least to my knowledge) that there are useful threads that researchers know to pull on that could yield such a technology. Emulating analog hardware with digital hardware in anticipation of some kind of breakthrough isn't going to have any material benefits in the short to medium term.

You're thinking small, think about software first. Analoga of differentials. Spatial. Specifics. If you get trapped in emulation, you're trapped in counting.

What does that even mean

Binary computers are toy prisons of counting that model in a reality that has no correct models, not even neuromorphic will gain validity here, leave them behind. Start over with the forms we know are valid in neural syntax and apply them to software.

That's a great sentiment but without code that does what you're describing, that means nothing. Put some code out into the world that demonstrates the idea, because otherwise it's just inspirational words without substance.

It's not a sentiment, it's real coding. Rhythms of the Brain by Buzsaki. Cinema is already a prototype of differences, using space, and coreelating topological states. It's analog software that's been running for 130 years.

Binary was always a dead end alley, we knew this going in. How do we escape it?

It's not real coding if you can't point to code and show people how to do it. Where's the GitHub link? Where are the tutorials? Where are the YouTube videos explaining why it's important and how to do it right? You're just saying words, show me something tangible that's not just a vibe.

Closed to public view.

Sure, of course.

You can start here, sports and cinema combined create the syntax of language as a code.

https://www.routledge.com/The-Constraints-Led-Approach-Princ...

Reply on news.ycombinator.com