,
16 tweets,
4 min read
Read on Twitter
2/ It typically takes about 100 milliseconds (ms) for a computer communication go through fiber optics across the ocean. It can be similarly fast for the megaconstellations in Low Earth Orbit. But it takes several hundred ms to go through a communication satellite in GEO.
3/ That extra time delay makes it difficult to play Smash Bros., or to teleoperate an aircraft, or to perform telesurgery on opposite sides of the globe. But yesterday I argued that ultimately we need gigantic geostationary satellites to keep up with projected data rates.
4/ I agree there will always be a need for low-latency communication, but really only a tiny amount of our data needs low latency. Blah blah blah...Most of this is boring, so here’s the weird part...
5/ Latency is relative; it depends on the data user’s psychology. I argued yesterday that information keeps exceeding its containers. Humanity is the latest container, but we are being exceeded. Machine intelligence will soon become the primary user of data communication.
6/ People have already calculated what is the biggest data rate a human brain could possibly process (full sensory telepresence...). Multiply that by 11 or 12 billion people, and that’s the maximum data communication our civilization could possibly need, right? Wrong. Because...
7/ ...by the end of the century or early next century there will be Trillions of equivalent human minds doing science, engineering, even literature, etc., existing in computers. They need not be organized like human minds or have any self awareness for this to be true.
8/ It might just be supercomputers crunching away on protein-folding problems, or computational fluid dynamics, like they already do. Or it might be trillions of simulated humans, because scanning & copying the human connectome might be possible by then: en.m.wikipedia.org/wiki/Connectome
9/ I don’t want to guess about this because I don’t need to, to make my point. Others have described possible futures dominated by non-human computer intelligence, e.g., @robinhanson “The Age of Em” (an interesting read!) amazon.com/Age-Em-Work-Ro…
10/ My point is that biologically human psychology won’t be the primary driver for latency requirements by the end of the century. Robin describes how some emulated humans may think 1000 times faster than others, because not every task is equally worthy of computational expense.
11/ Humans also have a giant latency built into our biological brains. We call it “sleep”, and it is when the processing apparatus does its maintenance operations. Computer intelligence based on human minds can simply skip that, and can manage the other latencies all day long.
12/ So in the future we’ll have layered data delivery, some faster, most slower. GEO satellites will carry most data. Fast data by fiber or LEO. It’s the same in shipping. Most items are transported across the ocean slowly on container ships. Urgent deliveries go by jet aircraft.
13/ A big question is why would we allow so many computers to crunch and transmit all that data, far more data than human minds could ever need? Why would we share our planet’s resources with machine intelligences? Let’s ban them to save the environment! I can relate, but...
14/ ...I am trying to be practical. There is already economic value in computer intelligence and we are unlikely to stop things that people value (and maybe we shouldn’t want to). It’s doing amazing things like crunching data to cure disease, and learning how to drive cars.
15/ It is becoming embedded in every part of our lives. I think it is far more likely that we can push all the heavy computing off the planet into space (to save the environment) than shut it down. And because we CAN push it into space, I think the future is exciting and bright.
16/16 So anyhow, I think it’s likely the data trend will continue, vastly exceeding human capacity, making latency irrelevant. Humans will be embedded in a grand civilization we can’t comprehend, like mitochondria can’t comprehend a human, doing amazing things we never guessed.
