Hacker News Reader

Comment by goatlover | Hacker News Reader

I'm not convinced that progress has continued to accelerate. Compare the past 70 years to the previous 70 years and ask yourself which period of time experienced more radical progress, upheaval, and scientific discovery. My suspicion is that there was more from 1877-1947 than from 1947 till now.

View on HN

You're looking at the impact on everyday life, which has nothing to do with the abstract advancement of technology. Superintelligence only requires continued exponential growth of computing power, which has documented consistency.

Not over the last 10 years vs the prior 10 years.

https://en.wikipedia.org/wiki/Logistic_function / Sigmoid function's look exponential, but there is no longer a lot of room at the bottom. When things where 10,000 atoms wide dropping that to 1,000 was no big deal yet massive speedup. When they hit 5 atoms wide we what use 0.5 atoms?

PS: Arrays of low power processors like GPU's are still making progress, but we are rapidly approaching the point where there is more than one core per pixel. We might be able to use 1,000 cores per pixel, but I doubt it. EX: At 1080p SLI GTX 1080 Ti = 289 pixel's per CUDA core. Sure, 4/8k will push this off for a while, but monitor resolution has not been going up all that fast.

Why are pixels relevant to a discussion on computing power?

Pointing out that add more cores is yet another wall because FLOPS is only a proxy and few things are truly embarrassingly parallel when your start talking 1+ billion cores.

And yet here we are, made of neurons.

An GTX 1080 Ti has 12 billion transistors which is closer analogy to the 100 billion neurons in a human than saying we have 100 billion cores.

Furthermore, adding cores is not an exponentially-growing process. On the other hand, there is some evidence, from the fact that intelligence is possible within the human brain, that intelligence is embarrassingly parallel.

I think neurons are closer to transistor analogs than core analogs. The brain wires them up into networks that have meaning vs passing along data that has meaning. And in that context we crossed the 10 billion transistor on a chip milestone recently vs 100 billion neurons in a human brain.

It also requires learning how to create software that has general intelligence. Some people think that current techniques will be sufficient, but that remains a conjecture at this point.

The bipolar junction transistor was invented in 1948, so I have a hard time going with you here.

Electrical applications were developed during the late 19th and early 20th century, so I don't have a hard time going there. And that's just one of many areas.

Power, and a few signal, applications were developed starting in the mid- to late 19th century, sure. But it strikes me as extremely difficult to argue other than that the development of solid-state electronics has vastly accelerated every field of human endeavor with which it is coincident - and, as we've broadened the range of its applications, that set of fields has turned out to be most of them.

In any case I would be very interested in hearing more about a historiography which holds otherwise!

By mid to late 19th century you mean to include 20th century up to 1947 including television & vacuum computers, right?

The combustion engine, powered flight, QM, Relativity, modern warfare (including the 2 World Wars that reshaped a lot of the world), commercial refrigeration, widespread use of vaccinations and penicillin, atomic weaponry, and nitrogen for fertilizer are pretty big deals.

The question to ask is which time traveller would have more difficulty understanding the world. The one from 1877 transported to 1947, or the one transported from 1947 to 2017?

That depends where the time traveler landed. In 1947, there were still a lot of places in earth that were unchanged from 1877. Now pretry much every place on earth has power and a cell network.

You know what the GP meant.

And even today, if you take a random poke at Eurasia map, there's an overwhelming chance to end up in the middle of Siberian forest with no cell reception.

From 1877 - 1947, the following became an everyday reality : the telephone, television, cars, flight, the light bulb, flush toilets, skyscrapers.

The computer age has changed things in a big way, sure, but it's pretty hard to overstate how profoundly many aspects of human life changed in that period.

There was a Planet Money episode recently about this exact topic : http://www.npr.org/sections/money/2017/05/19/529178937/episo...

Exactly, most things in use today had a long history before they became popular, then they entered a short burst of activity where they become smaller and cheaper, after which they enter a very long plateau again. These bursts create the illusion of ever increasing acceleration, but they happened in the past as well, and just as quickly, and the majority of them have been forgotten just as the majority of 'startup innovations' will be forgotten in 20 years.

I wouldn't even say there was a burst of progress. There was simply a long time of incremental progress before the respective technology crossed over into practical viability. It's only at that point that people generally start to notice.

Progress is fractal, by definition. Technically the rate of change possible is limited by the number of edges on the fractal and the number of people who can work on it.

Reply on news.ycombinator.com