The Law of Accelerating Returns

[andy]

The Law of Accelerating Returns
by Ray Kurzweil (http://www.kurzweilai.net/bios/frame.html?main=/bios/bio0005.html)

An analysis of the history of technology shows that technological change is exponential, contrary to the common-sense "intuitive linear" view. So we won't experience 100 years of progress in the 21st century -- it will be more like 20,000 years of progress (at today's rate). The "returns," such as chip speed and cost-effectiveness, also increase exponentially. There's even exponential growth in the rate of exponential growth. Within a few decades, machine intelligence will surpass human intelligence, leading to The Singularity -- technological change so rapid and profound it represents a rupture in the fabric of human history. The implications include the merger of biological and nonbiological intelligence, immortal software-based humans, and ultra-high levels of intelligence that expand outward in the universe at the speed of light.

-------------

http://www.kurzweilai.net/meme/frame.html?main=/articles/art0134.html?

I'd love to hear your opinion! This is scary and interesting at the same time.

[grifscoots]

Open the door, Hal.

[andy]

Open the door, Hal.

Yeah, kinda like that, just more dramatic.

[norriscooper]

Still waiting for my copy of "Fembots for Dummies"

[cfell]

Yeah.. reminds me of the commentary by Jim Croce...

".... several settings... dice, frappe' and OH MY GOD HOW DO YOU TURN THIS THING OFF!!!!! "

[dwsdad]

There's even exponential growth in the rate of exponential growth. Within a few decades, machine intelligence will surpass human intelligence, leading to The Singularity -- technological change so rapid and profound it represents a rupture in the fabric of human history. The implications include the merger of biological and nonbiological intelligence, immortal software-based humans, and ultra-high levels of intelligence that expand outward in the universe at the speed of light.

-------------

http://www.kurzweilai.net/meme/frame.html?main=/articles/art0134.html?

I'd love to hear your opinion! This is scary and interesting at the same time.

That's really not much of a stretch. Just look at Washington.

Did you ever watch Battlestar? Zylons are a lot closer than we think.

[bob_menton]

Where do the idiots driving cars on the road today fit into Ray Kurzweil's image of the future ???

I went to undergrad engineering school with Lenny Kurzweil, Ray Kurzweil's brother. Lenny was a smart cookie, too. Never could beat him at ping-pong (oops - table tennis).

- Bob

[bonafidebob]

Where do the idiots driving cars on the road today fit into Ray Kurzweil's image of the future ???
They're Fast, Cheap, and Out of Control. :)

[andy]

Where do the idiots driving cars on the road today fit into Ray Kurzweil's image of the future ???

The way I see that is that the picture Ray painted is ONE of the possible ones, albeit it is based on logic and we might come mighty near to that. But at the same time just because we (as in mankind) can do it e.g. download your brain into a computer does not necessarily mean everybody or even a lot will do it. The political and sociological implications are something though!

[WildBil]

We will get more than 100 years of technology growth in the 21st century! I am just not so sure the acceleration will be exponential.

SInce 1971, computer hardware has grown very close to (Gordon) Moore's Law (http://en.wikipedia.org/wiki/Moore%27s_law). He co-founded Intel and postulated that chips / CPUs double about every 24 months in a very linear fashion, as they have from 1971 to 2007. CPU speed hit the wall at 4 GHz. Its not clear how the next doubling will be reached (http://www.cise.ufl.edu/%7Empf/physlim/calendar.html), as chip density is very near the physical limits of silicon (heat, wire vs insulator thickneess). Intel, AMD and Montalvo Systems have basically grew from 2003 on by filling the bottom side of the board with transistors, dual and now quad core chips, to keep within physical limits, but bus communication limits the growth there.

Other, more exotic semi-conductor media is expensive and only slightly better that silicon; quantum computers (16 states versus 2 states) may take until 2040 or so, if Fehnmann's quantum electrodynamics ever becomes practical. It may turn out to be exponential overall, but its likely to come in spurts. Then again, Moore may be still be right in 2100.

And the "better than linear growth" of software is ahead of the hardware curve, right now. There is no better example of this than Vista. It has a lot of potential, but the hardware to make it run as fast as XP has not arrived yet. And as we give more and more CPU cycles to anti-virus, anti-spyware, anti-whatever we seem to be losing ground. Nothing is immune to that, not Linux, Unix, map, or phones. There are over 1200 known bluetooth exploits.

Great gains are being made in spite of this, but nobody is close to exponential hardware / software improvement. The next leaps will probably be in the area of integration, like the iPhone. Its not really anything new. Phones have done texting, played music, had browsers, used 3G for some time. The iPhone is a revolutionary integration of tools in a very usable package. That may be where the major gains will come from in technology in the next decade, an perhaps that was what you meant.

I hope, and believe that somebody will break through the physical limits for chips, just like I hope somebody figures out a way to get to the other side or the universe and back while we can still see it. The consequence of the 1929 (http://en.wikipedia.org/wiki/Hubble%27s_law) discovery that every star is moving away from every other star is that they will eventually be so far away that light will never reach other stars, night will eventually be black and distant space travel is getting harder by the second. Not a near term worry, but almost 80 years later, nobody has found a way around that phyical law.

[bonafidebob]

I am just not so sure the acceleration will be exponential.
Doubling every 18 (or 24) months is exponential. Moore's law isn't about clock speed, but rather about number of transistors per unit of cost. That means more RAM in addition to more CPU. There's a similar exponential curve for storage over cost (hard drive space), and even one for number of web pages.

One place this exponential growth shows up is for example in ability to render graphics. It's not machine intelligence, but it's definitely something that is getting a lot better a lot faster, such that now games (and game consoles) that are only a few years old are obsolete.

Now if only the fun of playing the games would double every couple of years, we'd really have something!

[ranks]

Clock speed has hit a plataeu. There are no 6Ghz chips on the horizon.

What is the game of the foreseeable future is multi core and multi threads. AMD and Sun have the jump on this game. The problem is, we need OS and software that can take advantage of true parallelism, not just multiple processes. This is a long way out from having the tools an average developer can use to produce parallel code in a rapid/easy fashion.

It'll definitely be interesting to see what happens in the next five years.

[andy]

Even if we are not getting the computers faster and faster in terms of clock speed we are still capable to faster and faster by other means. E.g. the multi core chips, Look at SUN's chips, currently 8 cores on one die, and each of the cores runs with 8 threads equalling 64 CPU's of 2002 - 2003. I am still seeing Moore's law, just slightly modified, but it's there.

Now that said, Intel and other chip manufacturer are currently talking 64 or more core chips ... And this is only for next couple of years, not even talking about the next 10 or 20 years.

[WildBil]

If you graph it, doubling every two years is linear. About a 45 degree angle, but linear. If you graph the Intel processors since 71, its linear too. If you want to call it exponetial, go ahead. 0 is a valid exponent.

The only current 5 GHz chips are labware, run in coolant baths and last about half an hour.

And yes, Sun has packed 64 simple instruction set CPUs on a die. Gotta love technology taking a turn when it cannot go forward. No if Intel, AMD and Montalvo could do that with large instruction set chips and come up with a better signalling protocol they might not be stuck at 8. I am sure they will figure it out.

[bonafidebob]

If you graph it, doubling every two years is linear.
If you graph it on logarithmic paper maybe, otherwise it looks like this, and it's exponential. The exponent is years/2.

[andy]

I have to agree with Bob. Doubling every 1.8 years means a doubling from the then status quo, not from an imaginary 0 point.

[mantan]

And yes, Sun has packed 64 simple instruction set CPUs on a die. Gotta love technology taking a turn when it cannot go forward. No if Intel, AMD and Montalvo could do that with large instruction set chips and come up with a better signalling protocol they might not be stuck at 8. I am sure they will figure it out.

I am not sure if you know but Sun has bought Montalvo. Also, the current and upcoming multicore Sparc chips are not exactly simple instr set CPUs. BTW, I work in the Rock team at Sun. If you google "Sun Rock" you can see the public info on it. It has been fairly exciting 5yrs working on it.

[jayjacobson]

Can't wait until only the cockroaches are left....