Monday, February 21, 2011
If you are concerned about the future of the world and whether we will avoid the many catastrophes that loom, try this bit of sunny optimism on for size. Cheered me right up! Of course, your results may vary.
Clown with a horn!
He should know that in science extrapolation is not always justifiable.
Besides, he is completely and provably wrong or disingenuous. Moore's Law regarding computer hardware has proven true due to incredible and ingenious technological improvements. The growth of solar is not due to technological improvement alone. There has not been a doubling of technological capability every two years as in computing. Only massive subsidy has ensured a rapid doubling of solar installations/capacity.
Ray needs to take the red nose and big shoes off and park his little car. The Guardian needs to learn journalism, something it hasn't practised in years.
No, he'll prove partially right. The commenters are correct that a nice linear flow is uh, unlikely, but energy sources are problems where we only need the occasional breakthrough interspersed with gradual improvements. Some things only have to be solved once.
What Kurzweil leaves out are the unintended consequences, and how humans will respond to the ever more rapid dislocations of technology making jobs obsolete. People don't eat in the long run, they eat every day. Boom and bust are very satisfying to look at in historical perspective, watching the Industrial Revolution make us all better off. But it was hell to live through.
Moore's Law has basically ended. We've hit the wall on transistor technology. All of the work going on to find more increases in computational performance are coming from parallelism (increasing the number of CPUs), rather than increases in performance (the power of a single CPU). CPU clock rates have hovered unchanged, around the 2 to 3 GHz range, for years.
The "wall", to be specific, is the power wall. We actually can run CPUs a bit faster (up to about 5 to 6 GHz), but the power output increases *quadratically*, meaning that a 2x increase in power becomes a 4x increase in heat output, so you need heroic cooling systems (liquid nitrogen). Such cooling systems are not practical for laptops or home computers, and they are not even practical for servers (data centers). Everyone in the industry (and I work in the industry) is scrambling to make advances in parallel processing ("multi-core").
Kurzweil is a smart guy, but he's far from infallible. His notion of the "singularity" is more of a religious vision, than any rational, useful way to see the world. We have a saying -- the Singularity is the Rapture for nerds.
There certainly will be advances in the performance-to-price ratio. In other words, as time advances, you'll be able to buy more computing capacity for less money. However, from here on out it is going to be a far more gradual change -- we are not going to see the same amazing "doubles every 18 months" phenomenon that we have seen for the last 30 years. Not unless we make a radical, and I mean *deep down* radical, discovery, like an entirely new transistor chemistry. It *could* happen, but it's far from inevitable, and a lot of smart chemistry/physics-type people have worked on exactly that for 30 years. So don't hold your breath.
Moore's Law was not based on clock speed but on the continuing reduction of transistor feature size. In the late '80s it was approaching one micron (one millionth of a meter) and they said it would soon dead-end. Well, now they are down to less than 30 nanometers (one *billionth* of a meter!) and it's still going, though the road ahead is getting harder.
In any case though, this has nothing to do with solar power. Moore's Law was meant specifically for transistors as digital switching elements and does not apply much to the analog (real) world.
"Moore's Law was not based on clock speed but on continuing reduction of transistor feature size."
True, but that doesn't change the fact that we've basically reached the end of the dramatic increases in *current* process technology. We can't decrease feature size much, again due to heat output. The smaller the transistor is, the larger, percentage-wise, the leakage current is. So even if we hold the leakage current constant, per transistor, if we increase the number of transistors by 50%, then we increase the leakage current by 50%. Good old "power = current squared x resistance" tells you that heat output will increase not by 50%, but by 125%. (1.5*1.5 = 2.25) Again, we have reached the *heat* wall.
Most of the easy gains that we will see in the next decade will come from market/scale improvements (more fab plants, more efficient fab plants), and from incremental improvements to transistor technology. The only potential "big" improvements will come from a significantly different computational model -- parallel computing. And that's a hard, unsolved problem.
That is, unless some bright lads come up with an entirely new transistor technology. What fun THAT would be!
Computing doesn't have that much to do with solar energy, does it? You have to be careful when you try to reason by metaphor.
The problem with solar energy is that not that much energy hits a square meter on Earth in a given period of time. That's a significant constraint, and I don't know that there's any practical way to work around that without ruining a lot of land at great expense.
There's a lot more potential in natural gas in the short run, and thorium reactors and/or little uranium nukes in the long-run. That's if we really cared about getting a real answer.
Moore discounts the political environment. He thinks it won't make any difference. He must know that green groups are successfully challenging solar projects in court and winning. And there are always friendly people in Congress who will agree with constituents who don't want nasty power lines spoiling their views, or nasty solar arrays blocking the sun from reaching flora and fauna on the ground. It's going to take a huge change to turn this around. Solar panels in space? Pu-leeze.
South Park had a show about a personal methane converter that an individual could wear at all times. (The show has a thing about the various permutations of anal probes.) Now there's an idea that makes sense.