Hacker News Reader

Comment by newyankee | Hacker News Reader

Not true, although he seems like he is doing crazy speculations, there is a method to his madness. It is driven by all the exponential growth laws which he documented and extrapolated very well.

View on HN

Exponential growth is not a law, nor is extrapolation a valid scientific technique.

"Not a valid scientific technique" is not a valid objection to a nonscientific process.

Kurzwel is making predictions. Extrapolation is like the only method of making predictions.

They can be, when used appropriately.

To give a particular example, the exponential growth of computing power has been a close-enough match to exponential since Kurzweil became famous: It was about 20 years ago that Apple was boasting that their latest PowerMac was banned from export to Cuba because a 1 gigaflops computer counted as a weapon, now they boast their phones do 11 teraflops — compatible with 1.5 year doubling time.

(That said: while I have met someone who expected the exponential to continue past single atom transistors, I do expect it to stop soon-ish; superconducting/3D/quantum computing may replace classical transistors, but I never expect to see multiple transistors on a single atom).

An atom can’t be a transistor. That’s like saying “a single atom banana”. An atom can encode spin state, which can be used for computation.

I don’t make any claim of being a material scientist, I just wish to note that there is something with the name “single atom transistor”: https://en.m.wikipedia.org/wiki/Single-atom_transistor

Maybe an atom can be something more powerful than a transistor and be able to emulate the properties of a transistor.

It is a "law" in the same sense as the MacDonald's Rule (countries that host MacDonald's franchises don't go to war.)

It works until it doesn't.

True, and I can say the same about Hooke’s law, Newton’s law of gravity, and so on.

Extrapolation outside your dataset is only a problem if you take it too far, and how far is “too far” depends on how well you modelled reality with your dataset in the first place.

Extrapolation, as the general act of generalizing from observed events to not-yet-observed events, is necessary for making any kind of prediction about the future, and for any awareness that goes beyond direct sensory perception. I think that's within the purview of science.

Moore's law, Wright's law may not be natural laws but have mostly proved to be consistently true observations.

Yeah, still not taking into account the energy we're using and the manpower this energy liberate.

But this energy is not infinite. Despite all innovations in solar/wind, and assuming efficient storage that does not exist yet, EU won't be able to absorb a 3% decrease in oil+gas importations by 2030. Rystad (https://www.rystadenergy.com/), pre-covid, was forecasting a 2 to 8% decrease in oil and gas import from the EU according to the think tank "the shift project" [0][1].

I won't talk about manpower much because there is still a chance that nothing more heavy than Covid will hit us in the near future, but meals will be hard to get this year in eastern Africa and while this year India did not suffer as much as in 2019, another drought like 2019 next year would be even more disastrous. For information, this winter could have been worse for Africa than 2011, luckily the wildfires and drought in russia were localized further north than a decade ago.

But well, we still have two silver bullets: fusion and renewable hydrolyse.

[0] https://theshiftproject.org/wp-content/uploads/2020/06/Study... complete data analysis

[1] https://theshiftproject.org/wp-content/uploads/2020/06/fig40... short graph.

In the real world, any period of exponential growth is part of an S-curve (coming to a plateau) or a bell curve (reaching a maximum, then decreasing) of some sort. You can't cheat with the law of physics and entropy.

I don’t think Moore’s law are laws in traditional scientific sense but more of a boardroom driven agenda that’s more of a self fulfilling prophecy

Reply on news.ycombinator.com