> The 21st century is a time of great scientific discovery. Cars are driving themselves.
These are the first two sentences of the article and already I'm skeptical. I'd say the self-driving cars are mostly a product of recent technical inventions and not scientific discovery - the scientific discovery happened mostly in the XX century, with discoveries which made integrated circuits, or lasers (for lidar) possible.
The deep learning concept itself did not "discover" anything, it's just a technical contraption that happens to work well for some classes of real-world problems. Granted, there were some maths advances in the eighties and nineties that made current SLAM possible, but again, new math is not "scientific discovery".
You're being overly pedantic over the exact difference between "scientific discovery" and "technical invention", a discussion which is completely irrelevant to the article.
That line isn't even important to the article, it's just a bit of flowery language intended to frame the subject of myth vs. fact.
Okay, but pedantry aside, the 20th century kicked the 21st century's ass when it comes to scientific discovery. Some stuff discovered between 1900 and 1920:
- Relativity
- Atomic nuclei
- Darwinian modern synthesis
On the technological application side, there was the airplane and mass production of cars, plus the growth of telephony and radio.
In the 21st century we have:
- found a bunch of extra solar planets but the work really started in the 90s, lol
On the application side, we have the cellphone/smartphone revolution and continued penetration of the internet, plus consumer EVs and solar panels. Not nothing, but the smartphone is the biggest change to daily life. Everything else is more just refinement of what came before.
I can't really think of other things that are profound scientific discoveries versus technical applications or filling in of minute details. We're just clearly in the far side of the S curve now, and the 20th century was the rollercoaster.
In much the same way that much of 2020's progress was actually started from the 20th century, much of e.g. the year 2060's progress will be stuff actually started from 2020. We likely systemically underestimate the use of today's discoveries simply since the big stuff mostly isn't useful yet.
I agree that we still don't totally know what happened 2000–2020, but e.g. the Eddington eclipse proof of relativity was 1919 and Einstein became an international celebrity around then.
I think I am underrated some math discoveries since it's not an area I follow, e.g. the Poincaré conjecture proof is probably a big deal. On the other hand, when did Poincaré make his conjecture? 1900. So in a certain sense, we are backfilling a known gap.
I feel like it was greatly due to the WWII. Those and so many other discoveries were the direct result of programs like project manhattan.
Maybe the human species really needs to be pushed to extinction to further tξhe human enterprise or maybe it was just the crazy funding similar to bell labs.
I'd just expect an author to be precise when writing something that will be read by thousands of people, as opposing to just spitting out some flowery language.
Wish someone had told me that I wasn't doing real science before I started in academia, would have saved me a lot of time and effort.
On a more serious note, seems like weird gatekeeping of what constitutes scientific discovery and what doesn't.
Scientific discovery is learning new facts about the world. Inventing new contraptions to help us in our goals is technology. I.e. I don't think anyone would call Henry Ford a scientist or Albert Einstein an inventor. I thought it's basic and well-understood distinction.
If you follow the constructivist philosophy of science (spoiler alert: I do) then absolutely Albert Einstein is an inventor. He invented a mental model (essentially a lossy compression) of some parts of reality that better fit observed phenomena than its predecessor. Our mental/scientific models are not descriptions of reality; objective reality is unknowable. What they are, are tools (technologies!) that we can use to predict effect, given a cause and a state.
Is constructivist philosophy of science an independent axiom or does it have testable predictions? E.g. is there a function from theorems of ZFC to {invented, not invented}?
it's a philosophy, so no it doesn't have testable predictions - but neither does objectivism, its counterpart. But then, testable predictions don't tell you about the nature of reality so c'est la vie.
> is there a function from theorems of ZFC to {invented, not invented}?
Can you derive definitions of inventions versus discoveries from set theory? No, probably not. You can't derive the smell of a rose from set theory either though so it's probably not a very good theory of everything.
I completely agree with you and I don't think it's being pedantic at all. Clearly the 21st century is bearing the FRUIT of scientific discovery and we're seeing incredible iteration and innovation that is very exciting. But much of this is the result of things which were DISCOVERED in the 20th century.
Language is important. "Discovery" means something just as "inventing" means something. I see these words used all the time for things like websites (Facebook/Twitter) or computer hardware (Apple products most notoriously). Something can be a wonderful and very popular implementation without being either a discovery or an invention. Often these larger things DO contain many smaller real discoveries and inventions along the way! But people always seem to be referring to the end product as a whole, which I find bizarre.
These are the first two sentences of the article and already I'm skeptical. I'd say the self-driving cars are mostly a product of recent technical inventions and not scientific discovery - the scientific discovery happened mostly in the XX century, with discoveries which made integrated circuits, or lasers (for lidar) possible.
The deep learning concept itself did not "discover" anything, it's just a technical contraption that happens to work well for some classes of real-world problems. Granted, there were some maths advances in the eighties and nineties that made current SLAM possible, but again, new math is not "scientific discovery".