You can definitely understand backpropagation, you just gotta find the right explainer.
On a basic level, it's kind of like if you had a calculation for aiming a cannon, and someone was giving you targets to shoot at 1 by 1, and each time you miss the target, they tell you how much you missed by and what direction. You could tweak your calculation each time, and it should get more accurate if you do it right.
Backpropagation is based on a mathematical solution for how exactly you make those tweaks, taking advantage of some calculus. If you're comfortable with calculus you can probs understand it. If not, you might have some background knowledge to pick up first.
> it's a language made by academics for academics to play with language design. It was a little weird it blew up in industry for a while.
Yep. They have always been pretty honest about this.
I think that it blew up in industry because it really was ahead of its time. Type systems were pretty uncool before Scala. It proved that you could get OO and FP in a single type system.
Actually, a big part of reason for doing Scala 3 was rebasing the language on a more rigorous basis for unifying OO and FP. They felt that for all their other big ideas, it was time to rethink the fundamentals.
I’m not up on programming language engineering as much as I should be at 37, could you elaborate a bit here? (To my untrained ear, it sounds like you’re saying Scala was one of the first languages that helped types break through? And I’m thinking that means, like, have int x = 42; or Foo y = new Foo()”
Not types, type-safety. Things like covariant and contravariant type declarations, implicit types (variables looked up by type instead of by label), and other things that you need to make a type safe system/service/application. The problem is that that feature of a language is massively oversold. Its nice but to pretend it prevents bugs or is even a great design goal is questionable and not backed up by research (as they claim).
> Its nice but to pretend it prevents bugs or is even a great design goal is questionable and not backed up by research (as they claim).
That's why people use JavaScript instead of Rust for critical systems, right?
Claiming in the year 2025 that strong static types don't provide massive advantages is almost laughable, TBH. This was settled long ago, and the whole industry now understands that type safety is inevitable to create reliable and scalable systems.
Papers like this are designed to fit into the conventions that allow knowledge to compound. Not that the conventions are perfect at doing this.
I would suggest that rather than changing this convention in a big way, there needs to be good pathways for communicating the most important takeaways to the general public. Unfortunately, there's kind of a chasm between academia and popular science.
You are providing the standard excuse. It is our job to advance knowledge. It is someone else's job to communicate it to a broader audience. It's just too bad that nobody is stepping up and doing that other job.
I don't buy it. In my experience, most scientific papers can easily be rewritten into simpler language. The act of trying to do so often catches mistakes - thereby immediately improving how well we are advancing knowledge. The resulting paper is easier to read. This makes it more likely to become better known. Both within its subfield, and in a broader audience.
The habit of doing this makes us better communicators. Which also helps academics in various other parts of their job. Including teaching the next generation.
Furthermore, easier to read papers are easier for science popularizers to understand. Which makes it more likely that the work will be popularized.
Yes, it is tempting for academics to deflect responsibility for their role in being understandable. But it is a mistake for them to do so. Their ability to communicate in an understandable way is their responsibility. The few that take up that responsibility benefit themselves.
I'm not saying it's someone else's job to communicate to a lay audience. Simply that a research paper doesn't have to be a self-contained device for doing that and accurately describing the research to people who already have a lot of background knowledge on the topic and methods.
I guess I will say that I have thought for a long time that serializing research into linear documents seems archaic at this point.
It would be nice if academics would move to BOTH publishing the technical write up, AND a more understandable write up of their interpretation of the result (in more detail than the one liner which is in all abstracts.)
The technical writeup is necessary. It's what spells out what they specifically claim to have done, and the specific results. "Specific" being highly technical and fundamental in the scientific community understanding the paper correctly. In particular, the in-depth statistics of many such papers is simply too complex for most of the population to understand, and that's fine. The technical write-up uses terms of art which do not mean what civilians read in them. (And while it's hard to do studies larger than this one, this is all the more essential in smaller studies.)
The interpretation would be useful because it's just plain dangerous to let your PR department write that. Even if they consult you. And it is interesting to focus on what the scientists themselves think they achieved. Both what they deliberately went for, and any ancillary result they think they notice. In this case in particular, they are very focused on this safety aspect, and they seem to not want to give too much attention to the efficacy aspect (which they probably did not plan for and is then suspect.)
Have you considered that it doesn't have to mean the end of your family? It's possible that there's no arrangement that meets everybody's needs. But, maybe there is.
I love the occasionally very surprising submission on here. Sometimes, something with nothing to do with technology touches this group of jaded techies enough to bring it to the front page.
I think she absolutely has a right to her judgment. She clearly has empathy for her father, but the rest of her family also suffered--greatly, it seems--from how he went about his life.
I'm not sure it is exactly the same. But even if so, someone needed to do the work to prove it. It's also worth noting that proving the undecidability of the halting problem is one of the reasons Turing is so celebrated in the first place.
I think a fair answer would be divide the current ticket cost by the amount of cash the average total asset value at the poverty line. Forget about net wealth, since that might well be negative.
On a basic level, it's kind of like if you had a calculation for aiming a cannon, and someone was giving you targets to shoot at 1 by 1, and each time you miss the target, they tell you how much you missed by and what direction. You could tweak your calculation each time, and it should get more accurate if you do it right.
Backpropagation is based on a mathematical solution for how exactly you make those tweaks, taking advantage of some calculus. If you're comfortable with calculus you can probs understand it. If not, you might have some background knowledge to pick up first.
reply