Learning Rust the wrong way - Ólafur Waage - NDC TechTown 2022
Вставка
- Опубліковано 15 чер 2024
- Approaching a new language can be scary and time consuming. The author has been a C++ programmer for quite some time and has tried to give Rust a chance many times over the years with not much success. But this has changed.
In this talk we will:
- Go over the basics of Rust (enough for the talk to make sense).
- Talk about different learning strategies and the strengths and flaws they have.
- How the wrong way to learn for one person can be perfect for another.
- How being a C++ programmer can both help and hinder you.
- Cover common mistakes of new programmers.
- And hopefully inspire you to learn in different and "wrong" ways.
Check out more of our featured speakers and talks at
www.ndcconferences.com
ndctechtown.com/ - Наука та технологія
I love that this applies to everything I want to learn, not just Rust.
Best cross-over ever. Im learning rust and Norwegian and I've been watching this guy's humourous takes on Norwegian for the past six months. Thanks !
Started smiling ear to ear when it dawned on me what he was doing with the constant interruptions to jump into another little topic. Great stuff. Well done sir, well done.
I was just looking to devour some Rust content, but this... This was one of the best conference talk I've ever seen!
I've taken college-level education classes with zero self-awareness, teaching that short blocks and regular breaks is optimal... In a 2 hour long class period with no breaks. Or a 10k word long, confusing to follow, textbook chapter arguing that it's best if you present in small, focused chunks.
The ability to teach education best practices without using them was astounding.
Anyway, this talk was the antidote, thank you so much!!
lol
The jokes write themselves
This was an interesting talk, I almost wish there was nothing on Rust since the "how to learn" portion was by far the most useful part of the talk.
pokemon example was great example that I enjoyed so much seeing it as pixel battle. really nice talk.
The rust portions help space the learning portions, reinforcing your memory of both
I came for Rust, but stayed for everything else, which is pretty much 95% of the take. Basically the talk should be titled "Learing the wrong way, with some Rust along the way". Thanks a lot to Mr. Waage for this refreshingly out of topic talk! I enjoyed every minute of it!
I just realized that you are also the guy I watch when I'm procrastinating about my work ... I hadn't thought about it, but I think I would have guessed that you were a school teacher based on your other videos :D
Jokes aside; this matches my own journey in Rust. I read through "The Book" twice but was unable to really grasp things before work threw me into a project where I had to build a REST API in Rust with an associated CLI client. The struggle that ensued was what finally made the language click.
Cool presentation but I didn't get all the learning tactics.
So what I have:
- Mixing up a different style (readinng books, videos, code)
- Waiting between studies
- Don't cram
- Consistent practice for long time
- Ask questions and find answers
Nothing new. I did'nt see any flashes of inspiration here.
@@greywolf271 maybe for you, but there was a lot of interesting information
personally 30:30 was so great it will greatly improve my future learning process !
I think the most important ideas here are about considering difficulty as en essential component of studying process and ask the questions while working with the topics. I like how Jordan Peterson gives advise to his students skip making notes during the lecture, and instead listen carefully then recall that new knowledge and ask themselves about what they have been learned.
And I also like how Ólafur explains interleaving approach with specific studies. I mean I agree those result are appearing to be not that obvious I'd may expect.
Well watching this video off and on today I was solving a Windows update error. Took me about two total hours to find three different parts to the solution. And what had me think that I've learned anything in solving my windows issue is the documentation I created that would allow anyone to fully recreate the solution. I even tested it one extra time. That's when I feel I really learned. I enjoyed this talk very much.
I'd give this talk 2 thumbs up if UA-cam could. I like how the discussion is backed by research articles and also his own input as well. Very well done.
I actually watched this by skipping past the non programming parts because I was in a hurry and then went back and watched the parts I missed for missing context later. Great presentation and I shouldn't have skipped ahead the first time through I should have picked a time to watch when I wasn't in a hurry.
You missed the point of the talk! Better to skim a bit and come back than just watch it once and forget :)
Thank you so much, wish we could hear the questions at the end though!
Thanks for a good presentation. I am currently learning rust, it is good to know that the approach is not wrong :)
Still one of my favorite Rust presentation. Or just presentations in general. Really helped me be less critical of myself and enjoy my failures to better learn from them.
I'm so happy to hear!
Illusion of fluency really hit hard. Definitely found this to be true with video tutorials. Not saying you can't learn from them, but you do have to go and try to take the concept that is explained in a video and apply it. Will change up my practice to challenge myself to do something with my own noggin and press myself to push pass the difficulty as opposed to reaching out for the easy answer.
"we are minmaxers" oh god, I will take this a key from this lecture :)
37:33 And most of the integration learning in the brain happens during REM sleep, so people perform usually much better the very next day, after a good night's sleep, according to Dr. Matthew Walker, in his book *Why We Sleep*.
Damn that's so true
i always read the comments first before watching these and so glad i stayed and watched! great lecturer!
The differences between spaced and massed learning make sense to me. At least, having been told the punchline I can retrofit a compatible set-up.
When you're learning to classify objects into sets (e.g. paintings by artist), what you want to learn is not just the common elements within each set, but also the elements which discriminate between two sets. Not just the middle of each set, but also its boundaries with its neighbors.
It also makes sense that you learn the most from difficult tasks. If the task was easy you had already mastered it, thus you didn't have a lot of room for improvement. It makes sense if this applies only (or mostly) when you're _successful_ at the difficult task: if you're successful and you know what you did, you can do the same thing again. Success becomes repeatable, and your brain can store the recipe for success.
Listing 46:00 where the video mentions that (at the time the conference was held) there are no rustling exercises for lifetimes, while I'm doing the rustlings quiz (branch named 5.6.1) exercising lifetimes. Which are number 65 - 67 from the total 96 exercises.
Funny coincidence😀
I love how you not only discussed many principles behind effective learning but you also used many of the principles within the talk
At 14:25 a baseball change-up is an "off-speed" pitch where the pitcher uses the same wind-up and arm motion to deceive the hitter. The pitcher uses less or no minimal wrist and finger motion (e.g. put the ball in the palm instead of the fingers) and therefore it travels at a much slower velocity. However to the hitter, it appears from the motion that it should be the same speed as full speed pitches. Note that some curve balls are also off-speed pitches where the wrist and fingers impart more spin and less velocity but faster than a change-up.
I was thinking the same thing. However, he is out of my league on RustLang
This is an awesome presentation. Thank you.
note to self. Occupied right now. Watch later. Seems like a nice talk
Very interesting talk. And it scares me because the learning methods in school are going more and more in the "wrong direction". It is made easier and easier for the children.
Thank you for motivating me to continue my studies especially now that I’m finding it super hard
Im speechless, so many clever points
Great talk, good insights about learning strategy in general 👍.
I benefited from this. Thank you.
Damned... i changed my mind halfway. First it looked like garbage going nowhere... then it works out nicely.
came here for Rust, got an even better insight on life. Thanks!!!
This was an amazing presentation!!
Great talk! The ideas presented are so valuable.
Like everyone else on the planet that likes computers I'm training toy neural networks for fun. Definitely didn't think this would be incredibly on point for that side of things. Great talk.
This was excellent :) Thanks !
This is gold!
Great presentation. Thanks
I have always learned a new language by building a significant project. The first one is not as good as the 10th, but it forces me to learn how that language solves real problems.
I'm a bit confused as to why it is 'counter intuitive' for spaced to be better than massed. Spaced obviously allows for identifying meaningful differences between styles, something that is important for the task of distinguishing those items
People generally like to focus on one thing when learning, mixing one new thing with another other new thing can make people feel like they are overloaded with new information. The research showed that people prefer focusing on one item or topic. It of course depends on the person and the subject but this is the general feeling.
If you would design some CS classes with the goal to introduce the students to three different programming languages: One interpreted, procedural language, one compiled OO language and one functional language (I'm just too lazy to think which ones I would choose). Would you design in a way that students learn every week new things from all three languages or would you start with one language, then have a block for the second language and finally a block for the last language? I guess most people would (intuitively) go for the approach with the three blocks.
Edit: When I studied the first semester was Pascal (yes, it's long ago), the second semester started with C, switching later to C++ ... it was a blocked approach.
This jives perfectly with the story about Kobe Bryant (RIP), who was asked to be interviewed while doing his daily routine.
So the interview/routine begins and the reporter notices Bryant, already a huge star, is spending an hour practicing the most basic maneuvers, and he questions Bryant on this.
Bryant's answer? To become legendary you have to have a solid grasp of the basics!
What a loss to basketball. But also, it seems this non-college educated dude knew more about life, learning, and being excellent, than many academics.
Haven't learned much about Rust, but was still worth the time.
It is a great talk about studying in general with some hard evidence on human memory and practising. However, it has very little on Rust.
Thank you so much
extraordinary talk
I like these videos. thanks!
They are for non professionals
Funny how these professionals get offended when something is not appropriate for their understanding.
He must be a really good senior to be admired by a junior!
Amazing talk!
I'm surprised to find him here. He makes such funny shorts about the Nordic countries. 😂
All the things he describes as counter intuitive or unexpected are the things that I found to be intuitive and expected :)
yes, we learn both by similarity and difference; and a lot goes on at subconscious level, we should not expect to control the learning process
I agree! I feel the best when comparing both similarities and differences.
Nice research of mr Ólafur !
Not the rust I was thinking about, but still very interesting.
Most studies mentioned in the talk is system-1 related, I wonder if the results transferable to something system-2 heavy like Rust
Refreshing talk 🎉
Bravo! This is fantastic
Programming video conferences aren't about learning programming, they're about learning the culture and thought processes of other people. Variation is key here too, where you want a broad span of viewpoints in order to hit the target, which is ability to communicate effectively and mesh with other peoples' thought processes.
I learned programming prior to any talks because I didn't want to have to unlearn anything they might have suggested, but more importantly, I wanted to be able to ignore the programming itself and focus on the motivations and thought processes that led someone to that moment.
Training on the job is far better than training in a university in basically all cases because you have to jump around, take breaks, run into issues, etc. Watching an expert attempt to solve a problem they don't understand, and maybe even failing, is the absolute best possible learning experience because you see the chain of events necessary to make the process function and the priority the expert places on each part.
Really great talk.
After watching this hour long video, I am a Rust Expert now!
very very nice talk
Wait a minute, it's the Iceland youtube shorts guy
I know right!? I was like wait I'm here for a Rust video what the heck?
Holy crap you're right.
great talk!
Useful talk
Best way is the hard way let's go 👍
I thought returning constants (like `str`) in your function exposes you to lifetimes in a rather gentle way.
I was confused by "borrowing into" I think the author means "lending"? (ps, I know borrowing is the technical term `borrow checker` but in this case speaking from the caller (main) POV it would be lending)
24:57, it doesn't work, but it actually should work. The presenter frames it as Rust caught some potential bug, but in reality a perfect Rust compiler should have accepted this code as correct. See the Polonius borrow checker for more details. You are borrowing from two separate memory regions, there is no basis to restrict simultaneous references.
You can do this with slice.split_at_mut(..) or with slice.iter_mut() to work aroudn this borrow checker limitation
This video helped me procrastinating harder xD
35:20 I‘d really like to see the study showing that cramming has no point compared to not studying at all. While I cannot remember any of it, I‘ve successfully written soo many tests where I would not have any knowledge about the topic at all if I hadn‘t crammed the night beforehand.
I mean in the long run. Yes you can pass a test but if you actually really want to learn and remember something, cramming is not good for you.
You can look into "Forgetting as the friend of learning: implications for teaching and self-regulated learning" and "Optimising Learning Using Flashcards: Spacing Is More Effective Than Cramming"
Can someone share rust exercises link Olafur mentioned. Thanks
You can literally watch this video on 2x speed and right arrow skip through it and read the captions, finishing the whole video in 5 minutes and get the same comprehension of all the content said
I expected this guy to explain to me why it's easier to teach rust to a toddler than it is to teach it icelandic
Can someone explain why the Pokemon code at 22:31 doesn't compile? The error says you cannot borrow self.pokemons as mutable and immutable at the same time. This is fine. But we want to borrow two different objects in the vector and not the entire vector. Is there some sort of a workaround or a way to tell the compiler the objects we want to reference are different?
In this case we don't need the objects at all. As the speaker said ( 26:08 ) the programming was correct. Its the Rust compiler is telling us to reconsider implementation. So the proper thing would be to only work with the mutable behaviour instead of the entire object.
This is just a limitation of the rust compiler, a perfect rust compiler would be able to detect that the borrows don't overlap and allow the code to compile. There are some workarounds involving slices, but this will eventually be fixed by a compiler update.
@@typecasto yup. I did a bit of reasearch after watching this video and there is a function which splits the array (or vec) at a given index, and the compiler lets you reference 2 values from each slice, because they never overlap
@@2teaspoon yes but it forces you to write code in a different way even though your solution was correct and safe. That's what i felt when i tried rust for the first time. You end up fighting the borrow checker until you learn how to write code "the rust way"
"van go"
Someone should do a Dutch 'G' course. From below the rivers, I don't fancy listening to a chainsaw.
This validates my ADHD way of learning
I learned Rust while deep-sea diving.
Can’t even finish watching this bc there are SO many ads…
My first impression was that the beginning was a bit long before starting the talk.
Then you had a pretty long preamble about meta points about the talk.
Later I saw the point to the intro and taking that time was actually quite nice.
The baseball section still felt long and I skipped forward a bit and still got the point fine when you got to the point.
The conclusion was a bit vague in my opinion. Like do a little bit of everything over long time and hope it works. The idea of encountering the various pieces of the language semi randomly is probably sound, granting that the studies generalize to the situation which I’ll grant. And that its fine if it feels hard because you might still learn.
But I imagine there are plenty of situation where your training data, maybe the tasks you get at work, or the way you happened to pick problems to work on might not automatically cover a broad range of language features or might not cover them in a semi randomized manner.
So you might actually need to plan to get that kind of distribution.
Those are just some thoughts.
I like the talk in general, though it could have hade more rust in there =)
What excellent feedback. Thank you.
I'm doing the talk again later this year and I'll definitely take these to heart.
Basically this stuff is written up in a review article: Teaching the science of learning, 2018 (Yana Weinstein, Christopher Madan, Megan Sumeracki).
And also in a book callled Make it Stick.
Can someone summarize it, honestly I'm not getting what is the way to do this
Forget structured learning. Don't cram learn if you can (i.e. Reading weeks worth of knowledge in a single day) .
Learn everything you can, and don"t selectively learn stuffs. Read, test, fail, and find out what doesn't stick
An even shorter and incomplete summary could be "Learn the similarities and differences between the knowledge you learn"
❤️
man this was an adhd fueled talk
I use GPTs
hold up... THAT'S THE GUY FROM TIKTOK
"Rust is focused on productivity ..." 🤣
omg thats the tiktok guy
so
Oh, those Nordics...
35:37 I disagree: if I never crammed, I would have failed everything. It was often the first time I had seen the material.
Intelligence is based on discrimination, no confusion. Of course you can't learn to discriminate between things by looking for similarities. For AI builders, all of this is quite obvious.
(feel free to contact me for math models or philosophy about that)
Oh like some math models or philosophy about that! I've been learning cybernetics and I'm dying to see contemporary research on similar themes from an AI perspective, after having sunk my teeth into this university AI course.
@Weaver Marquez Well... if you accept that everything can be modeled in binary, then it's pretty obvious: you need at least two distinct states (usually written 0 and 1) to describe things. Three states are superfluous, because we can always express in base 2 the equivalent of numbers in base 3 or more. On the other hand, you can't do anything in base 1, because there is no distinction to work with between 0 and 0. In base 1, no matter how many identical symbols you write, it remains the same number, the same object. Unless we count the number of identical symbols we write, but then we need more than 1, so at least base 2.
Distinction is the basis of intelligence, because it is essential for signaling, perception, discrimination, comparison, decision, classification, memorization, and therefore for identification as well... well, for everything to know.
Even if you write a method that looks for similarities rather than differences, you end up implementing tests, and those tests make sense only because they can return either true or false, which are distinct things by definition. No matter what you call these things, no matter what symbols you choose, they must be distinct for your program to work. So all models and all knowledge depend on this difference, fundamentally.
From my point of view, this is why people who seek the absolute unity of everything at all costs necessarily give up intelligence and end up very blind. We only perceive contrasts. We only think of differences. One thing alone without relation to anything else that is different, no one can conceive. Knowledge require structure, structure require relations, and relation require differences. This is also, I think, evidence that the universe cannot be perfectly described, unless the description is the universe itself, simply because there is only one universe by definition.
Learners simply minimize a distance between their current state and another desired state in a vector space structured by differences.
Thus, *unperfect* repetition is just an excuse to try many *different* things, to perform many *different* steps. If not, it's not learning, it goes nowhere, it never approaches the optimal state, it's stultifying. The only thing people can learn by perfect repetition is... repetition, without even a meaning.
Whenever a programming language advocate starts talking of "hubris", the language is summarily disqualified for that reason alone.
This video is entirely about programming the programmer, and not at all about programming in rust. There is a tradeoff between what part of the effort lies in writing the program and what part lies in programming the programmer. Rust simply fails that tradeoff. Learning advanced concepts and newest experimental results in cognitive science should not be a prerequisite to learning a programming language. In the case of Rust, it apparently is. There's way too much effort required to program the programmer, and it's much closer to something like "a factor of 20 too much" than to something like "20% too much".
Impasto? Sounds sus.
after 19 minutes of this talk I finally had a reason to write this and close the tab. The dude talked about painting and painters (Van Gogh and what not), then he talked about baseball, and then he went on to talk about Pokemons *and Rust.* So it seems watching this *is the wrong way of learning Rust,* just because the dude spends so much time and effort for presenting a simple idea with pictures and data from totally unrelated topics to programming. Out of these 19 minutes he talked 6 minutes about the basics of borrowing in Rust. I'd leave the room at 19 minutes into the talk.
Sounds like you’re pissed that the description of the presentation was accurate. Hmm…
@@sageinquisitor Exactly. When I read the title I thought the guy would tell what's wrong with other tutorials and show the right way. But he bluntly wasted time. And his crap was recommended. That's why I don't like Rust. It's that the language is horrible, but the hipster hype is.
Give me my 40 minutes back
Warning: this talk will waste an hour of your time. Go read the Rust Book for an hour instead.
Almost no information. A lot of cultural gibberish, philosophical rubbish and anecdotes.
Why wouldn't actionable strategies for learning qualify as information? What parts were cultural and what makes them gibberish? What parts about his philosophy were rubbish? Or is philosophising in general rubbish? Given he cites actual studies, does that prevent the examples from being mere anecdotes by definition?
Though I have to agree the relation with rust specifically was missing.
Congratulations on missing the point so thoroughly
@@mclovin9210 Indulge me if you will, what is the point exactly? I'd like to get it.
isn't it ironic that programmers, who value conciseness and accessibility, constantly churn out these bloated talks. the guy in the video seems like a nice guy, but would love to know wtf this is really about from the onset.
@@raymanovich3254 Use those 4 tactics to learn any new skill efficiently.
a 58 minutes of copium 😂
What a waste of time that was.
If you watch this at 2x, you'll still waste approximately 26 minutes
Thank you for the free knowledge
The Pokemon mut borrow example isnt the greatest, because this code actually is safe and it is a limitation of the borrow checker that it can not handle partial borrows on slices. You can btw. to that with slice.split_at_mut(..)