Video
Đăng bởi Ban biên tập OpenER ·
Tác giả giữ toàn bộ bản quyền (All Rights Reserved) — không phải giấy phép mở. Bạn vẫn xem và học miễn phí trên OpenER; muốn tái sử dụng cần xin phép tác giả.
BORO NIKOLIC
And record. Yeah.
DAN GARCIA
Perfect. Wonderful. And we'll keep you on the site. We'll spotlight Boro and myself and keep you on the background until we pull you in, if it's Okay.
That way, you can just relax for a second.
CRAIG ZILLES
Perfect.
DAN GARCIA
Okay.
CRAIG ZILLES
I won't need the mic on?
DAN GARCIA
Yeah, exactly. Here we go. Just spotlight this guy. Stop video.
Spotlight for everyone and spotlight for everyone. At spotlight. Perfect. All right.
Good stuff. And we have 36. All right, welcome everybody. It's great to see you folks.
BORO NIKOLIC
Welcome, welcome.
DAN GARCIA
Boy, busy time in the political space if you're a — I just got my ballot. I'm very excited. By the way, everybody vote. This is recorded.
I want you all to vote. I want you all to exercise your rights, folks. This is very important. So make sure you do that.
BORO NIKOLIC
We're going to remind you every week. We're going to remind you every week.
DAN GARCIA
Yes. And be informed. Read up on the ballots. Look on that.
There's lots of propositions. Look on the propositions on the ballot. Make sure you understand that.
Craig, we have these propositions which are kind of — any citizen can put a proposition forward, and it becomes law if enough people vote for it.
BORO NIKOLIC
Yep. I grew up in California, so.
DAN GARCIA
Oh, yeah. All right, you know.
BORO NIKOLIC
Yeah.
BORO NIKOLIC
I got my ballot last Friday or so.
DAN GARCIA
Yeah. Definitely. Actually, I spent three hours with my best friend, who lives in California. We just went through every single one.
We marked it, and then we talked about it. And we thought — it was great, really great. Do that. Do that kind of thing.
All right. Let's get started. Ladies and gentlemen, welcome to week eight, once a week lecture, RISC-V Datapath Control and Pipelining. Whoo!
Great to see everybody here. Wonderful, wonderful.
BORO NIKOLIC
All right.
DAN GARCIA
Let's get started. So here's our agenda. We're going to do, as we always, welcome to week eight. We'll do a little computing in the news.
We have a wonderful guest, a colleague and friend, Craig Zillis, from University of Illinois Urbana-Champaign, who's going to join us in a moment or two.
Then we'll talk about our week's plan and any — we have some information about the upcoming exam. I'm sure you want to care about that.
But we'll let Craig take our leave after we get done with that.
BORO NIKOLIC
Yeah. Yeah.
DAN GARCIA
All right.
BORO NIKOLIC
Important thing. We are past the midpoint. We are past the midpoint of the class.
DAN GARCIA
Exactly.
BORO NIKOLIC
It's all just downhill from now.
DAN GARCIA
I love it. I love it. No, it really is. It is.
It's like my — I'm over 50, Boro. I think you are, too. So it's like, just like our lives. Just downhill, just coasting.
Put it in neutral. Let my Tesla recharge. Perfect. I just got it.
I got it there. Computing in the news. All right, computing in the news. Waymo.
You know that Google split from Google to all of its different versions, A through Z, I think they called about it, and Waymo is one of them.
And for a while, they were having these little teeny kind of mini, looks like a VW minivan, but a much smaller version that had no steering wheel.
And they were doing test drives around Mountain View. They went a little bit bigger.
And they went to Arizona, where the roads are really wide, and a lot of people running around like New York City or a big city. Very slow and soft kind of a space.
They did all this test prep for this. And in fact, I just read on the article — the article is linked here from Bloomberg.
When people were doing these beta testing drives, riding in these Waymo cars that had no users. For a while they had a person there with their hands on the wheel, just making sure.
Then for a while, they actually had nobody there. This has been beta test mode, just to see if it worked. Really amazing.
And people, everybody riding in those cars doing beta testing had to sign an NDA, a non-disclosure agreement, saying, if the car ride is terrible and it stops er, er, er, every two seconds, they won't complain.
They won't post on social media. They literally had a silence. It's like a silence clause they had everybody do. I mean, they certainly asked them to give feedback, how did it go, and blah-blah-blah.
They're ready. They launched. Can you believe it? The future is here, folks.
Elon Musk keeps telling me my Tesla will be able to drive me from here to there without my hands on the wheel still. But like, roads might — even Elon does highway.
So this is it. Look, they've got the system in Arizona. I think there are 300 or 500 cars, so it's still small, still small scale. But they've got cars that have no driver.
This is unbelievable. This is like Disneyland or Fairyland.
So I think, I would hope that people who are using these cars will share some video, high-def iPhone video of here's me getting in a car. There's no driver.
Look, aaahhh, we're driving around. I cannot wait. So exciting, exciting. I hope, I wish everybody's safety.
I certainly wish for this to succeed. But it's here. The future is here. All thanks to LIDAR sensors, crazy machine learning.
And the same thing we talked about last week — Boro, we talked about this last week and the week before. These high-end machine learning systems, they are built into these systems.
That's how they work. It really is the key to — thanks to all the hardware advances that we've talked about.
BORO NIKOLIC
And thanks to a whole bunch of our alumni who went there. I mean, the chief architect of this is Pierre Entralls, who got the masters from Berkeley.
DAN GARCIA
Yeah.
BORO NIKOLIC
Yep. And a bunch of other people who just went through as our alumni outside of just taking classes. Some of them with advanced degrees.
Some of them with bachelor's degrees. And just made all of that.
DAN GARCIA
Yeah. Exciting. Exciting space. Grand time to be here.
Wonderful. All right, let's keep on going. Ladies and gentlemen, let me introduce our guest of the day, Professor Craig Zilles of UIUC, University of Illinois Urbana-Champaign. So Craig, let me now spotlight him.
Give me a second. I'll spotlight him here. And add spotlight. And I'll leave it on spotlight ourselves just to focus on you, if that's all right.
Remove spotlight and give it to him.
CRAIG ZILLES
Am I on spotlight?
DAN GARCIA
Yep. You're on spotlight. Here we go.
BORO NIKOLIC
There you go.
DAN GARCIA
So Craig and I met each other. Craig has been part of the computer science education. Craig is a regular professor. He does research in architecture.
And most recently, he's been working on how to make exams scale online. What's the future of exams?
Just like, in a way, Piazza is some sense the future of online forums, as is Slack and Discord and other things, just like Gradescope is the future of exams with computers.
Future of homework and auto-grading and supporting that, and scanning exams.
Craig's really research is thinking about — and maybe, Craig, correct me — is thinking about, could we build a system that will allow us to randomly generate questions, to allow us to be really creative.
Rather than having to write exams every year, what if we split the time into actually writing question generators and packing them all together?
And Craig's really cared about, how would you then assess a campus? So he's worked on something called the Computer-Based Testing Facility, or CBTF. That's a space where students come to take these exams.
So just like you're taking our exams — by the way, just for FYI, the system that you are using for your Quest and your midterm and your final is a system that, again, thanks to — it's called PrairieLearn.
And it's thanks to Craig and his team for building that PrairieLearn system.
Craig's cared a lot about building a place, an actual physical location called CBTF on campus to allow people to go and drop in whenever they want to, when they show their ID and they take an exam.
And they can take a range of time. You can have as much as you want to or you can have a limited time. There might be a list of things you can bring in.
All those things are part of the CBTF to try to make this really rigorous. And what's nice about Craig is he's done research to show how effective that is where you see cheating.
It's really interesting stuff that Craig has done. And so Craig and I have known each other for many, many years, and it's great to have you here. Welcome, Craig.
CRAIG ZILLES
Thank you.
DAN GARCIA
So tell us a little bit about how you got to where you are. Did you always know you wanted to be a professor? Where'd you grow up? When did the academic bug bite you?
CRAIG ZILLES
Yeah. So I grew up in the Bay Area, South Bay Area, just North of Santa Cruz. I did my undergrad actually, surprisingly, my first two degrees are in mechanical engineering.
I was the kind of kid that — DAN GARCIA: I didn't know that. I didn't know that. CRAIG ZILLES: --always was taking things apart and trying to put them back together.
So I studied mechanical engineering at MIT. And then I did my master's degree in robotics, which I think of as the gateway drug between mechanical engineering and computer science.
DAN GARCIA
That's true.
CRAIG ZILLES
I then worked for a year and went to graduate school and studied — basically, I worked for a year, and I found like, oh, I really want to do more of this computing stuff.
And I got interested in digital electronics pretty deeply. And I realized that my background, even though I had taken a number of computing and electronics courses, that I didn't have even a full degree.
And so I went back to graduate school in electrical and computer engineering, and ultimately transferred to computer science for my PhD.
I got super interested in computer architecture there. So I did my graduate work at the University of Wisconsin-Madison, which has a very strong group of computer architects.
And so I did my PhD in computer architecture. And that whole time I was studying, my intention was to go and build processors.
So work for the IBMs and Intels, back when there was Digital Equipment Corporation, and there was a lot more variety of computer companies back then.
DAN GARCIA
Right. All on Route 128 on the western part of Massachusetts, right?
CRAIG ZILLES
There's the Massachusetts group, the West Coast group in Portland. HP was in Colorado at the time.
And so I actually — I almost left with a master's and took a job.
I had a very interesting conversation with one of my mentors, who was Joel Emer, who is now a Fellow at Intel and actually is retired from Intel and teaches at MIT.
And he sort of said, don't necessarily be in a hurry to work in that when you graduate, you'll have 30 or 40 years to work, that if you want to get a PhD, take the time, that there's sort of no experience quite like it.
And that was — work will be there for you, four years later. You'll still have plenty of time to work. So that was pretty compelling.
But even then, most of the time I was expecting to go into industry when I graduated.
And so again, there was a mentoring experience that some faculty that were on my PhD committee pulled me aside and said, we think you'd really be a really good professor, and so you should think about that.
And so I guess my recommendation would be to seek out people to give you advice about that. So I've taken sort of a crazy path from mechanical engineering through computer architecture.
And now, as Dan was saying, a lot of my work relates to how do we teach, particularly teaching at scale.
The project that Dan mentioned is really focused on — so our campus, the University of Illinois, is the flagship campus for the state of Illinois, just like Berkeley is for California.
We are the second-largest engineering school in the country.
And so we have some of the same challenges that you guys have, which is we're a really good institution, but we also have a lot of students going through our classes.
And so how do we maximize the quality in the presence of the number of students?
And so our project, the Computer-Based Testing Facility, was really about how do we make sure assessment, in the form of exams and quizzes and things like that, meet the demand of our large classes without losing a lot of quality?
You guys know how much time you spend studying for exams. But you might not realize how much time your professors and their core staff spend preparing and running exams.
Typically, the standard model of maybe like two midterms in a final that was pretty common with pencil and paper exams is because it's really painful for faculty and their core staff to run exams.
And we thought, that's really not good for the students, that having these sort of big, heavyweight exams relatively uncommonly would be problematic.
And so we are thinking, like, how could we sort of make this a better experience for our students. And that's sort of what led to us thinking about the Computer-Based Testing Facility.
DAN GARCIA
And so tell me a little bit — tell me, if you're not going to give two midterms and a final, what do you do?
How do you do — and by the way, you teach a course. I understand. CS233, very similar to 61C. It's almost exactly the same course.
It was MIPS, and we do RISC-V. But we also did MIPS about four or five years ago. So we actually almost taught exactly the same course.
Tell me what you do in your vision of assessment.
CRAIG ZILLES
There's two main things that we do. One is, we do very frequent testing. Basically, every two weeks we run a 50-minute midterm. And this starts in like week three or four of the semester.
So the first one, the idea is that we want to frequently test you so that if you find that you're falling behind, you get that feedback early.
What we don't want to have happen is that your first exam is in week six or seven of the semester. And by then, if things aren't going well, it's insurmountable to catch up.
So we want to do frequent testing. Give you constant feedback.
Then if we're doing frequent testing, then each exam, it's only like 5% or 10% of your grade.
It's not quite as scary, and the amount of studying you have to do is not as much.
And then the second thing that we do is we do second-chance testing, which means that — DAN GARCIA: I like the sound of that. I'm sure the students do, too.
Tell us more.
CRAIG ZILLES
The students love this. The idea of second-chance testing is, if you go in and you take an exam — as Dan mentioned, we use PrairieLearn — you get immediate feedback.
So you know when the exam is over, because we do auto-grading on almost all of the exams. You know how well you did on the exam.
And so what we do is the following week, we run an optional second-chance exam where, if you want, you can study the things that you didn't do well, come back, schedule an exam that following week, and take it.
And we do some form of grade replacement.
DAN GARCIA
Do you do a full replacement, or just some frac — like, do you have to own some part of the exam the first time, or can you kind of get a zero the first time and ace the second one, and somehow get an ace by the two of them?
CRAIG ZILLES
It's a different — this is actually one of the things that we do research on is understanding these policies.
What we found is that — and I'm sorry to tell you this, students, but this seems to be true — is that if there's no penalty for, there's no consequence for doing poorly on the first exam, people will tend to blow it off.
DAN GARCIA
Ooh.
CRAIG ZILLES
There's a — yeah.
DAN GARCIA
Because we have a full replacement policy here in our class. It's called the "clobber policy," if you do better on a later exam, but only on the same question.
Like the later, the midterm might have one question that was on the previous exam. So if you aced that one, it's a full replace — we're doing a full replacement.
Usually, it was a curved replacement. But this year, we're going to do a full replacement. So whatever your grade is, if it's higher, it's higher. If it's lower, it's okay.
It's just averaged in. But if it's higher, completely replaces the earlier one. But you're saying that people don't study for the first one.
CRAIG ZILLES
I'm saying that some people don't. So we definitely see — we've had instances where people just don't even take the first one, if we use that kind of policy.
So we like to have some skin in the game for the first one.
Because I think what the research suggests is that some of the benefit of second-chance testing comes from taking the — studying for the first exam, taking it, getting the feedback of what it is that you still don't know, taking that feedback and studying those things, and coming back and demonstrating that you now know them.
Going through that cycle twice, rigorously, is important. I actually started doing second-chance testing when we were doing pencil and paper, and we were grading everything manually.
DAN GARCIA
That's a lot of work. Wow.
CRAIG ZILLES
It's a lot of work.
DAN GARCIA
Oh, my gosh.
CRAIG ZILLES
And so basically, at that time, we wanted to — there was a consequence of a student if they just didn't study, and they went in and they took an exam.
They'd just be like, hey, if I can get by without studying, great.
And I totally understand you, students, your desire to optimize your life, because you have three or four other classes that are going on.
DAN GARCIA
And they don't have that policy. That's the danger, right? You're the one who's soft and squishy with your policy. They don't.
So they really have to study for those classes.
Then they come to you and say, going to lean on you to borrow time from your class for their class, because that is no retakes, in that sense.
CRAIG ZILLES
Right. Yeah. Everybody's an optimizer, right? Actually, this is tying this back to computer architecture.
That's very much what computers, the way we design computers, is all about, optimization problems of like, okay, how can I, given a finite amount of resources, how can I do the best with those available resources?
So I don't fault any student from being optimizers.
But it just means that there's pressure on us as educators to design policies that get the behavior that's going to be most beneficial for the students.
And so what we found is having some amount of skin in the game of the first exam.
An example policy that people seem to like is 10% from your lower score plus 90% of your higher score. So it's almost complete — DAN GARCIA: Clobber.
Yeah.
CRAIG ZILLES
Yeah. Complete clobber, but not perfect. So that's what we found.
DAN GARCIA
And overall, what's the reac — I'm sure those students love the retakes. But overall, do the students appreciate? Because there's some things — it's funny.
I have this theory, which is there are some things that are better for students. But it's like broccoli. Eat your broccoli.
Even though you're not going to like it, you're going to tell me with the surveys you don't like this. But it's better for you.
It's better for you to do the reading, and force you to do the reading before you come to lecture, because you'll get more out of lecture. Says, I don't, blah-blah-blah.
They don't do it. Aside from the retakes, which they love, what do they think about the more rapid assessment versus the bigger high-stakes things? Do they appreciate having an exam every two weeks?
Is that overall? Because it sounds like they do better at the end. From the point of view of better learning, that's better. But what's the student perspective on that?
CRAIG ZILLES
I think they, in general, like it up to a point.
DAN GARCIA
But there's no break. Because like, I got an exam every two weeks. I can't have a break here.
CRAIG ZILLES
Right. And I think, especially if you imagine that you're starting to get three or four of your classes that are testing frequently, it can be fatiguing.
But I think — so that is a concern. But in general, students are very positive about smaller, more frequent assessments, especially if they have flexibility.
Our system back before COVID, we have a physical space. We have a couple computer labs on campus, and we give students a three-day window to make a reservation.
And they come in. They take their exam whenever it's convenient for them.
And so that's another feature that they really like, is because if their psychology class has an exam on Thursday, they can choose to take the exam in my class on Saturday instead of the Thursday that I would have run it on.
So they get to pick a time that's convenient for them. Another thing that they like is picking the time of day.
Some kids work well at 9:00 PM, and they'd like to take their exam then. Other people want to get it done at 10:00 AM in the morning.
And so having that flexibility is really beneficial.
DAN GARCIA
What do you think about having untimed exams? We had a 24-hour window to take an exam that was some number of hours for some people.
Here, we're now going to have a 48-hour window for our midterm. And we've played with it. We played with having it basically on time, 24-hours exam.
And the second one, we're actually trying, we're going to talk about this, maybe limiting that time per question. We'll [INAUDIBLE] with the question.
So what do you think about the trade-off between untimed versus timed exams in terms of that?
CRAIG ZILLES
I strongly believe that you should have enough time for the exam.
I think one of the other things about why we are interested in frequent testing and early testing relates also to having enough time.
One of the things that we want to be true is that we want your score in our courses to be dictated by how well you understand the material, not necessarily by how good your test-taking skills are.
So we don't want to design exams where, oh, the right way to take the exam is to skip the first three problems and do problem 4. Then if you have time, go back.
Definitely, having enough time set aside for your exam I think is super important.
There's definitely a point for a bunch of students where they've done what they can, but they still have 20 minutes left, and so they stay there for 20 minutes.
One of my concerns would be, is this an extra burden on those students that feel like they should spend more time, even when it's not being productive?
I guess the other concern I have, one of the things that the literature suggests that is a mitigation for cheating in online exams is having time for them.
So if that is a concern. I know none of the students at Berkeley have ever cheated in history, so.
DAN GARCIA
That's certainly true. Actually, I was going to bring that up. Tell me. You've done some research into cheating.
Tell me what you found. CBTF is a place where you can guarantee, as much as you can, that cheating wasn't happening because there's somebody watching the space as a staff member.
You're in a booth, in a kiosk, and it's just you and nothing else and whatever you are allowed to bring into that exam, calculator or whatever.
But now that you can't go to that place and you're kind of asking these students to take these exams at home, what are you doing in terms of — is it is an honor code?
Are you having them bring a camera? Are you recording the space, recording the screen? How are you managing cheating?
Because cheating is a reality, even though we know no one has ever cheated before, either. What are you doing, at least to make sure it doesn't happen in the future?
CRAIG ZILLES
Yeah. I think preventing cheating in online exams is basically impossible, so actual prevention.
So we go for a number of mitigation strategies.
We've instituted — since we can't run the physical CBTF, a group of people have organized what we call the CBTF Online, which is some Zoom proctoring.
So we do have people that are one-to-many proctoring. So that's one thing. We're not trying to lock down the machines.
I think one of the big things that we do is, as you mentioned earlier, we do a lot of randomization on our exams.
And this sort of came from when we started running exams asynchronously.
The thought is, if you let different people take the exact same exam at different times, then student A goes in and takes the exam, comes out, and talks to student B.
And student B has an advantage over student C that doesn't have friend A. So if we randomize the exam so different students have different questions, then it means that it's very difficult.
You're more dependent on what you know than what other people have told you. And it also means that we can give out practice exams that are very similar in structure to the actual exam.
DAN GARCIA
Sure.
CRAIG ZILLES
So a heavy degree of randomization is, I think, one of our keys.
This semester, we've been running synchronous exams to try to mean that at least students in the class at the same time have to be focusing on their exams rather than potentially helping other students.
DAN GARCIA
Right.
CRAIG ZILLES
Yeah. We don't — DAN GARCIA: No, it's hard. We've got to support multiple time zones. And so all of that with 1,200 students has been just a challenge.
And so we're mostly leaning on the honor code, and doing what we can in terms of randomization, some other things we can to try to help with that.
Tell me the things you're excited about. Where are you kind of cutting-edge research, the next five years? This is what we always ask any candidate, a faculty candidate.
What are you excited about the next five years in terms of AI, mixing with PrairieLearn, mixing with CBTF? What's the space? What gets you up in the morning?
CRAIG ZILLES
I've been recently teaching one of our CS1 classes for Berkeley people. This is close to the CS0 that you guys teach, The Beauty and Joy of Computing.
And I have a couple things that we're working on there. We've been focusing on a very reading-before-writing kind of curriculum.
One of the key things there is the research literature in learning to program suggests that being able to read and explain code in English is a step, is a building block step to actually being able to write code.
And so my group has developed a natural language processing system that allows us to present pieces of code to students and have them type a description in English and automatically score that.
DAN GARCIA
Wow. Wow. It's kind of the pseudocode, kind of describing the pseudocode in English, they're kind of saying it. Well at first, grab the number.
But are they doing, like — because when I've asked students to do this, they go step by step. They literally read. Well, this one, 4i.
They're not able to step back and say, I don't care about the 4i. Tell me that you're iterating over the list and doing something. But they can't do that at the early stone.
So what are you finding your students are doing? Are they line by line, or are they able to pull back and say — CRAIG ZILLES: Right.
So we've trained this autograder to not give credit to the low-level description.
DAN GARCIA
Wow. Wow.
CRAIG ZILLES
As you face. Our goal is to — DAN GARCIA: It's impressive.
CRAIG ZILLES
We want them to see the forest for the trees. So it's like — DAN GARCIA: Right. Right.
CRAIG ZILLES
This function is making a new list out of the even numbers that are in the provided list.
DAN GARCIA
Right, right, right, right, right.
CRAIG ZILLES
How many regrade requests are there? There's some. It's not perfect yet. It's still research.
It's about 90% accurate, which is really good for homework. We put it on exams.
But every time we put it on exams, we have a human go through and make sure it doesn't make mistakes.
But I think the key thing is — and I think the whole point of automation, the power of automation, is that any time a student wants to study, they can get immediate feedback on their work.
So to me, it's almost a more important formative tool than actually putting it on exams.
DAN GARCIA
Definitely. Definitely. There's a question, actually a very good question from Benjamin, who's also asked good questions in the past.
He's saying, "For a heavy degree of randomization, how can you ensure that each given exam is the, quote, unquote, "same difficulty"?
Since it seems the same question with different numbers and parameters would be easier to randomize but also make cheating easier, whereas truly randomizing the questions could lead to differences in difficulty of the exam between students."
Boy.
CRAIG ZILLES
This is a great question.
DAN GARCIA
A very good question.
CRAIG ZILLES
And it's something that we've studied. I think the key thing is we're not perfect. We admit we're not perfect. Our goal is to be at least as good as semester over semester.
So historical practices in fall semester, Dan would write an exam. In spring semester, Dan would write a different exam. And those exams wouldn't be perfectly fair.
So maybe if you took it in fall, Dan wrote a slightly easier exam versus his spring.
DAN GARCIA
And even within a question. Like, my STS question might be easier, [INAUDIBLE] digital systems easy in the fall. But the cache is easy in the spring.
That's even with some variants, even in the question.
CRAIG ZILLES
Right. Right. There's going to be some variants. And so basically, we're translating that variance that used to be between semesters to within a semester.
And so that variance is there. We've studied it. Sometimes we make mistakes, and we need to fix them.
I've definitely had the case where a problem that had an average score of 37% was in the same pool as a problem that had a 70% average.
DAN GARCIA
Wow. Wow.
CRAIG ZILLES
And we're like, ooh. Something bad happened to — DAN GARCIA: So how do you correct it? Do you throw it out? That's the hard one.
Do you the double it up, just double the, double the lower one, too?
CRAIG ZILLES
Yeah. And so one of the key things is I tend to have many, many questions on my exams.
And that's another thing that makes me, helps me sleep at night is that there's a bit of central limit theorem going on, is that if you ask enough questions, that even if — DAN GARCIA: Large numbers.
CRAIG ZILLES
All large numbers. You're going to get some hard ones and some easy ones, and somebody else is going to get different hard ones and easy ones. And it's going to work out.
DAN GARCIA
It's work out. Yeah.
CRAIG ZILLES
So we do that. One of the things that I'm looking to do going forward is there's something called Item Response Theory, that when you did that — DAN GARCIA: Yes, indeed.
IRT. Yeah. I served on the college board. And we spent — I learned that from them.
I was in the meeting. I had no idea what it was. And they said, let me teach you what Item Response Theory is. And I learned it.
Wow, this is so cool. Most people don't know this. You didn't know this in teachers' school, you know.
DAN GARCIA
So all of you took some exam like the SAT or ACT in order to get into college, or most of you did?
DAN GARCIA
Yeah.
CRAIG ZILLES
In general, those are different exams, but somehow they uncovered a score for you. It's not a particularly transparent mechanism.
And so I don't think it's the main mechanism that we want to use when grading students, because students value transparency.
But I think as a background mechanism of, like, let's go ahead and run this in the background.
And if somebody got hurt by the main mechanism, we'll boost their score up to what IRCA would have said they should have gotten.
So if you are that one person out of a thousand that gets really unlucky and got more than the average number of hard versions, we'll bump you up at the end of the semester.
So those are some of the things we were thinking about. Go ahead.
DAN GARCIA
Go ahead, Boro.
BORO NIKOLIC
We spent quite a bit. We talked about this. I mean, what happens if you discover this in Gradescope, that we have very different outcomes in score.
So our approach was that we are going to go ahead and renormalize somehow. But we'll see. We are hoping that we are going to make our exams perfect from now.
CRAIG ZILLES
Right.
BORO NIKOLIC
But we are preparing for them not to be perfect.
DAN GARCIA
And actually, just for the students, 77 students here on the call. We actually threw away — we wrote, I think, seven versions.
It was seven different versions of this, the question, the second programming question. Section 3, the splitting question. And then we threw away three of them because they said these aren't the same difficulties.
So we already have been thinking about it. We already decided that only the bottom four or whatever, some number, were the same. We made some more that were variants.
But the point is we threw three of them away because they said, these are definitely — and actually, I want to give CeeCee credit. Our head TA, CeeCee, came up with it.
Said, you know, Dan, I don't think 1, 2, and 3 are the same as 4 through 7, so let's throw 1 through 3 away and not use them.
So I thank her for thinking about that. But that's an issue.
BORO NIKOLIC
And they're perfect practice questions for going forward. I mean, once when we use all of that in the exam, they're good for practicing. We don't have to be completely fair.
There is not the same scrutiny when we give something, give out as a practice question.
DAN GARCIA
Makes sense.
BORO NIKOLIC
And we definitely have pretty hard ones also. Not going live in any exams anytime soon. I have a quick question — DAN GARCIA: I love this.
I love this — BORO NIKOLIC: --and then we'll let you go.
DAN GARCIA
Yeah. I love this natural lang — I just want to say, I love the natural language to work. That's really exciting. I'd love to see it in 6C or any other conference.
I want to see some results from that. That's wonderful.
CRAIG ZILLES
There's a paper that got accepted at [INAUDIBLE].
DAN GARCIA
Hey, congratulations. Yeah, it was a good year. It was a good year. Any more questions [INAUDIBLE] there?
BORO NIKOLIC
I have a quick, I have a quick question, a different thing. So I'm assuming some of our students want to go to grad school.
So how do they get — do you have any tips on how could they get into UIUC? We heard UC is good in architecture.
CRAIG ZILLES
Yeah. So we're good in a number of things. We have a very strong computer education group now. We got Colleen Lewis.
DAN GARCIA
Got Colleen Lewis. Our own student, yeah. She's one of our all-stars.
CRAIG ZILLES
Yeah, former Berkeley grad.
DAN GARCIA
Yeah, definitely.
CRAIG ZILLES
She's amazing. Yeah, like so I think the key things I would recommend applying to grad school.
One of the biggest challenges of computer science is it's so big, so trying to figure out what it is you're interested in is a challenge.
I think that's something that you guys as undergrads need to figure out as you work through your elective courses. But get to know faculty in that area.
Letters of recommendation are super important. What grad schools are looking for is somebody who understands what research is, that can sort of say, oh, yeah, this person has the characteristics.
That means that they're likely to be able to do research, and whether the person says that because you've done research with them or because they've seen enough of you in other contexts that they can say that.
That's what you're looking for. So I think making sure that you are making some relationships with potential letter writers early.
And then the other thing that faculty know is if you say, I'm interested in this subject. What are some interesting schools? Because not every school is equally good in every dimension.
Again, CS is so broad these days that not every department nails it in every sub area.
And so finding the faculty in that area that you're interested to be able to tell you oh, okay, here's the list of schools that are particularly good in my area that you should be targeting.
Because it's just like undergrad. When you're applying to undergrad, there's just so many places, and how do you make sense of all that information.
The point of grad school is a little different in that — DAN GARCIA: It's who you want to work with, really.
CRAIG ZILLES
Yeah. It's a lot of who you want to work with.
But I think one of the neat things is that once you get admitted, a lot of the schools will invite you out, and you'll have an opportunity to visit.
And that's just sort of a super unique — it's almost the time where you'll get more interaction with faculty, per unit time, than almost any other time in graduate school.
DAN GARCIA
That's true. That's true.
CRAIG ZILLES
So taking advantage of that opportunity. And when faculty that are trying to recruit you to come there, ask them questions. Like, what do you think are the important problems in this discipline?
What should I, as a new graduate student, be thinking about?
DAN GARCIA
Definitely. Definitely. Wonderful. All right.
Thank you. Craig, thank you so much — BORO NIKOLIC: Thank you. DAN GARCIA: --for taking the time. It's always a pleasure.
It's always great to see you.
BORO NIKOLIC
Thank you.
DAN GARCIA
And thank you also for what you've done for PrairieLearn and all of us doing these online exams. Thanks to your work and your team's work. It's outstanding. Let's give folks a hand.
Thank you very, Craig.
BORO NIKOLIC
Thank you.
DAN GARCIA
Whoo! Good stuff. All right. Take care.
Thanks again.
CRAIG ZILLES
Thanks so much.
DAN GARCIA
My pleasure. Wonderful.
CRAIG ZILLES
Thanks for the invitation.
DAN GARCIA
All right. Yes, indeed. All right.
BORO NIKOLIC
Back to the main program.
DAN GARCIA
Back to the main program. All right. We're back to here. And I'll remove Craig's spotlight.
Perfect. Okay. All right, we're back. So Boro, you want to take us through this upcoming schedule?
It's a fun week.
BORO NIKOLIC
Yep. So what we got here. I need my glasses over here because — DAN GARCIA: Small fun. Small fun.
BORO NIKOLIC
My computer is misbe — now I know. My computer is misbehaving, so it doesn't want to project on a big screen. I have to disable that. Otherwise, I can't pin.
If you've noticed on the recent videos, I'm not scribbling because I can't. It's refusing to do that when it's connected to the screen. Yeah, it is still totally annoying.
So where are we now? I have at least one lecture I'm pipelining. I am hoping to get the other one out tonight, and the other one later, the Monday's lecture.
So the Wednesday lecture has been released, I think, on Saturday. Friday's lecture will go out today. And in the next few days, we're going to have also the next Monday's lecture.
And that essentially finishes all the material for the midterm.
So you have almost two weeks to sit on that material, and adjust it and make sure that you're comfortable with it.
We have one more lab that is pipelining, that is preparing you for our second part of the project 3. That's the pipelining lab. It does take a bit of a time.
Get ready for that and use it to your benefit. There is no lab next week. Just prepare. Use that time to get ready for the midterm.
And that's all. And there is homework. By the way, the homework that we have out now is going to take — we shortened it a bit.
We took out — one question is left as an optional. That was like a 20-minute, half-an-hour question.
But the homework is still going to take 2 to 2 and 1/2 hours, so budget for that between this week and next week. But the questions there are good.
The questions totally make sense as a preparation for the midterm.
DAN GARCIA
Yeah, I was going to say, I can't think anything, any better way to study for the midterms than tell them to do the homework. We always suggest that.
Great, great stuff. All right. So we've got some announcements. Let me click on that.
All right. So Quest regrades, finally done. Holy, moly, I have never seen a regrade that was as painful as this one.
We had virtually one in four students needed to talk to us with a lot of — it wasn't just, like, I want points in this.
But it was like, there was some technical or some this or that, and all those things. Some of them actually required us to rerun.
We would make a change, and then rerun it on everybody's, and so all those things. Anyway, it was painful.
But thanks to a great set of TAs who helped coordinate all that to make it work.
So about one out of every four students, which I think is the highest fraction of regrades we've ever had. And so that's why it took us so long.
So thank you for all your patience with us. We've got a final date. We've got a Quest retake this Sunday, 24-hour period. It's going to be exactly the same.
We can't make it — as much as we want to give you test cases, we can't.
Has to be exactly the same to be fair to anybody who just had trouble technically with it and needs to redo it. So that's the idea. I was just looking here.
I've got my — there's a whole thread on Slack where my staff is right now, working through all the students who had challenges and what the scores were.
And we're going to — we won't make everybody do it, but we're going to try to figure out the final list by Tuesday. So we'll let you know by Tuesday.
Just don't send us mail. And by Tuesday. We'll send it out. We'll send a big Piazza post, saying, here are the people who need to take it again, and then you just do this.
But it's essentially the same. You're already past the material. Hopefully, the C is already deep within you, the number representation. And all that, by the way, i great preparation for the midterm.
Hopefully, you don't have to do any extra studying. You're just basically studying for the midterm and kind of refreshing yourself. That material is also fair game on the midterm.
So you're studying for this in general with this stuff. But we'll let you guys know, and that'll be between Tuesday and Sunday.
And hopefully, there's enough days for people to get some time to do whatever prep. And also, by the way, it's open notes and open book.
By the way, did I ever tell you this story? My first semester teaching at Berkeley, fall 2000. I said, I'm going to shake it up here.
I'm going to have an open book exam, untimed. Like, literally untimed. I started at 7:00 PM in Wheeler. Okay?
And — ready for this? It's the craziest thing. So I look around. And by 7:00 PM, by 11:00 PM, there are people still taking it.
Four hours in, for a one-hour exam. But I gave them infinite time. It's 11:00 PM. And the janitor comes in and says, I got to kick you out to clean it.
So I took the remaining — this is at midnight, and I was, 11 students or 20 students there. And we walked. I collected their exams, and they walked silently, single file, back to Soda.
I found an empty room, and they took the exam for two more hours. Now that's funny story number one.
Funny story number two is, after I handed the exam out, I saw somebody in the front. And they took the book out, and they went, "Page 1." And they start reading the book.
Like, they didn't think she — they just figured, come on, open it up. I got infinite time. You said infinite time. I'm going to read the book from page 1.
Page 1.
BORO NIKOLIC
[INAUDIBLE] DAN GARCIA: Welcome to Computer Science. Crazy. I was, don't do that. Don't do that.
That's my advice. Don't do that. All right. So we also have information about the midterm.
Whoop! All right. The midterm is scheduled. Thank you, everybody, for voting what days work for you.
Sunday and Monday are the days. People say, we need a day. By the way, this exam, we think the exam should be doable in about two-ish hours. That's our current thing.
And in fact, we're so confident about it that we are going — BORO NIKOLIC: It's not 14, it's 2.
DAN GARCIA
Yeah, yeah. We are so confident about it, we are going to have every TA who beta tested. And if the TA — we think, I want the scale factor to be four.
Maybe you'll argue the scale factor could be two. But if I tell you two hours, then every TA should finish the entire exam in an hour.
And I don't think that was true for the last Quest, that they could do it in half an hour.
So I want it to be at least two, if I could have them finish even earlier than that.
We're going to say, whatever we have for the TAs, we'll end up multiplying that by some factor to give you more time. Here's the idea. You got five questions.
Many people said, don't give us infinite time. We will take 48 hours. Limit it in some way.
But we also know — you saw from the conversation from Craig — it's appropriate to give people enough time to show what they learn.
I don't want have people rushing through and not be able to finish. That's not appropriate, either. So we're thinking, what's the middle ground?
Roughly, we're going to write it so that it's about 20 minutes a question. So it's about a three-times scale factor for five different questions. So each question can be done in 20 minutes.
If not, if our TAs can't do it in 10, at least a factor of two. If our TA's can't do it in 10, then I'm not going to have an hour.
I will give you more time for that particular question. So there's five different sections. I'd just say, five different sections. I should have edited this.
I should say "sections." Just like before, we had three sections. There are now five sections.
Each section will have a time limit, and we're going to basically have all the TAs take it, figure out when they can take it, and then triple that time or something, and that's how much you can take it.
So we'll make sure that you have enough time. So the time isn't an issue. And then that'll be the limit. So we'll kind of make both sides of the room happy.
The side that says, don't give us 48 hours. We'll use it all up. Okay, we'll limit it. But the other side says, I'm a slow test taker, which I am myself.
Give you more time. I want to make sure that you guys can finish it in there. So that's the idea. Okay?
So it should not be a five-hour exam, but we're going to give you one. And as I said, we'll make that more per question if that particular question is too hard.
You take it any time in a 48-hour period. You can start one question the first hour, and then take a 12-hour break and take the next question.
Whatever you need to do, same in terms of style. The topics are these five topics. It's going to be a Quest redo, basically summarizing all that you saw on the Quest.
Everything up to week 3 and 1/2 or so when the exam ended. Bloop, freeze that. That's one question. There's going to be a float question.
BORO NIKOLIC
If there's a question, is there C? Yes, in the Quest there is C, so [INAUDIBLE]..
DAN GARCIA
Is there C? Yes, there is. So get ready to do that. Make sure you're prepared for this.
There is C. There's a RISC-V assembly question in which you're going to be able to do some coding.
And we're going to have a practice midterm, so you'll see what kind of thing it is. But you're going to be in Venus.
We're going to ask you to do something, and you'll code it in RISC-V. That's the plan there. There also is going to be a datapath control pipeline. That's this material.
And there's SDS Logic, which was last week's material. Okay? That's what I got. I'm with five minutes left, and that's what I have for the plan.
But I hope that addresses a lot of the concerns that the students had in terms of what's on the midterm, what's the scope, what it is.
There's no more — there shouldn't be any more surprises, okay?
And by the way, we mentioned last week that we will be giving you test cases for your C and test cases for your RISC-V. So we're going to say, do this thing, and by the way, here's some test cases.
And I told you last week, we're going to reserve a couple test cases so you don't just hard code those things in there.
We'll try to make it so that the timing isn't an issue, but we've got to make sure you guys can get above the bar and code RISC-V, you're above the bar and code some C.
And again, we hope, we want, Boro and I and the entire staff want you to totally crush the Quest one. The Quest is going to be a full redo plus a curve thing.
So whichever one is better for you, we'll do that. But we want you to crush that Quest we do. And it's all in one question. There's going to be a lot of partial there.
But again, folks, it's got to work. You can't give us C that doesn't work.
If you give us C that doesn't even compile, that's not going to give us — we're not going to fix your code for you. We've never done that for projects.
We don't do that for exams. That's no different than we used to ask at Quest. The Quest used to be, write this line of code. And mostly, we'll give you, if it's mostly right.
No. You have the compiler. You have all the time. Get it working.
So hopefully, you'll have enough time for the C. Again, we're going to take all the TAs, run it through that question.
And whatever they take, we'll triple that or whatever, and that'll be the time you get. So we'll give you enough time to get that working. Make sure you get it working.
BORO NIKOLIC
And keep these questions going. Keep these questions going. We got two questions. Is the grading, testing the same?
I'm not sure exactly what that means, but we will provide the full test coverage.
DAN GARCIA
Right. Right.
BORO NIKOLIC
If that is what is meant. Yeah.
DAN GARCIA
Right. Yeah, we're going to still do the same thing. We're not going to grade it live. We'll pull it out of it.
Drop it in Gradescope. Write some autograders to try to grade it. So very similar way to do this. There's a question.
Will the practice exam be reflective of the actual exam in terms of length and difficulty? The answer is no.
We're not going to — we don't have time to write a second exam and give that away.
So we're just going to give you, like, here's the kind of question we'd write RISC-V. Might be much easier. So it's not at all. I'm telling you no.
No, the practice exam will not be assessment of that. It'll be, you're going to write something. But we're going to make sure the practice, when it gets you practicing with Venus and doing something.
We'll ask you to do something. It may be harder. It may be easier. But it's not going to be the same.
We're not going be able to write the float. I can't write all the float questions. I can't write all the things. All the things you saw, it's just too hard.
Look at our old exams for that kind of thing. We're just going to give you something to give us some basic, a rough test of that.
BORO NIKOLIC
But the point here is it will test exactly the mechanics of the — DAN GARCIA: Yeah. Practice in mechanics.
BORO NIKOLIC
Whatever is in — DAN GARCIA: Practice in mechanics. Exactly right.
BORO NIKOLIC
Yeah.
DAN GARCIA
Yeah. And, by the way, if you had any trouble technically with your setup, I'm letting you know you have two weeks to get that debugged. Okay? Whatever you need to do.
Practice with the zoom. Plug your — I can't tell you how people didn't plug their phone in. What, are you kidding me? Plug your phone in.
Whatever you do, get an extension cord. Plug your phone in. Get it set. Whatever you had trouble, whatever the cause was for the redo, you got to get that working.
I don't know if we have time. But we had to write a sec — we have to write a Quest, and then a Quest retake. We don't have time to write a midterm retake.
We may not be able to do that. So you've got to get this working. We haven't yet figured out what to do if people have technical problems.
But we don't — you have to get it debugged. We had enough time. Quest. Quest, retake, midterm.
You've got enough time to practice, practice, practice on that to get that stuff working. So get your AV, whatever you're doing.
Whether you're zooming or whether you're time-lapsing or you're moving, whatever you're doing, get it debugged. Make sure it works. Make sure it'll work for whatever the maximum time is.
By the way, the other advantage of a time limit per question is, I only have room on my phone for a time lapse that holds an hour and a half.
Well, if no question is more than an hour and a half, then you're fine. That's the kind of thing. So make sure you get that stuff done.
BORO NIKOLIC
Yeah. Exactly. This was one of the reasons we also, why we limited the time. There's a question.
If we get to 100% on the Quest redo coding question, does our Quest score become 100%?
DAN GARCIA
Yes, it does. 100%.
BORO NIKOLIC
That's right.
DAN GARCIA
First time ever.
BORO NIKOLIC
Exactly.
DAN GARCIA
By the way, first time ever. We've never done that before. We've always done a curved thing where we compare you to your neighbors, and if you moved up relative to your neighbors.
It was a "curvedy" way of doing it. Not this semester. That Quest was so low. It's a full clobber 100% map.
So we want you to ace that Quest question, just rock on that. Beautiful.
BORO NIKOLIC
Yeah. And then, that is new to us, and due to the semester being different than any other semester there have been.
DAN GARCIA
Yep. Love it. So the second question. Will the Quest redo C have test cases or just the post-Quest C?
The Quest redo C. The Quest redo does not. The Quest redo has to be exactly as before. So unfortunately, no test cases for the Quest redo. It has to be the same.
People can't get free test questions if you had to redo, but you didn't get it when you took the real test. That has to be the same.
But the Quest re — how to say this. The Quest redo C on the midterm, so the Quest redo on the midterm will have test cases, if that makes sense.
So not the one that's happening this weekend. That's the retake for the video problems. That will not have test cases.
But on the midterm, the Quest redo coding question will have, and the RISC-V will have. So all the coding, we're going to give you test cases.
And we're also going to, as I said, keep a couple ourselves so you don't hard code any answers there. Got that? All right. Beautiful.
Answer live.
BORO NIKOLIC
Good [INAUDIBLE]. So — DAN GARCIA: I don't know. Recording policies will be the same on the Quest. Basically, exactly the same thing.
You'll get the same options. And by the way, I really encourage — you have two weeks. I encourage you to try every option. First, try it time lapse.
Then try it the video. Then try it the Zoom proctoring. Whatever that thing works, make sure it works, beta test it, triple beta test it. Make sure that works.
But yes, the same rough policy as before. Is each section equally weighted? The answer is no. We have to think about that, and think about that place.
Usually, what I'm seeing is, float is one lecture, but it's one question. That can't be the same weighting as the Quest redo, which was like four weeks.
And it can't be the same as RISC-V, which was probably four or five or six lectures, if you add them all up.
BORO NIKOLIC
Seven lectures. Six or seven lectures.
DAN GARCIA
Seven lectures. Seven lectures, right. I, mean that's great. So we can't write — because it's above, it's there, it's the machine.
So we certainly are not going to have equal weighting, but we have to think about the balance of all that.
And probably, if you look at our old midterms, it's probably the similar weighting that we did before. Our float questions were smaller, and other questions were bigger. So think about that.
But it's a very good question. Thank you for that.
The section will be labeled very clear. The question is, will the sections be labeled so we can review the material before taking that portion of the exam? Yes.
I mean, I'm telling you. Here, look, folks. Ready? The topics are there.
The topics are right there. See the bullet? Wait, where's my hand? Where's my hand?
Over there. Right here. There's the topics. Okay?
There's five questions, five topics. We're telling you what's on the exam. It's basically the whole, all the material up to now. So it's beautiful.
BORO NIKOLIC
Yes. So one thing to keep in mind. There are things that leak over a bit. Like in digital systems, you need to digital systems in order to build the datapath.
DAN GARCIA
Right. It's not so isolated.
BORO NIKOLIC
It's like a stepping stone.
DAN GARCIA
Yeah, it's not so isolated, exactly. Yeah, yeah. But we tried as much as we can to put them there, but sometimes they leak over. Boro is exactly right.
Beautiful. All right. Quest 3 grades. I don't know if the Quest 3 grades have been released.
The Quest 3 grades are done. I just told my TA that they just finished them, but I don't know if they've actually pushed them back to release them back.
But the Quest grades, from our point of view, are done. They probably have a little bit of job of turning it, adjusting it, and then pushing it back out.
But thank you for that question. Good, good, good, good, good.
BORO NIKOLIC
And by the way, related to that, if we mess up any of the — if there is a surprise in your grade, in the overall grade, do let us know, because there are some things that have been regraded manually, and we have had issues before with that.
So don't get freaked out if some grade gets lower.
DAN GARCIA
Perfect. All right. So that one's all done. I don't know what the first one.
So it'll be hackers, rank, style, where we can run test cases and see how many non-hidden ones pass. Yeah. The idea is we'll give you something that lets you check all, some basic.
Maybe it'll just be one test, but it'll be a test, at least. And we'll give you that test to make sure it's there. Set it up.
You write your code, and then run it at least through that one. By the way, I would run it on other ones, because that might not test every case, obviously.
But it's going to test enough. It's more of a sanity check there. But we have some test cases. And again, we're going to have the real one that has the big problem.
Does it do the big one that you can't see? But it'll be enough to give you a very good sanity check sent to that. So that's a good question there.
We got that one. All right keep them coming. This is great.
BORO NIKOLIC
On the Quest, what if we only know we did it wrong — DAN GARCIA: Did wrong.
BORO NIKOLIC
A little hard to fix it.
DAN GARCIA
How hard to fix it. Could we get autograding test after Quest? We do. I don't know what that's asking.
BORO NIKOLIC
It's asking us to release the alpha beta test.
DAN GARCIA
The test cases, yeah, yeah. We may release that. Let's think about it. We may release it after we do the Quest redo, because it might be a similar thing.
So we'll talk about that. It's a good question. Boro, let's try to remember to bring that up to see if we would release the test cases for that.
Thank you for that question. And on the Quest, the sections labeled 1, 2, blah, blah, blah, on the exam section will be labeled C, for example, because [INAUDIBLE] from the start.
I think, I mean, I don't mind — I don't know, Boro, if you have opinion on this.
I don't mind labeling them sections 1 through 5, even calling it Quest redo question, float question, RISC-V assembly question, so you'll know what you're getting.
I don't need, I don't think we need to surprise people.
BORO NIKOLIC
Yes.
DAN GARCIA
So I'm okay to label them.
BORO NIKOLIC
We usually label them. One thing, though, there is this leakage. So don't expect all — can't ask us any additional systems in datapack. It is a digital system.
So there will be a little bit of that. But yes, the main theme is going to be labeled.
DAN GARCIA
Yeah. I don't have a trouble with labeling the main themes. Yeah, that'll be easier for that.
BORO NIKOLIC
There is a question — DAN GARCIA: And some students, they gave a handful — BORO NIKOLIC: [INAUDIBLE] separate.
DAN GARCIA
Okay. Go ahead. A handful of students actually didn't realize that there were three sections, so we're going to make it clear. There are five sections, each labeled.
Make sure you finish all five sections. You're not done with the midterm until you finish all five sections. Thank you.
BORO NIKOLIC
Yeah. And we need to wrap up in a minute or so, because I think there is another class, I think, after us.
DAN GARCIA
Sure.
BORO NIKOLIC
There's a question. Do we have any extra credits for this class? And we do.
It's called the EPA. Dan, do you want to explain?
DAN GARCIA
Sure, sure. We mentioned in the first lecture, EPA is Effort, Participation, Altruism. There are ways for you to show excellence and show that you're helping other students. In the altruism case, you participate.
You're coming to these sections. You're coming to these things. You're coming to office hours. People are getting to know you.
E is for effort. Effort means you do every assignment. You don't blow any labs off. If you show that you've really put your time into this class, there's ways to get that.
As well as if you take the performance. If you try that, there's a chance for the competitions. If you submit a competition, it means you're putting extra effort in.
So we'll see that, and we'll give you some extra bonus. So there's some extra bonus there with that. We'll play with that as well.
Exact timings for each section will be visible before starting a section. Yes. I don't see a problem with labeling that so they'll know how much time to take for that.
Set your timer, blah, blah, blah, boom. Cut that there. Good question, Benjamin. I love that.
That's a good question. Discord might be kind of an EPA if there is a way for us to kind of get some numbers about it. That's a way for our TAs.
At the end, by the way, it's an internal number. And if our TAs. Can get to know you through Discord, then certainly, discord is a useful as a mechanism to get to know you.
Just kind of like office hours.
BORO NIKOLIC
Very much so. I mean, we're going to ask our TAs to tell us — DAN GARCIA: Who's bubbled to the top? BORO NIKOLIC: --what they've learned about [INAUDIBLE]..
DAN GARCIA
Who's bubbled to the top, exactly.
BORO NIKOLIC
Always visible.
DAN GARCIA
I'll save that one, [INAUDIBLE].. For later. That's a little secret thing all right. Good.
BORO NIKOLIC
All right.
DAN GARCIA
Thank you. Thank you so much.
BORO NIKOLIC
Great. Thank you.
DAN GARCIA
Well, it's great to see you all. We'll see you next week.
BORO NIKOLIC
[INAUDIBLE]. There was a lot of questions. This is why we have these live ones.
DAN GARCIA
Yeah, exactly.
BORO NIKOLIC
Questions.
DAN GARCIA
This is exactly. We love that connection.
BORO NIKOLIC
Yeah.
DAN GARCIA
All right.
BORO NIKOLIC
Thank you.
DAN GARCIA
I'm good. Thanks, Lois.
BORO NIKOLIC
And we have exciting — more exciting guests are coming. So get your — people from industry coming, and there will be people from academia.
DAN GARCIA
And by the way, people have done very well in the projects. So everyone's done — I mean, I know the question was a challenge. We're hopeful to get them replaced with the midterm.
But I think folks are doing great in this class in terms of the scores I'm seeing on the project. Boy, have you seen those numbers in the survey?
The median is 100 for every single section.
BORO NIKOLIC
I thought median is 100.
DAN GARCIA
For every single one. Project 1A, project 1B, project 2A, median is 100. Keep it up. We want that, folks.
We don't need to — I want to give you all A-pluses. It's not about making a, reading you out. Come on. We'll put a bar.
If you all go above the bar, we're happy. We all did our jobs. So keep it up. Keep on.
Keep helping each other, right? That's why we had you work together. Help each other, teach each other. Learning community, we love it.
All right. We'll see you next week, folks. Take care.
BORO NIKOLIC
All right.
DAN GARCIA
Thanks 2 and C. Whoo!
BORO NIKOLIC
Thank you.
DAN GARCIA
See you, folks. Take care.
BORO NIKOLIC
See you, folks. Bye.
Đăng nhập để đánh giá học liệu này.
Đăng nhập để bình luận — chưa ai bình luận về học liệu này.
Không có khóa học liên quan.