I'm in the middle of watching the show Start-Up on Netflix. I just finished episode 5 (some spoilers). Some background: Seo Dal-mi and Do-san are our main characters. They and Do-san's machine-learning-developer buddies are all on the same team at a hackathon. At the end of the hackathon, each team presents their product and business plan to the judges and an audience, and the top 5 teams (out of 40 - so this is pretty competitive) will be admitted to Sandbox, a fictional Korean Y Combinator. Seo Dal-mi has spent years working at crappy part-time jobs and entered the competition because she was inspired by Do-san's "company" to start her own. Do-san and his friends technically do have a company, and are very good at getting deep neural nets to attain state-of-the-art accuracy on classification tasks, but are terrible at every other aspect of business, and their company is currently falling apart. Dal-mi's sister In-jae is also in the competition, but the two have a terrible rivalry, went 15 years without speaking to each other, and mostly despise each other. In-jae forms a different team and, out of spite, does her project using the same dataset that Seo Dal-mi's team chose: a computerized collection of handwriting images from a bank.
What follows is the best machine learning dramatic face-off I've ever seen in anything. Dal-mi's team has made an excellent forgery detector: it takes handwritten text as input, and outputs "forgery" or "not a forgery" as output, with super-high accuracy. In-jae's team, meanwhile, has made a font-generator: it takes some sample handwriting as input, and outputs a full custom font-set, that you can then use for whatever digital application you want, like design. The presentations are going straightforwardly, until a judge asks... what if we put some writing from this font-generator into the forgery-detector? Would it identify it as a forgery? The two teams go on stage to set it up, In-jae's machine-fabricated font is fed into Dal-mi's fabrication-detector, and everything is on the line: the fate of their projects, and whether they get into Sandbox, all depends on what the detector network outputs.
This is the best. I've heard of this type of contest before, in real life: the contest between email spammers and spam filters; between doctored photos and videos and the ability to detect the doctoring; between auto-generated text in comments sections and the ability to identify it well enough to auto-collapse it. But I'd never seen it dramatized in media, and was wonderfully surprised to see it dramatized so well. The history and animosity between the two sisters, and the stakes of trying to win entry into a start-up incubator, makes the face-off between the two deep networks something I can feel, and puts it in the form of interpersonal conflict, which works very well in television.
There's more. Seo Dal-mi is the CEO of the hackathon team, but knows little about machine learning. A few minutes before she's to present to the judges, Do-san's teammate gets nervous and starts to freak out. He and his two developer friends have won an international contest in this machine classification stuff; Seo Dal-mi first learned about machine learning a few weeks prior. He says Do-san should present instead: "she doesn't even know what a hyperparameter is!" As the viewer, at this point, I'd already seen these guys bungle business opportunity after business opportunity, and understood how bad they were at presenting this kind of thing. But at the same time, I've felt this same instinct: it's late in the game, the next step is critically important, and also the next step is out of my hands... and I want to take it into my hands! But running around trying to do every important thing yourself, instead of letting your teammate do it - even when you have no reason to think you're any better at it than your teammate - is a big mistake. "Do your best and then trust your teammates" is a lesson I've learned, and that has served me really well over the past few years at work, and I love seeing it dramatized here in a context I'm so familiar with. Later in the episode, Dal-mi passes the same test: when she's asked whether she thinks their detector can detect the forgery from her sister's network, she says: definitely. She has no clue, of course; this is a hard question even for someone skilled in the field, and she isn't. But she's the CEO, and she was asked in extreme public whether her team's creation was good enough, and she had to choose: do I trust that my devs did better than their devs, or not? And she says: yeah, we did better. Definitely.
My favorite scene is when Seo Dal-mi actually presents their model and business plan to the judges, before the face-off. She does a great job, because she motivates the "detect forgeries" task so well. A recurring frustration in my life is that I can always think of cool technical stuff to do, but can much less often see how I could make it directly useful to others. Do-san and his developer friends have this problem even worse: they are totally unable to explain or even reason about the ways in which their projects might be worth money. Dal-mi's presentation makes a knock-down case. She motivates the problem before mentioning any part of the solution, listing out all the institutions that use handwriting for verification purposes: banks, prosecutors, the National Foresic Service, and the National Tax Service (and of course, all these are possible customers for a forgery detector). She uses concrete numbers: - "the rate of forgery is 8%"; "there are only 20 forgery specialists registered with the court". The show's filming of her presentation is intercut with shots of Do-san realizing how poorly he's been doing on the selling and motivation part of business, and with shots of the judges seeing how her reasoning makes sense and how the thing could be worth real money. And, uh, it's all set to sentimental Korean-drama music, which also works pretty well. I literally teared up during this scene. But I also felt like Do-san: I've been to hackathon's, and I've worked really hard, and then me and my team have done a bad job presenting why what we did is useful. I think we've even gone into detail about technical details of the technologies we used, without prompting, which is... not advised. So this is close to home. Seeing a heightened-reality, super-dramatic presentation of a hackathon project is nice and emotial, but also super useful. I love dramatization of normal activities that actual people can actually do. Like: I can actually go to a hackathon and write up an ML algorithm, or present a business plan, just like in this dramatic scene! It's like if, after watching a Fast and Furious movie in theaters, you could actually drive 100 miles and drift under semi-trucks on the way home. I can be this; I can do this.
There is a broader theme here. The show takes for granted that tech can improve the world, and that succeeding in business and at work is a worthy goal. I feel the same way. But it can be hard to find these messages in art. So when I do find a good show or book with these messages, I get way into it. Other recent fiction that dramatizes work in ways I really like are the show Industry, and the novel Sweetbitter.
Industry follows finance interns working on the trading floor at a fictional Goldman-Sachs kind of place. Banks can be unpopular, but this show does not hate banks at all. It frames working on the trading floor as a perfectly reasonable thing to do, that smart and kind people could reasonably choose to dedicate themselves to. Characters do overwork themselves, and yell at each other, and do too many drugs, and get into serious trouble - all the high-finance stereotypes are here. But since the viewer is always following the main character, and the main character has no doubts about her goal of working at the bank, or about the banking system in general (in one scene, annoyed by trash-talk about capitalism at a party, she snaps back that she wrote an 8,000 word paper on the moral case for capitalism), there is no feeling of "banks are bad". Instead, it's more like "look at all these interesting dramatic problems these people encounter work through at their job". In this show, the job is the setting, not something to be discarded in order to follow your real dreams. Success is celebrated.
Sweetbitter is about a girl who moves to New York City to... I'm not sure what, exactly. I don't think she knew either. She becomes a server at a restaurant. But not as a stepping stone to some higher goal. At the interview, she is asked: Why New York? She thinks:
I knew so many said: I came here to be a singer/dancer/actress/photographer/painter. In finance/fashion/publishing. I came here to be powerful/beautiful/wealthy. This always seemed to mean: I'm stopping here to become someone else.
I don't know what it is exactly, being a server. It's a job, certainly, but not exclusively. There's a transparency to it, an occupation stripped of the usual ambitions. One doesn't move up or down. One waits. You are a waiter.
The book is all about the restaurant and the people the girl works with there. She starts as a back-waiter, and wants desperately to be promoted to a full server. She is laser-focused on learning the trade; she spends serious amounts of her free time reading books about the minutiae of wine-grape-growing. She looks up to a server at the restaurant, Simone. As she becomes closer to Simone, and better at her trade, her confidence grows, and by the end of the book she is not the child she was when she arrived. There is a love interest in this book, but on my reading, he isn't what changes her from a child to a woman. She gets that growth by gaining mastery in her trade. (Sweetbitter is my favorite book of all time and I highly recommend it.)
In all these media, there is a lot going on besides the work stuff. But in each, the core motivation of the characters is inextricably tied up with work and competence at it. Being good at stuff that people pay you to do means so much to these characters, and as a viewer, you can feel it. This is art that says: work can be important; work can be worth it; and work can be awesome and dramatic. I love this message.
'Competence' is one of my favorite subjects in narrative media; probably because I too enjoy gaining it myself both at work and outside of it.
Yes, but what tools do they use?
Python, actually! (Who would have guessed?). The camera zooms in on Do-San writing correct Python every now and then. I mean, he keeps writing a function called sigma_prime, which, like, maybe he should import? But it is tech literate even there!