Deep Dive
Shownotes Transcript
Discussion keeps the world turning. This is Roundtable.
You're listening to Roundtable. I'm Steve Hatherly, together today with Niohung Lin and Fei Fei. Coming up, the hiring game has new players. AI recruiters are flipping the script. Blink and you'll miss it candidate sorting? Yup, and a whole lot more. But behind the speed lies a minefield of ethical questions. We'll dissect the good, the bad, and the algorithmic.
After that, "Hey, are you free for dinner tonight? You want to join me? Maybe we could go out for some Buddha jumping over the wall." "What's that?" you ask. "Oh, okay. How about some husband and wife lung slices?"
Translating Chinese dishes is no easy task. We'll see if we can tackle that problem today. Our podcast listeners, that's you, can find us at Roundtable China on Apple Podcast. And don't forget, we love hearing your voice. You can send us an email with a voice note. That would be wonderful. Anything that we discuss on the show, you can send that topic along. Roundtablepodcast at qq.com. One more time, roundtablepodcast at qq.com. And now...
The way companies find and hire new talent is changing, and it's changing fast. We're seeing a big shift with artificial intelligence stepping into the recruitment process, bringing some truly amazing levels of convenience like faster candidate matching and much more efficient screening. But as is the case with any powerful new tool, there are also some real downsides and some important questions that we need to think about. For example,
How does it impact fairness? And what about the potential for new types of fraud? We're going to talk about this new era of hiring and what it looks like, where it's headed, and maybe some of the surprising or perhaps even not so surprising challenges that it's bringing up. So when you were doing your interviews, either for this company or for any job that you had, can you kind of walk us through what that experience was like?
Well, first of all, you need to have your resume and your CVs ready. And then you, well, depends on the age. I think in the old days, you will need to mail some of the CVs out. And in these days, you are sort of dropping them, sort of checking up on, then later you need to email them in their email boxes. And these days we have like different online platforms. You can just search for the job that you want. And they sort of,
Have different requirements. Yes, and then you can just drop your online digital format of CVs into their sort of inbox on that online platform.
So that's kind of the way it was everywhere, I think, right? And then you go in and you have an interview, and if you pass, you might have another interview. That still exists, but the process is vastly different now for some companies in New York. So how is it getting different? Yeah.
Well, it's different because AI is being brought into the process. It's easy to imagine some things that AI can do. AI can speed up the process because when they're selecting potential candidates, if you have a deal breaker in the very first place, you tell your AI screener they can definitely cancel out or just omit the CVs they received and be
based on your deal breakers. They can speed it up. They can streamline it because they can arrange the CVs that a human HR would see in the right list. They can give you the best candidates to the least possible candidates. At the same time, they can also fix problems for the problems that might be brought into the process because of human bias.
So this is also easy to predict. But they can also do things like they can provide possibilities that human HR can't really do if you use an AI interviewee because they can analyze the facial expressions. They can analyze the way the interviewer is using the candidate.
candidates' language patterns and correlations, and these things can hardly be analyzed by a human HR on spot, yet it can be achieved by an AI interviewer. So the process is vastly different.
Yeah, the simple stuff is keyword matching, right? So you can use an advanced algorithm to process different data points for candidate profiles, specific skills listed, educational background, work history. Yeah, that's the easy stuff, right? But the AI can now predict, they say, how well someone might perform in a job.
They can recognize patterns and correlations within data. And the AI looks to forecast potential job success. And the goal here is to help pinpoint top talent more efficiently and ideally, thereby reducing the hiring bias, as you said, Ngo Hung Lam,
by carrying out objective criteria. It's basically automating and refining the early, very, very, very labor-intensive stages of talent acquisition, and it frees up human recruiters to focus on more complex, more human-centric tasks. And you know how crazy it is?
In theory, these AI recruiters can get better and better because they can learn from their successful examples and they can learn from their mistakes. If they can successfully hire 10 people who are perfect for their job and failed on two cases,
And based on their past experience, the data they have, they would know how to do a better job next time and the next time and better and better. And another thing I would argue is that with AI powering so many stages in these hiring processes,
And I guess it's a good thing for HR themselves that they don't need to go through piles and piles of resumes. They can focus on more sophisticated jobs. Yeah, because if you think about it, okay, if you're applying, if you're a teenager applying at a small mom-and-pop convenience store, these types of tools are not necessary, right? They might get 13 applications, something like that.
If you're applying to a conglomerate, right, and there's maybe 50 jobs, how many resumes do you... How many applicants do you... Thousands. Thousands. Hundreds. Right? Without tools like these in place, it would take...
so much longer, the efficiency would be near the bottom of the scale as opposed to near the top. You know, here in China, in a conglomerate making or manufacturing cars, they say that with the help of AI interviews,
they can interview people as many as 100,000 a year. So that is definitely something that only with technology that we can do. Yeah, that's true. Oh, so it is happening here. It is happening here in China. It's happening here in China. In different ways. I watched, you can go online and you can see some clips of some people who are...
recording themselves during an AI interview. And it's quite, I don't know how, I don't even know. It is what it is, right? You're sitting in front of a computer screen and the AI is asking you questions. And I think I was watching some sort of engineering position interview and the AI was asking very technical questions about trying to assess, you know, do you actually know the answer to this question or do you actually not? The response has been, yeah,
split, I guess, from the responses that I saw online. Some people think it's a good thing to go through a process like this. Others don't. So with the promises, I think they're quite obvious, aren't they? From the company's point of view, it can save them a lot of time. It can save them a lot of money.
What are some of the problems though that we're looking at here? Because I think there's definitely more than one. From the example that you were talking about earlier, whether or not this person has the knowledge of the specific questions the AI interviewer is asking.
From right there, if I am facing a computer and I'm answering the questions, specific technical questions, I can certainly glue a cell phone in front of me and use another AI tool to answer for me. All I need to do is to read out what's on my smaller screen and look at the bigger screen a little bit. Definitely.
hiring fraud would be a growing concern. And that is only one situation. Furthermore, the rise of generative AI means candidates can now easily create highly convincing but entirely fake qualifications. So those would be
hurt the trust system in the hiring process, making it harder and harder for employers to improve the process to make sure, for example, in the past, maybe we'll be happy to allow a
tele-interview because the candidate is far away and living in a different city or something. But now we dare not to do that because we want to make sure they're not using it to help them cheat in the process. So it is making things difficult for everybody if this continues to happen. You could argue then that that's actually making HR's job harder. Yeah.
Because when they get their final candidates, do they actually have the best candidates available? Yeah. I actually saw a video online a couple of days ago. Someone is definitely reading off their AI tools and the EHR, they're video chatting. So the EHR is asking, are you using AI? See, that's the thing.
And that interviewer said, no, no, I'm not using it. That's the thing, because a human HR would still sense a little bit of the weird reaction from, well, at least for this stage, I'm not really sure in 10 years later, AI can be much smarter than I am. But at this stage, a human HR would sense the weirdness in the conversation, but an AI HR cannot unless you specifically program it to be. And
programming it to be is also part of the problem because you said, and that was a fair point you made about AI interviewers, some bias can be eliminated there, but some bias can also be reinforced there as well because AI is simply trained on historical data. We know that that already exists. So if there were cases of
discrimination on gender or race or age, then the AI is going to learn that. It's going to perpetuate that problem and make it even worse. That being said, I still think every case could be different and every interview can be different. An interview based on the type of the interview and the aim of the interview, because like Steve said, if you send your CV, you are going to go from interview to interview because different interviews have different aims. If the
interview is to test out the knowledge, the technical or professional knowledge of the candidate, it can, in the perfect case, one, the interviewee is not cheating, it can be carried out by AI interviewers quite well. It does not have to make a real human being sit in front of a
a computer to ask those questions. And also for some jobs, the interview is to only to select those who do not have the deal breakers. So instead of finding the best out of the best, they're just trying to eliminate those who are not okay for the job. I get it. A massive hiring process. I get it. I get it, but I don't like it.
Because it removes the human element, right? Exactly. We've heard this. Hire the person, not the resume. And this removes that opportunity for the human to make a great impression. We're now having to impress a machine before we can impress the person who makes the decision first.
on a career development possibility for me. I kind of get what he's saying. Especially for like creative jobs that requires a lot of creativity and also, for example, human interactions. It's not something that you can program. No, I think it's for all jobs. I think it's for all the positions because we love a story.
where this candidate goes into knock on the door of the CEO and saying that I do not hold the degree. I do not have the skill, but this is what I have. And I have passion and I want to work and study on the job. And I promise you after a year, I will be better than 90% of your employee. And he gets the job and becomes the next CEO or something. We love that relationship.
That story. Now, will AI be able to listen to and evaluate? Precisely. I mean, maybe in the future. But humans evolve. You are a different person, for example, 10 years ago. You change, you evolve. That's not something AI can predict. And so I think AI is just a tool to help HRs, to help companies or help candidates to
But it can't be the only platform, especially in this process. Interesting times we live in. Interesting times. That is for sure. Coming up, the translation of Chinese dishes into English. Always challenging. Often entertaining, actually, from a non-Chinese speaker's point of view. But should we discard dish names for simple descriptions of the ingredients? I don't think so. So much of that is lost when we do that. We'll talk about it.
Looking for passion? How about fiery debate? Want to hear about current events in China from different perspectives? Then tune in to Roundtable, where East meets West and understanding is the goal. ♪
You're listening to Roundtable. I'm Steve with Fei-Fei and Niu Honglin. Translating Chinese dishes into other language is a linguistic tightrope, I've seen it described as that, that is almost impossible to walk. How do you capture the soul of a dish like, help me please, 佛跳墙? That's translated as Buddha jumps over the wall. How about this one? The direct translations, red braised pork. Well, that's 红烧肉.
But when you translate it to red braised pork, there's no poetry in there. And Buddha jumps over the wall doesn't really make any sense to a non-Chinese speaker or to someone who doesn't have any familiarity with Chinese food. Translations are almost impossible, even for the ones who are best at it when it comes to dishes. So how can we tackle this problem? What do you think?
I don't think there is a one-size-fits-all solution. It depends really on the dishes. And even for one dish, there are different regional varieties and there are so many culture and historical things going on behind. It's just...
you just to look at it case by case, dish by dish. For example, some would argue like the Shizitou, some would argue just keep the Chinese original name Shizitou, but I think it lose the... The grandness. Yes, I actually personally prefer the English translation as a lion's head, at least.
get some of the mystery and some of the charm behind that is the thing i do agree with feifei because i think we when we talk about shizutou when we talk about those are the meatballs right meatball huge meatballs well bigger really bigger ones and when we talk about fuqi feipian the uh examples yes husband and wife lawn slices it's lawn slices created by a husband a
pair of couple, husband and wife. It was not from them. So that is the idea. The reason that as a Chinese person we know this is only because the dish is very famous among all Chinese. It's not because as long as you speak Chinese, you would be able to know what the dish is. And this happens in a lot of other
Not that universally, by saying universal, I mean around China, universally known dishes. For instance, recently we visited Guizhou. It's a business trip. They have a local specialty called Si Wawa. From the surface of it, as a Chinese person, I can only understand Si as thread and Wawa as a baby.
So I have no idea what that dish is until I arrived and I realized it's... They would slice the ingredients into threads, so 丝, and then they would wrap them up like a taco kind of wrap, but you wrap a baby. So that's...
See, we do not find it improper for the dish to be named as Sibaba in Chinese. Why are we talking about when we translate, we would have to change the nature of the name only because people might not understand? I don't see it as a problem. I don't think we're talking about Chinese dishes overseas, right? I don't think we're focusing that. I think what we're looking at here is for tourists who are coming into China and perhaps
trying dishes for the first time, right? So which side makes more sense? Do you keep the poetry of a dish
you know, when you talk about Buddha jumps over the wall, that's beautiful. A dish called Buddha jumps over the wall. But what is that? So for me, if I'm a tourist and I see that on a menu and there's no picture of what it is, I'm not ordering it because I have no idea what it is. Add a picture. Okay. Now, but the other side of that is, okay, xiaolongbao. Now that translates to, that was one of my favorite things in the world. That's,
That translates to soup dumplings. Well, that's boring, isn't it? Right? So keep xiaolongbao because it's beautiful. It's artistic. It's not that hard to say for a non-Chinese speaker. So I think I agree with you, Fei Fei, when I say I don't know if there's one answer to all of this because it's...
You know, some cases it might work, some cases it just might not work. My suggestion is add a picture and use the original name. Sometimes the pictures don't help a lot. And also in like Italian or French restaurants, they'll have like smaller lines of notes to their dish, like how they cook it, what ingredients they use. It's also something that we can use.
So I don't accidentally order four pieces of music played by violin, but no dish at all for my meal. And also another thing is I realized if you keep the mysterious element in the name of the dish, it would provoke conversations and it would allow server to give you the background story of the dish. Because I have encountered a dish called global governing dish.
And you would never know what global governing is as a dish. But I was in Shandong province. And Shandong province is the hometown, was the hometown of Confucius. And in Chinese, there is a saying that with only half of the analects of that is the saying of Confucius, you would be able to govern the world.
the entire area under heaven. So 半步轮雨至天下. And the dish you get is a tofu cold dish, but they basically print the pattern of a 轮雨 of a Chinese annulet on the tofu. And the idea is poetic. The dish might not be that significant, but if
visit Shandong and you order that dish, you would get the story, which is more important. Which I love, and that's a great story, right? But here's the challenge. Okay, so if we combine some of these elements, let's say that for...
Gong Bao Ji Ding, which is stir-fried spicy chicken with peanuts. Let's put the Chinese title, the transliteration I think that's called, where you just kind of phonetically write it out in English. Then let's put the, you can keep the name, then you put the description of the dish, right? And then maybe let's tell the story behind that. And then, right, okay, that's perfect.
Now, how do you do that on Wayshin when you scan the QR code? You can't. You can't put the whole story in there. There's no space for it. So that's why this is an even bigger challenge, right? I'm old school. I like a menu where you can have all that information. But it's 2025 and everything's QR code here now. So don't you think that just makes the problem almost impossible to solve?
I think so. Especially if you consider, you know, there is a difference between regions and provinces and cities in China, like mantou, the steamed bun in northern China is just steamed bun. But in southern China, it means baozi, you know, buns with stuffings inside.
It's just impossible to translate. My suggestion would be create an app with all this information in it, with the stories, with the ingredients, with the real names. And you can jump from your ordering app to this app to get the information.
That's an enormous task. You should actually do that. That will take you 20 years to accomplish that. It's a challenge, isn't it? It's a challenge and it's not an easy solution. Otherwise, the solution would have come to us already, right? That will do it for Roundtable for today. Thank you so much to Fei-Fei and Niu Hong-Lin. And thank you to you, as always, for sharing your time with us as well. I'm Steve, and we would love to have you back again next time.