Recently, the bad reviewer encountered a strange thing when recruiting people-

Suspected encounters with AI face-testers.

A young man (hereinafter referred to as "Little A") submitted his resume, and his job search direction was science and technology editor.

I asked Little A to write a trial essay about the tech giant moat. Although Xiao A's style is not very compatible, the angle is quite comprehensive, and you can interview a wave.

Because I was abroad at the time and HR was in the company, I made an appointment for an online video interview.

During the interview process, I was not very satisfied with Xiao A's performance, and there were a few doubts↓

In fact, during the interview, I was already a little resignedAI 面試不cheatI thought about this piece, and a message from HR after the interview made me more sure.

HR said that the opposite network was down several times.

Hiss, obvious sense of reading the manuscript - scientific and technological common sense that contradicts the level of the article - suddenly disconnected when answering questions several times -

Is this little A using AI to assist in the interview?

The so-called disconnection is just a disguise of the time when the AI outputs the answer.

Of course, this matter, the bad reviewer can't really hammer it. In order not to affect the parties, I have replaced many of the details.

You can listen to it as a story, but there is indeed a not-so-good trend behind the story:

There will be more and more AI interviewers around you and me.

On social media platforms, you can see it by searching for the keyword "AI-assisted interview".The use of AI to cheat in interviews has become a phenomenon.

Someone shares how to use AI to answer questions for you in an interview, and someone teaches you how to cheat in an interview so you don't get noticed.

And on the other side -

There are also HR who are complaining about whether everyone can stop using AI for interviews, and most of them can be seen;

There are also interviewers who write a long essay directly《 我和 AI 驦的攻防戰 》to share how to hack AI cheats.

In fact, the reason for the emergence of this phenomenon is not difficult to guess-

There are a bunch of AI "assisted" face-to-face apps on the market.

These apps upload live videos on major videos and social media platforms with attractive titles to promote their selling points.

What is "remote screenshot written test black technology", "Java back-end job hopping success 35K".

The reviewer has probably looked at a few apps, and the functions are almost the same, and they can be used with their hands-

Open the app before the interview, and hide your phone next to your computer.

The app listens to your voice, transcribes the interviewer's questions, and then uses AI to present the answers in real time.

All you have to do is just read it.

Even if it's a programmer's technical written exam, these apps seem to be able to cope with it.

Direct access to a remote screenshot function, automatic identification of questions, answers in seconds.

Of course, anyone can brag, so the bad reviewer placed an App to test.

Don't say it, the functionality is quite complete.

It will ask you to fill in the job title, responsibilities, and requirements before uploading your resume. This way, the AI can make the answer more relevant to you based on your experience and strengths.

In case you don't know how to use it, the app also has a "mock interview". It is possible to choose the interviewer's personality, question preference, and length of the answer.

After starting the simulation, the interviewer asks a question, and the answer generates a sentence, and you just read along.

Finally, the app will also let you read a piece of text.Add your voice signature。 The bad reviewer guesses that this is to distinguish your voice from that of the interviewer and make it easier to identify the question.

There may be a poor friend who sees this and is a little moved.

But I advise you to use it,Practice acting skills for two and a half years, prepare 50 bottles of fast-acting heart pills, and finally prepare yourself to be a leek.

First of all, you'll need to top up your money to use the app's "Formal Interview" feature.

Half an hour 188,0 hours 0.

It's not a buddy, I understand the money.

But ChatGPT and Cladue are all calculated on a monthly basis, and it is only 120 in a month, and you are 0 hours and 0, is it an exaggeration.

I want to have 2 more rounds of interviews, isn't it gone to be a few hundred?

Okay, it's all here, buddy will be a good leek today.

The bad reviewer pulled a colleague as an interviewer, and I was responsible for playing a person named Wang Haiyang, preparing to interview an algorithm engineer.

After the "Formal Interview", the bottom line of the app will be instantaneousRecognize the interviewer's voice, and then生成問題lastPresent the answer, I'll just pretendRead as it isThat's it.

It looks OK, and the recognition is quite accurate.

But the problem is, the latency is severe.

In normal human communication, one party can answer the question within 5.0 seconds after the other party finishes speaking (after all, when you hear more than half of many words, you know what they mean).

But this interview assistant's AI requires the interviewer to finish the conversation thoroughly5.0 secondsregeneration problem,再過5.0 secondsout of the answer.

When you see the answer, you react and start reading, and you have to ask for another one5.0 seconds。

So, there is always a delay of 2.0 seconds ~ 0 seconds when you answer a question, and most people can see such a high ping, right?

You said you can act, you can make excuses.

Indeed, when the bad reviewer was in the "interview", in order to cover up the delay, sometimes "emm + cheek support", sometimes "sorry, the network is stuck"...

But you can't do it every time. You are not embarrassed, the interviewer is embarrassed.

Even if the latency is high, it has one of the most fatal problems:Flip the car。

Maybe I didn't read it all in the original text, or maybe the sound recognition was not accurate, but in short, the app was used several timesIdentify "my words" as "interviewer's questions".

I was reading the answer, and the app suddenly refreshed the answer!

Brothers!!

Originally, the interview was nervous enough, I was still cheating, and the "artifact" suddenly overturned again.

Just ask who doesn't have a brain down and whose CPU doesn't burn at the moment. If it were me, I'd unplug the network cable too.

After the interview, the bad reviewer asked the interviewer how he felt, and he said: I almost didn't hold on.

To sum up, the bad reviewer really thinks that this thing is not good and unreliable.Not to mention cutting leeks, heart disease can be committed several times.

That being said, there are already plenty of live examples of AI interview cheating online.

The programmer forum V2EX has encountered this situation with many old brothers.Everybody said,

用 AI 作行太明顯了。

When the other party answers a question, he is often stunned for a few seconds, and then suddenly his mouth is like a river, gushing, his tongue is like a lotus, and he nods his head with a stubborn stone.

But one expands to questions that have no fixed answers, and one question does not squeak.

The same thing happened abroad.

A remote working company in Poland encountered a "sweet spot" when recruiting people.

This buddy's resume is invincible, graduated from a prestigious university, and has also worked in well-known companies, as well as in technical interviewsA person who completes all programming tasks within the first 2 hours.

The boss almost made an offer, but when I finally interviewed him, I felt that this guy's background was really a bit suspicious. For example, he graduated from the Warsaw University of Technology in Poland, but he doesn't speak Polish, forget it.

The most annoying thing is that 2 months later, the buddy used DeepFake face swap technology to come again, and the resume and technology are still indiscriminately killed.

How did you find out? The boss told him to spread his hands in front of his face.

As a result, the man refused.

Of course, there have been people who have succeeded in cheating with AI over the years.

一個舅 Roy Lee 的哥倫亚大学學計算機系學生,就需要自己開發的 AI 工具,前後得 Amazon、Capital One、Meta、TikTok四家大頭的 Offer。

If it weren't for his werewolf self-detonation, I guess no one would know about it now.

Some of your friends may see this and say:

AI cheating is really wrong, but what does it matter to me, isn't it HR and boss's problem?

Let's look at it from another angle:

A person whose level is a lot worse than you, relying on AI to cheat, earns more than you, and you don't agree.

Although there are still many HRs who can see AI interviews -Look at the candidate's response delays, speaking style, and ask more detailed questions.

But we also have to take it into accountAI is still evolvingandSurvivor bias- HR who doesn't recognize cheating won't say it online.

The bad reviewer thinks that if it goes on like this:

Sooner or later, HR will be forced to learn some skills to recognize AI, and even interview software will have to be connected to AI expression judgment and eye tracking systems.

AI offensive and defensive battles, you don't have to fire the first shot in a workplace interview, right?

When writing the article, the bad reviewer saw someone asking:

Why should cattle and horses be embarrassed by cattle and horses. Isn't it awesome if a person can solve all problems with AI, and doesn't that mean he has the ability?

At first glance, it may seem impossible to refute it, but there are too many loopholes.

First of all, the real interview is an exchange of ideas and a demonstration of strength, not a "better" AI tool than anyone".

HR wants to know your thoughts, but you give someone else's voice – it's not an aid, it's all soul outsourcing.

其次, AI 不可能solves all 問題。

An interview is an interview, and the fact that AI can help you pass the interview does not mean that it can handle every project in the actual work, and every Party A who makes decisions with his ass.

You have to know: people have "one pair" of eyes or "two pairs", "everyone" is followed by "good" or "bad", these answers are just AI from the database by calculating probability,Guess itThe best option.

Guessing the wrong answer, hallucinating, and talking nonsense are also the norm for AI.

This kind of external skill, which is full of uncertainty and instability, cannot be called one's own ability, let alone completely entrust one's daily work to it.

Otherwise:

Every time you work, you are using AI as a wishing machine, and then you put your hands together, like a cyber nun, and pray quietly in front of you, which is the right answer.