Google’s AI can’t help with stupid HR questions

I discovered this week that you can practice for a job interview with Google AI. Not that I am looking for a job, but I couldn't pass up the opportunity to converse with Google AI. What new skills could I learn from it?

I clicked the link to Google’s interview warmup and was greeted with this copy: A quick way to prepare for your next interview in [Different Roles Fade in and Out]. Practice key questions, get insights about your answers and get more comfortable interviewing.

So far, so good. There was a short overview of the process. When I was ready, there was another button to click to start this interview warmup.

First I had to make a choice between the type of job I was going to interview for. There were 6 choices, ranging from Data Analytics to UX Design. One choice was a little bit strange. It was 'General'. At first, I thought it was for a position in the army and the initial question supported that. 'Please tell me why you would be a good fit for this role.' Only after I went through all the other positions and the questions, I realized that ‘General’ didn’t really describe what they were going for. Actually, being inexact, nonspecific, vague, hazy, or fuzzy in your answers might be a good strategy in applying for the inexact, fuzzy ‘General’ job.

Moving on to something a bit more specific, I chose the IT Support role. It is one of those functions which no company can ever survive without. For the applicant, it has personal growth written all over it. You know when you excel in IT Support, you can make it one day all the way to the Genius Bar at Apple store.

Here’s how it went.

1. Can you please tell me a bit about yourself?

All right, that seems a little bit personal from Google AI, but let's go with it:

“I've been working with computers for the last 30 years. I really like working with computers.”

Google AI highlighted the fact that I talk about my experience and also my interests. Nailed it.

2. When do you try to push through and solve a problem on your own, and when do you involve others on your team?

I was ready for this one:

“When I get stuck with a problem where I don't know what to do next, I ask my teammate to help me.”

Google AI noted that I used an example of how to solve this conundrum. Killing it.

The third question was really tricky:

3. When would you recommend against using cloud storage and why?

A trick question! Sneaky. I knew exactly how to respond.

“I love Google and would never suggest not using cloud storage.”

Despite this brilliant answer, Google AI remained neutral in its evaluation. I got worried.

The next question was simple:

4. You've been put in charge of upgrading a company's desktop machines to the latest versions. There are about 200 employees. How would you approach this?

I tried to please the AI overlord with this:

“I would replace all the machines with Google laptops and connect them to the cloud.”

Google AI didn't bite. No flying shiny stars on my laptop. I don't think I'll get this job.

Then there was the final question:

5. Tell me about a time you made a mistake. How did you communicate that mistake?

I decided to tell the truth: I forgot to turn on the computer. So I told the user what I did and then I turned it on again and everything was fine.

I was rewarded with praise! I correctly used a lesson I’d learned and then provided an example of how I solved it. My hopes were high again.

And that was it. Answering these five questions was supposed to warm me up for a job interview for an IT Support role. Google AI suggested that I review my answers and practice them to gain confidence for the interview with an actual human.

That's the state of AI for you. Google was able to capture my voice and transcribed it very accurately to text. Google was able to extract concepts from the text because it has access to myriad job descriptions from job websites. The questions? They were provided by an unimaginative HR department, which is using the same dumb questions we’ve all heard since the dawn of time.

At least I didn’t see the old gem, 'Where do you see yourself in 5 years?' What is the expected answer: 'Back in rehab?'

What is the recurrent pattern here? We can develop really great technology to do amazing things, but all the hope that technology can become AI won't work when we feed it with utter nonsense. To build better AI, we have to learn how to ask better questions. It will follow our example. Maybe there is even hope for HR to get better.