The potential of Artificial Intelligence (AI) applications in the workplace is both exciting and scary. We’ve all started to see how useful AI proves to be when handling loads of data that we humans can’t effectively fathom. We’re also very well aware of the threats it may pose for privacy and equity.
At CodinGame, we’re cautiously but resolutely turning to AI and all it can enable for both recruiters and candidates.
What is the actual power of AI? How will AI transform the HR industry? How can we keep being human when algorithms start doing the job?
Unlocking the Positive Power of AI
AIs are Dumb
There is this myth around AI that it must be so complex and advanced to grasp that only a few scientists can deal with it.
Not really. A few if conditions and for loops, and you already have a bot — a very basic kind of AI. I’m always reminding players in our community of this.
I know that people tend to fear what they don’t know or understand — like neural networks, genetic algorithms, Monte Carlo Tree Search algorithms or any other algorithm with a fancy name — but they shouldn’t be afraid of AI. AI programs are just tools that developers create to do one job really well. Yes, they are very efficient at the tasks they were programmed to tackle. But these tasks only; they are not capable of performing other tasks.
When you think about robots and what they can achieve, one funny thing is that they easily surpass us on some difficult tasks like chess, but they can also have a hard time on other relatively simple tasks. Like piling up cardboxes, for example. This is called the Moravec’s paradox:
“It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”
Opportunities for a Better World
We usually classify AI into two types:
- Weak AI (or narrow AI), which is the dumb type I’ve described above
- Strong AI, which can think and perform tasks like a human
Today, strong AI doesn’t yet exist, and, to be honest, is far from existing. So let’s forget about strong AIs for a moment when contemplating the potential benefits of AI in the world of tomorrow. Because the number of untapped opportunities is huge.
Do you consider AI applications as dehumanizing? To me, AI offers opportunities to gain time and concentrate on things that matter. Yes, I’d like my house to optimize my energy consumption. If my fridge could remind me I haven’t eaten food that is approaching the expiry date, that would be great. And I would definitely prefer to interact with a chatbot that can solve a problem with my bank account in 5 minutes rather than have to wait for the opening hours to call my adviser.
We’ve got the tools and they’re improving. Let’s use them.
A Revolution in the Workplace?
Is AI stealing jobs?
I think AI is and will be replacing more and more jobs. I find it inevitable, and it actually makes good sense. If a more efficient way to perform a task exists, why would you keep the old way? We’re talking about repetitive and boring tasks here, like fruit-picking.
On the other hand, I believe that a number of new jobs will appear — meaningful work that weak AI cannot replace. We’ll need more people able to:
- Create AI software and maintain it
- Organize and optimize the workplace around these new artificial agents
- Communicate and span the bridge between humans and machines
AI will save businesses a lot of time. Call me naive if you want, but I think this time won’t necessarily be transformed into less workforce and more money. The companies that will effectively harness the power of AI will turn their focus back on customers. And this will make a positive difference.
Recruitment 2.0
At CodinGame, we’re in the HR tech services business. Here are three major changes that AI could bring to this business.
People to Positions Matching
On one hand, we have countless open job positions from companies all over the world with multiple criteria: years of experience, technical skills requirements, maximum salary…On the other hand, we have so many qualified candidates, either actively searching or simply open to new opportunities, with their own criteria: minimum salary, specific benefits, location, preferred technologies…
How can both parties be matched effectively? That’s a major concern today for the recruitment industry.
For companies, the matching should be:
- Reactive. If they need developers for an upcoming project, their time frame is usually weeks, not months.
- And reliable. Any time spent interviewing, or even worse, hiring a candidate who eventually doesn’t fit and leaves the company during the trial period, is lost. It’s ok not to find the perfect candidate on the first shot, but every lost candidate costs.
For developers, the matches should be:
- Measured. Candidates can’t focus on too many companies at the same time. Moreover, they’re already being solicited a lot today.
- And relevant. Some companies will eagerly release a criterion, since their job offers often target the perfect candidate. Developers, on the other hand, may have very specific demands on which they will not budge, and they expect the matching to respect that.
It represents a huge problem for humans, but AI can help!
A program won’t have any issues managing and sorting large lists of candidates, companies and criteria. An AI would make it possible to: easily re-include archived applications from already qualified candidates, avoid risky matches based on previous performances, or even ensure freshness of data in both candidate profiles and job offers with regular updates.
The most important improvement of AI-improved matching, though, would be the reduction of bias. AI will not be influenced by information such as physical aspects, names, gender and origin on a candidate’s application: the likelihood of discrimination will be greatly reduced.
Candidate sourcing can become data-driven yet still focus on improving the experience of candidates and recruiters.
The Perfect Assistant
With so many positions and candidates, and so many details to validate on both sides, it becomes more and more complex for intermediates to personalize the experience for both parties. Here again, AI can make a difference.
What if a robot could assist candidates all along the way, from the first encounter with a company to the signing of a contract? Natural Language Processing (NLP) can work wonders and will continue to improve. In a few years, it might be difficult to tell whether we’re talking to a bot or a real live person online.
Candidates could have conversations with their recruiting assistant about their demands, their expectations, their interrogations…
Such robots will have key merits:
- They don’t forget about the criteria and expectations of both parties.
- They can plan the whole process (interviews, tests, calls) and provide advice all along the way.
- They’re always available if candidates or recruiters need some information or help.
Keeping up with several hiring processes can be overwhelming for both candidates and recruiters. By making it possible for candidates to only have to talk to one intermediary, an artificial assistant could simplify communications a lot.
Intelligent Assessment
Assessment tests have become an integral part of the hiring process. MCQ tests, code exercises, oral examinations…Even a motivation letter can be seen as a kind of assessment test today.
Most of the time, before being reviewed by the recruiting party, these tests are first corrected by a machine. What if a machine could understand the candidates’ thought process and determine if they are a good fit for the company – even better than a human reviewing the test answers would?
An AI could analyze how candidates complete the tests: how many times they fail a code exercise before actually finding the proper solution, what parts they got right first…More importantly, an AI could compare the results and the patterns of the completed test with other candidates’ answers.
We could also imagine that an AI would actually modify the test questions according to the candidates’ previous answers to better qualify their skills.
In a few years, we will be very far from a test with right-or-wrong types of questions.
Asking a candidate for 30 minutes of their time to take an assessment test will probably give a lot more information about them, their qualities and their skills than 3-hour test assessments today.
The Stakes of a Successful Transition
In an ideal world, the workplace would easily integrate AI everywhere possible and allow humans to benefit from its power. But I’m not that naive. There are a few key points we should be wary of.
It’s Not a Kind of Magic
One of the major hindrances to AI today is, in my opinion, its failure to be welcomed in our society. Not because of cinema and its AI-caused apocalyptic scenarios, but because of how we build and use AIs.
Imagine you’re asking for online chat support. If you’re misled into thinking that you’re talking to a person, whereas it’s actually a chatbot, you will most probably have a bad feeling about the support provided. On the other hand, if you’re told you’re talking to a chatbot that can understand this, that and the other, the feeling you’ll have will be completely different, even if the conversation happens to end the same way.
It’s a matter of perception.
I think that every time an end-user comes in direct contact with an AI, the AI should state what it is, what it does and how it does it. Let’s stop using black-box magic effects like “Our intelligent system will find the best recommendations for you.” Everyone should be able to understand what AI is and isn’t.
Once people understand that what they’re interacting with are just powerful programs that can manage a lot of data and perform some tasks very efficiently, we will have made a major step in having more and more robots in our life.
Fighting for Good
We build machines in our image. It’s perfectly normal, but we’re quite imperfect, so we should be all the more mindful of the biases we can unintentionally introduce in AIs.
An AI won’t realize it’s discriminating against people based on their race or gender if it has been taught to do so. When an AI is unable to recognize a black face, we’ve got a big problem. Its creators probably didn’t mean any harm (they most likely only tested it with white faces); they didn’t think enough in their lab setting about how the AI would work in the real world. This example highlights the importance of diversity in the teams creating new AI tools and applications.
Moreover, AI is learning from us (as in Machine Learning). AIs use the data of today’s world to compute, analyze and deduce things for tomorrow. In this light, teaching AI ethics, common sense and good manners will be a real challenge. If not solved technically, safeguards and principles should be put in place, a bit like the Hyppocratic Oath, to make sure the AIs we create will always work for the benefit of all humans.
Empowerment versus Enslavement
Sometimes, a world with AIs seems too good to be true. Leave the complex, tiring and boring tasks to robots while we humans enjoy our lives – right?
However, AI should be an enabler for us to use our time better, build great things and never stop learning. We don’t want to live in a world where humans are too dependant on robots that do everything. If we did, what would happen to us when AI systems crash or get hacked?
The disturbing future in Wall-E
It reminds me of my parents telling me to continue exercising mental calculus even though we had calculators. Not learning how to perform a task simply because it can be done more effectively by robots would be a terrible mistake.
Would you trust artificially assisted surgeons if they never actually performed surgery on their own? Would you blindly hire a candidate that a machine chose for your company?
Let’s keep using AI as a tool, however advanced it may be, and remain the only really intelligent party in this new artificially-empowered world.
The Artificial Intelligence revolution is just around the corner and will profoundly reshape the world of work. There are many benefits ahead, but also many bumps along the way. We must actively adapt and prepare the ground for the big changes. AI is coming.
