Here’s the latest prediction concerning the future of AI, based on a survey of the leading researchers, conducted by Oxford University: the experts predict there’s a 50 percent chance that AI will be better than humans at more or less everything in about 45 years. Apparently they’ve learned from Nostradamus, who made a lot of predictions that have supposedly all come true. The secret is to be vague on the details and timing.
Making a specific prediction is usually a bad idea. That’s why the Mayans lost so much credibility when the world didn’t end in 2012. That’s especially true in technology. Video phones were first introduced in 1964 and were supposed to become common in a few years. Flying cars were supposed to have become commonplace long before now. The same goes for AI. The first practical demonstration of it occurred in 1956. Then it was predicted that we were 40-50 years away from AI becoming better than humans. And as the survey mentioned above proves, that prediction turned out to be true; we are 40-50 years away from AI becoming better than humans. In 40-50 years it may still be true.
The HR and recruiting technology world is abuzz with predictions about how AI will revolutionize the space. Predictions run the gamut from recruiters being completely replaced to not a huge change occurring. So what are likely to see? First, we should be clear that there’s no such thing as AI. True AI means a software system having the flexible, general-purpose intelligence of the type which allows an individual to learn to complete a vast range of tasks. That does not exist anywhere. No software available today is capable of mimicking even the intelligence of a three-year old. What is labelled AI is machine learning — the ability of certain software products to change and improve at what they do when exposed to new data, without being explicitly programmed. Of course the term “AI” sounds so much better than “machine learning.” The 2001 movie titled “AI” likely wouldn’t have been much of a success if it had been titled “machine learning.” Maybe a movie about Wall-e going to college?
Predictions are that AI systems will outperform humans in the next 10 years in tasks such as translating languages, writing high school essays, and driving trucks. These aren’t exactly profound predictions since the technologies have already been demonstrated. What these tasks have in common is that the technology focuses on a narrowly defined task. Make no mistake: it’s still a very complex undertaking to get a truck from one place to another, but it’s well understood what’s needed. So it will be for machine learning and recruiting.
Multiplicity
Recruiting will be affected by AI through “multiplicity” — diverse groups of people and machines working together to solve problems. An article in the Wall Street Journal mentions that solving complex problems requires training machine learning algorithms using data from a varied group of humans to demonstrate appropriate responses to varying situations. This is how the algorithms for a driverless car are trained. In recruiting, the two areas that would benefit the most from machine learning are predicting a candidate’s motivation and fit.
Predicting Motivation
Evaluating motivation is about improving sourcing, which is typically a low-yield, labor-intensive business. Every recruiter knows that reaching out to candidates who have not applied often produces few results because of low response rates. However, a machine learning system can identify people who are more likely to to consider a solicitation for a job; in other words, those who are more motivated to change jobs or accept a new one. There’s an abundance of data on social networks and other places that can be tapped for this purpose. For example, Google’s Timeline tracks your every move (check it out) and can be used to accurately determine a person’s commute. A candidate with a long commute is more likely to respond to a solicitation than someone who has a short one, especially if the former travels through heavy traffic.
Combine a candidate’s travel information with other data — such as remarks posted on social networks that can be indicative of their engagement levels in their current job — and you’ll very likely boost your response rate. Currently Google doesn’t let anyone see other people’s timeline, but that’s by no means guaranteed to continue. The cell-phone providers all have the same data as well and already sell it for targeted marketing campaigns.
Evaluating Fit
Knowing how well a candidate may fit in is another aspect of recruiting that will benefit from AI. The challenge today is one of defining the culture of the organization, which may not be the same everywhere. It’s easier to know if a candidate will fit in with a group using data from social networks. Using the profiles of people in a team, it’s possible to predict if a person will make friends with them. That’s a good proxy for evaluating fit and eliminates any need to define the culture.
Machine learning can also predict what the impact of adding a person is to a team’s productivity. It may substantially increase the team’s overall productivity or have no effect at all. It may even be negative. This can also be turned around to figure out what kind of a candidate should be hired so that a team can achieve certain goals.
The Revolution
The impact of revolutions can take a long time to be realized, and they often are not what anyone expected. There are more recruiters today than before the creation of job boards, applicant tracking systems, and social networks — all of which were predicted to eliminate the need for recruiters. AI will change the work recruiters do but not eliminate the need. Given the combination of a recruitment marketing system, an ATS, and assessments, it’s already possible to automate every step of the recruiting process. That recruiters haven’t suffered the fate of the dinosaurs isn’t because the work can’t be automated, but because it fundamentally involves human interaction. AI can help, but only up to a point. The algorithms need continuous input for improvement. The WSJ article also mentions that if people stopped providing input, these systems would quickly become outdated and would deteriorate. Of course, that may not be true in 40-50 years.