Why AI is already dead (and what’s coming next)
Using machine learning to predict the end of “machine learning”
According to CB Insights, VC investment dollars in AI startups surpassed $5B in 2016. That’s nearly double what it was just two years ago, and almost ten times more than in 2012.
Ask any technologist what they do right now, and you’re liable to hear stuff like, “We use AI and ML to help you make better sales forecasts,” and “We use AI and ML to tell you what to have for dinner,” and even “We use AI and ML to predict which Game of Thrones character is going to die this week.”
We’ve previously written about how you can tell what’s legit from what’s baloney when it comes to AI claims in HR tech, but this week we wanted to look at this from another angle. With all the AI and ML talk going around, how do job posts with these terms perform?
We know that the language that works best in job posts changes over time, sometimes quite rapidly. This is a huge bummer for all the people who keep reusing the same job post year after year and expect it to keep working. But it’s a major plus for you, since you’re using augmented writing software to write your job posts and always know what’s current!
Our hypothesis was that, as performance in job posts goes, AI and ML are going the way of Big Data — once cool, then cliche, now beginning to feel hopelessly dated. But is it true? Do jobs containing these terms perform better or worse than jobs without them? And if AI and ML are on the wane, what’s coming next?
Here’s what Olivia Gunton, a Data Scientist here at Textio, found when she used Textio to examine these questions.
AI and ML jobs outperform other engineering jobs, but the margin is narrowing
As someone with deep knowledge of this area, Olivia decided to begin by looking at the phrases artificial intelligence, AI, machine learning, ML, and machine intelligence, and she included the last two years of Textio outcomes data in the analysis.
These terms were hot a couple of years ago. In 2015, engineering job posts containing artificial intelligence or AI filled nine days faster than average engineering jobs — that’s 28% faster! Though machine intelligence showed up much less often than any of the other phrases, job posts containing it did even better, filling an incredible 18 days faster than average.
In 2017, these jobs are still hot, but the effect is cooling. Today, jobs with any of these phrases fill between one and two days faster than average. This a huge drop-off from the dramatic difference this language made to job performance just 18 months ago.
AI language is on the rise, so why isn’t it working anymore?
As tech investments in AI have increased, this language has shown up more and more in tech job posts. In 2015, usage was already reasonably high, with 13.3% of all new software engineering jobs containing at least one of these phrases. Today, usage of these phrases in engineering jobs is up to nearly 20% — and still increasing.
These phrases appear to be following the pattern of big data before them. As AI language has become more common in job ads, the less interesting it has become to potential job applicants. No surprise: when everyone is using the same language, no one stands out.
The takeaway: Want to fill your jobs quickly with great people? Don’t sound like everyone else.
What comes after AI?
Now let’s get to the fun part: making some predictions!
Textio looked at a bunch of tech terms that appear to be trending based on their recent usage growth. While all of these phrases still occur much less often than AI or ML, they’re all on the rise in a statistically unexpected way.
This, of course, is how trendy tech terms begin, first used by a few people in the know. As a term begins to gain credibility, more and more people adopt it. These phrases are on the rise now, but big data, AI, and machine learning have shown us how this story ends.
Enjoy your 15 minutes of fame, chatbots. Everyone knows what happens next.
Learn more about how language impacts your hiring at textio.com
