. BUT I would argue it took about 15 or 16 years to even get to that point of being able to recognize red, recognize letters and numbers, recognize the shape of a stop sign, recognize and react to a ball rolling in front of you etc. Essentially millions and millions of repetitive tasks every hour of every day until you understood it enough to even get to the point of being able to sit behind the wheel of a car.
Now translate that to AI. If you wanted to train a computer to source or "recruiter" first you have to get it recognize "stuff" like job titles, contextualize meaning of words in grammatically correct and grammatically incorrect sentences, first person, third person, spelling of words and common misspellings like manger versus manager etc, then you have to train the AI to match that (usually poorly written) resume to a (poorly written) job description and then teach the AI to stack rank without bias. That should take about 15 years. Right now we are about 3 maybe 5 or 6 years in depending on who you ask and what line of BS you are willing to believe. Now train it by running it through millions and millions of tasks and test it and make adjustments etc
The difference is that with humans if I wanna learn how to drive it will take me 15-16 years, if my wife, kid etc wants to drive it will also take them 15-16 years. There are no short cuts. No matter who you are, given a certain mental capacity, anyone can learn to drive but it takes 15-16 years give or take.
With AI even if it takes 20-30 years to perfect a task, once it's perfected it can then be cloned and updated and modified across all systems and from then on it takes an instant to "train" a new system on how to do that task; be it autonomous driving or sourcing resumes.
From the demos I've seen on AI resume search today in 2019 it's good, not perfect. It's usually as good if not a bit better than a human recruiter but still needs a lot of work. If it is good now in 5 years it will be amazing and in 10 years it will seem magical.