advertisement
advertisement

To keep up with demand, AI may have to build itself

To keep up with demand, AI may have to build itself
[Illustration: John Lund/Getty Images]

One day, artificial intelligence may no longer need humans to build it.

According to the New York Timesthere are only about 10,000 people in the world who have the necessary skills to build the complicated, mathematical algorithms necessary to create next-gen artificial intelligence. That’s not enough supply for the potential demand, and while today’s students may try to add the skills to their resumes, until then someone—or rather something—will need to fill the void.

While companies like Google, Facebook, and Microsoft can throw money at the problem and pay “millions of dollars a year to A.I. experts,” per the Times, other companies are taking matters into their own hands and developing tools and neural networks to help companies build their own AI software. That’s where technologies like Google’s new AutoML come in. It hopes to automate the AI-building process by outsourcing it to AI itself. Per the Times: “Google said AutoML could now build algorithms that, in some cases, identified objects in photos more accurately than services built solely by human experts.”

It’s not just algorithms either, but AI is getting really good at teaching itself skills like speech recognition, machine translation, and all sorts of other things thanks to a process that experts call “learning to learn,” or “meta-learning.”

On a related note, The Terminator with all its self-aware machines is streaming on HBO Go.ML