Zuckerberg: AI is both closer than we think, and really far away

20 Dec 20168 Shares

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

Image: John Williams RUS/Shutterstock

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

Mark Zuckerberg has revealed his latest side project, the AI tool Jarvis, which will gradually change how the family home is run.

Mark Zuckerberg is a man of many talents. He is CEO and founder of Facebook, he’s pouring resources into cutting-edge medical research, and he’s pretty focused on an AI-led future.

The latter of those three things is as much a personal project as it is professional.

Zuckerberg

Zuckerberg sets himself annual challenges to complete throughout the year and in 2016, they were to run 365 miles, cumulatively, and build an AI home helper that everyone could get along with.

While the first challenge was straightforward, the second was complex, a bit unknown and, surprisingly, took less time than all that running.

“I’ve built a simple AI that I can talk to on my phone and computer; that can control my home, including lights, temperature, appliances, music and security; that learns my tastes and patterns; that can learn new words and concepts; and that can even entertain Max,” wrote Zuckerberg, the latter name a reference to his young child.

Using several AI techniques, “including natural language processing, speech recognition, face recognition, and reinforcement learning”, Zuckerberg wrote the programmes in Python, PHP and Objective-C.

Zuckerberg said he undertook the challenge to see what areas of AI we were furthest away from, and what we’re closer to than many people thought.

He said the coding written for Jarvis are all “variants of the same fundamental pattern recognition techniques”.

Zuckerberg thinks it will take less than a decade to get to a stage where AI systems are more accurate than people for each of our senses. “It’s impressive how powerful the state of the art for these tools is becoming,” he said, though words of caution followed.

“We are still far off from understanding how learning works,” he said. “We know how to show a computer many examples of something so it can recognise it accurately, but we still do not know how to take an idea from one domain and apply it to something completely different.”

Zuckerberg spent 100 hours on Jarvis. He doubts that ten times that amount of labour would lead to an AI learning skills completely on its own.

“In a way, AI is both closer and farther off than we imagine,” he said. While it’s great for things like connected cars, we’re still getting to grips with “real intelligence”.

Gordon Hunt is a journalist at Siliconrepublic.com

editorial@siliconrepublic.com