While Facebook’s artificial intelligence (AI) lab is focused on advancing machine learning, the social networking firm’s chief Mark Zuckerberg has set himself a personal AI goal for 2016.
“My personal challenge for 2016 is to build a simple AI to run my home and help me with my work. You can think of it kind of like Jarvis in Iron Man,” he wrote in a blog post.
Zuckerberg plans to start by exploring existing technology with the goal of developing a system that will respond to voice commands to control things in his home like music, lights and temperature.
The next step will be to enable his digital assistant to recognise people and open the door to friends, to keep an eye on his daughter Max and help him with his work.
“On the work side, it’ll help me visualise data in virtual reality to help me build better services and lead my organisations more effectively,” he wrote.
According to Zuckerberg, he sets himself a challenge each year to learn new things and grow outside his work at Facebook, such as learning Mandarin and meeting a new person every day.
Coding an AI assistant should be “a fun intellectual challenge”, he said, promising to share what he learns over the course of 2016.
Despite warnings from researchers about the potential dangers of AI, large tech firms are racing to apply the technology to consumer products.
While AI that could pose a threat to its creators is still considered to be a long way from becoming reality, machine learning is making rapid progress due to technological advances that have enabled cheap storage and increased computing power.
Under Zuckerberg’s leadership, Facebook is investigating techniques such as neural networks and natural language processing. The social networking firm’s AI lab, led by neural net expert Yann LeCun, is focused on advancing these types of machine learning.
As a result of the AI lab’s work on neural nets, Facebook rolled out its digital personal assistant called M in August 2015. Built on top of Facebook Messenger, M’s AI technology also relies on human assistants, who monitor how M is working and supplement its responses to questions.
The AI-plus-human system is helping Facebook build a model by analysing how and when humans intervene, according to Wired. Eventually, M will be built on neural nets, but even then humans will need to be in the loop to continue training the technology.
Facebook is trying to find a way of building machines that have a kind of artificial common sense, but that will require finding a way to enable machines to learn by themselves.
LeCun told Computer Weekly's sister-site SearchCIO that Facebook is looking to use a technique known as vector embedding, which can be used to introduce a degree of basic reasoning into AI systems,
He told SearchCIO that the algorithm works by making an "educated guess" based on a comparison of detailed mathematical descriptions of known and unknown responses when it comes across an image or text that it does not understand.
However, LeCun is among those who believe that developing a common sense program, or artificial general intelligence is still a long way off.
Artificial general intelligence should be approached with caution, according to Stuart Armstrong, research fellow at the Future of Humanity Institute at Oxford University.
In his book Smarter than us, Armstrong lays down a few points that humans need to consider about general-purpose AI.
“Never trust an entirely super-intelligent AI. If it doesn’t have your best interests at heart, it’ll find a way to obey all its promises while still destroying you,” he wrote.
Read more about AI
- Socially aware general-purpose artificial intelligence in the form of a dog could be the ideal form factor to take over the world.
- Two artificial intelligence luminaries talk about the future of AI – what’s possible and what’s still a ways off.
- Harvey Koeppel offers 10 steps CIOs must take now to prepare for the AI technology era ahead.