Using New York City as a model, the Facebook's AI research was able to come up with new algorithms to help travellers or any user prevent going around in circles.
The Facebook Artificial Intelligence Research (FAIR) group in New York made two AI programs to help users navigate New York City: a 'tourist' lost in the city, and a 'guide' engineered to help the lost 'tourist.'
The lost 'tourist' sees photos of the physical location, while the 'guide' sees a two-dimensional map with landmarks. Both algorithms are designed to get to a specific location.
With a 'tourist' feeding natural language instructions to a 'guide,' the latter will be able to learn and relate to real-life objects such as a park or cafe. Like a parent who teaches a toddler a thing through real objects and actions, a 'tourist' will be able to assimilate what a 'guide' is getting at.
In the field of artificial intelligence, language is a steep learning curve. This makes for a big setback for algorithms to understand more complex commands, just like a human's common sense would. With simple language, it means less sophisticated AI programs capable only of a bare minimum of work.
"One strategy for eventually building AI with human-level language understanding is to train those systems in a more natural way, by tying language to specific environments," the Facebook researchers wrote in a blog post. "Just as babies first learn to name what they can see and touch, this approach—sometimes referred to as 'embodied AI'—favors learning in the context of a system's surroundings, rather than training through large data sets of text."
With FAIR's novel innovation, this would help other research practitioners in the field come up with their programs based on this algorithm and help make digital assistants and chatbots more intuitive.
There's no mention of Facebook's plans for the algorithm if a massive rollout will commence later down the line. But if the social network giant is ready for it, it would be a huge help for travellers go beyond what Google Maps can offer.