Facebook’s AI Research team has been working on its Deep Learning systems for some time, and company CTO Mike Schroepfer has been talking about the development of a Deep Learning system based on neural networks.
The challenge that Facebook is attempting to solve is to create a system that can closely mimic the way in which humans store information.
Schroepfer explained: “A lot of the techniques we’re looking at now, that’s mostly for short-term memory. The human has a small short-term memory and much bigger long-term memory, so you want a closer memory architecture to that.”
Facebook has already demonstrated systems capable of complex reasoning by grafting a short-term memory onto neural networks with the aim of giving a system the ability to respond to complex queries.
Describing this as a fundamentally missing component of AI, Schroepfer said that AI systems need an in-built learning capability and that involves having the ability to retain new facts it has never come across before.
AI that understands
The new memory networks technology from Facebook was demonstrated when a system was shown that could answer questions about the plot of a movie after being given a film’s synopsis. Successful answers to questions about characters and items in The Lord of the Rings were given.
Schroepfer explained the importance of the successful experiment by pointing out that “you can see basic reasoning and basic understanding of relationships of the ring with the person, where they are and where they were.”
Facebook wants to use AI to help it organise posts from more than 1 billion people a day, and with the increasing image-heavy content across online platforms, this means having an automated solution that understands what is being shown visually.
The process involves segmenting an image into component parts so that a machine can better understand the whole, and whilst the science behind the new developments is nothing short of revolutionary, the day-to-day applications could be more mundane.
Better AI-filtered screen content could mean a user’s News Feed could regulate what type of images were seen on a basis of personal preference.
“All of this is about understanding the world as it exists today and better helping to filter and manage that,” said Schroepfer.