Data Backup Digest

Artificial intelligence (AI) is growing more sophisticated every single day. While past systems were only capable of giving binary commands in the form of a yes or a no, some modern AI systems are capable of mimicking human thought, behavior, and speech with a great degree of accuracy. In fact, one of the latest AI systems, known as PLATO, reportedly develops intuition at the same rate of a human baby.

What is Intuition?

Human intuition has long been an obstacle to achieving lifelike AI. Nuances like a human’s gut feelings and impulsive reactions aren’t easily replicable by AI systems – until now. Developed at an AI laboratory in the UK known as DeepMind, PLATO – which stands for Physics Learning Through Auto-encoding and Tracking Objects – is a next-gen AI system that’s currently making headlines around the globe.

But PLATO is hardly the first AI system to make news as of late. What sets this one apart from all the others?

Intuition Training

The PLATO system has been heavily trained in the basics of human intuition. To achieve this, the development team shared 28 hours worth of animated videos with PLATO – all centered on the subject of basic physics.

Most of these videos were centered on various objects and the common laws of physics that exist between them. In some cases, scenarios were presented that purposely defied these norms. After watching all 28 hours of the training material, PLATO was able to learn intuitive physics – including how it’s possible for an object to move from point A to B or how multiple objects cannot share the same space simultaneously.

According to the original research paper: “Current artificial intelligence systems pale in their understanding of intuitive physics, in comparison to even very young children. Here we address this gap between humans and machines by drawing on the field of developmental psychology.”

Just in case that wasn’t impressive enough, the AI system behind PLATO actually seemed surprised when it watched videos that purposely defied the laws of physics. The system essentially knew that things weren’t adding up and that something wasn’t quite right. Moreover, PLATO was able to effectively use the intuitive physics it had previously learned in order to predict what might happen next.

In order to achieve this, the development team behind PLATO tapped into a synthetic dataset that was furnished by MIT researchers. The dataset uses various visual aides to probe and disseminate physical knowledge – including specific objects that had never been seen or experienced by PLATO in the past.

The authors of the original research paper behind PLATO continued by stating: “The data suggest that intuitive physics knowledge emerges early in life but can be impacted by visual experience. Of course there is extensive debate and legitimate uncertainty about innateness.”

Regardless of what you think of the project, PLATO is certainly one of the most sophisticated AI systems we’ve seen to date. It will be exciting to see what innovations come out of this system in the coming weeks, months, and years.