Meta’s AI Chief Yann LeCun Publishes Paper on Autonomous AI

Imagine a world where artificial intelligence can experience and learn in a truly human-like manner. That’s exactly what Yann LeCun, VP & Chief AI Scientist at Meta imagines in his latest paper which imagines a world of “autonomous” AIs. A world that many in science fiction have worried about for decades. A world of superintelligent programs that can match humans in self-awareness and possibility, exerts itself.

If the world that Yann LeCun imagines does come to pass, it will likely shake the entire foundation of human civilization. As it would mean, for the first time since Neanderthals, humans will share space with entities of comparable intellect. In his paper, LeCun breaks down the path for an autonomous AI architecture by training it to learn more efficiently. This is because, at this point, AI has yet to be able to predict or plan changes in a real-world environment outside of its training sandboxes.

In Yann LeCun’s vision, we could get past current AI limitations by this architecture flow which was put together by Freedom Preetham, founder of Kena.ai:

1. The configurator module performs executive control.
2. The perception module receives signals from sensors and estimates the current state of the world.
3. The world model module constitutes the most complex piece of architecture. Its role is twofold: (1) to estimate missing information about the state of the world not provided by perception, and (2) to predict plausible future states of the world.
4. The cost module computes a single scalar output that predicts the level of discomfort of the agent.
5. The actor module computes proposals for action sequences.
6. The short-term memory module keeps track of the current and predicted world state, as well as associated costs.

Altogether, the point of Yann LeCun’s paper is to allow AI to learn like humans and animals, by gaining copious amounts of information from the world by observation, with a bit of interaction, if possible. Though this doesn’t mean SkyNet quite yet, what has been published could lay the groundwork for future AI systems to learn with greater effective ability, which then will allow these systems to train themselves to better adapt to real-world conditions. 

In the future, the benefits of autonomous systems can’t be understated. The ability to have systems in place that are able to manage uncertain situations and events in a world full of uncertain situations and events could lead us to a level of artificial intelligence currently reserved in science fiction.