In the latest leap towards the greater goal of accomplishing human-level cognition in machines, DeepMind and Google Brain have developed an AI agent that can predict immediate future based on its past learning experiences. The agent is named Dreamer learns long-horizon behaviours from images purely by latent imagination. 

The research is outlined in a paper titled "Dream to Control: Learning Behaviors by Latent Imagination", authored by Danijar Hafner, Timothy Lillicrap, Jimmy Ba, and Mohammad Norouzi.

The researchers accomplished this by predicting the state values and actions in the learned latent space and propose an actor-critic method that optimizes a parametric policy by propagating analytic gradients of multi-step values back through learned latent dynamics. 

According to the paper "Dreamer outperforms previous methods in data-efficiency, computation time, and final performance on a variety of challenging continuous control tasks with image inputs." 

The researcher further demonstrated that Dreamer is applicable to tasks with discrete actions and early episode termination and future research on representation learning can likely scale latent imagination to environments of higher visual complexity.