Planning is a data efficient decision-making strategy where an agent selects candidate actions by exploring possible future states. To simulate future states when there is a high-dimensional action space, the knowledge of one's decision making strategy must be used to limit the number of actions to be explored. We refer to the model used to simulate one's decisions as the agent's self-model. While self-models are implicitly used widely in conjunction with world models to plan actions, it remains unclear how self-models should be designed. Inspired by current reinforcement learning approaches and neuroscience, we explore the benefits and limitations of using a distilled policy network as the self-model. In such dual-policy agents, a model-free policy and a distilled policy are used for model-free actions and planned actions, respectively. Our results on a ecologically relevant, parametric environment indicate that distilled policy network for self-model stabilizes training, has faster inference than using model-free policy, promotes better exploration, and could learn a comprehensive understanding of its own behaviors, at the cost of distilling a new network apart from the model-free policy.

"The essence of general intelligence is the capacity to imagine oneself" -- myself Recognize that to gain the perspective that comes from seeing things through another's eyes, you must suspend judgement for a time -- only by empathizing can you properly evaluate another point of view. Moravec's paradox is the observation made by many AI researchers that high-level reasoning requires less computation than low-level unconscious cognition. This is an empirical observation that goes against the notion that greater computational capability leads to more intelligent systems. However, we have today computer systems that have super-human symbolic reasoning capabilities. Nobody is going to argue that a man with an abacus, a chess grandmaster or a champion Jeopardy player has any chance at besting a computer.

Researchers behind a new study, published in Science Robotics, have developed a robotic arm with knowledge of its physical form – a basic sense of self. This is nevertheless an important step. There is no perfect scientific explanation of what exactly constitutes the human sense of self. Emerging studies from neuroscience shows that cortical networks in the motor and parietal areas of the brain are activated in many contexts where we are not physically moving. For example, hearing words such as "pick or kick" activate the motor areas of the brain.

Having a sense of self lies at the heart of what it means to be human. Without it, we couldn't navigate, interact, empathise or ultimately survive in an ever-changing, complex world of others. We need a sense of self when we are taking action, but also when we are anticipating the consequences of potential actions, by ourselves or others. Given that we want to incorporate robots into our social world, it's no wonder that creating a sense of self in artificial intelligence (AI) is one of the ultimate goals for researchers in the field. If these machines are to be our carers or companions, they must inevitably have an ability to put themselves in our shoes.

A robot able to'imagine' itself has been created in a step towards the self-aware robots envisioned in the Terminator movies. Skynet and other sci-fi machines are able to learn and decipher from scratch but real-world robots have yet to master this art. Now, scientists have managed to create a machine that can learn without prior programming via'deep learning'. After an initial 24 hours of behaving like a'babbling infant' it was able to grasp objects from specific locations and drop them with 100 per cent accuracy thanks to 35 hours of training. Even when relying entirely on its internal self model - the machine's'imagination' - the robot was able to complete the pick-and-place task with a 44 per cent success rate.