Artificial General Intelligence (AGI) is a general purpose system that is smart enough to adapt to new situation by applying previous experience with an performance similar to or exceeding that of a human 7 year old. There are a lot enterprises trying to get closer to AGI. Computers today has an excellent pattern-matching ability of a type that might be called holistic thinking. These deep learning systems can answer all types of questions in a way that seems intelligent but is basically just a system that finds out what feels right in the situation. It's mimicking with no deeper understanding.
No system has yet combined this holistic thinking with reasoning, critical thinking and self awareness. AGI needs to understand what it is uncertain about, direct attention to what is important at each moment and make plans to resolve uncertainty by experimentation using available tools. There should be an integration between the feelings from the deep learning networks with the rational symbolic logic representation.
Emotions are basically global analog values that changes gradually over time. They influence associations, attention and goal selection. What you should focus and act on changes if you really needs to sleep or eat or in presence of danger. It depends on if you are in a learning phase or production phase. It depends on the social situation. This is similar to what exists within Recurrent neural network as a sort of working memory. Some people think that computers are good at math and bad at emotions. The state of the art shows us that the situation are the reverse. Todays AI are all feeling and no thinking.