SCEIWA - What is an AI?

An informative guide, by the team at SCEIWA.

AI, or Artificial Intelligence, was first defined as any task performed by a machine that required a human level of intelligence. A lot has changed since the 1950s! Today’s AI are divided into ‘narrow’ and ‘general’ intelligence. Most AI fit the first category: developed for a singular and specific task, such as text generation or image identification. Artificial General Intelligence are broader and aspire to a “human-level intelligence”. Omi is the Software and Computer Engineering Institute of Western Australia's AGI project.

  Myths About AI 
AI will steal everyone’s jobs - While some jobs are automated by AI, others are created!
AI are impartial judges - AI trained on biased data will be biased so we developers must be constantly working at improving our AI, and aware of the possibilities for harm.
AI will become evil overlords - While we can’t see the future, we do know that there’s no AI nearly advanced enough to do that now! We also use responsible development policies in 
all our AI work. Read more about SCEIWA’s ethics here [hyperlink broken].

Most modern AI operate as a neural network, which is a series of linked nodes that take input data and output data based on a training model. Omi is a parallel linked neural network, constructed from a series of specialized neural networks running in tandem contributing to create a single AI. This is how Omi is better able to approximate AGI.

  So What Does This Look Like in Practice? 
User Input: “Give me a picture of a dog.”
Communication: Language Engine recognises the phrase, breaks down key parts of the statement, identifies it as a request.
Cognition: Knowledge Management confirms the request and pulls images of dogs from the database, which are passed back to Communication.
Communication: Vision Engine generates a unique picture of a dog from the database and Language Engine forms an appropriate response.
PLNN5 Output: “I made you a picture of a dog!”