AGI is a large field. My focus is to make a meaningful contribution to AGI research.
While I do believe some form of AGI will be recreated in computers one day, I think it is improbable that we will see it in our lifetimes. I believe my goal is closer and attainable. I want to combine several areas of research to create a data structure of composable, generalized, abstract models of our world that are usable by software agents. These agents would be separate from the world in either physical or simulated form, so their knowledge would be grounded, which means they could inspect the world similar to how we do. If the data structure works as envisioned, then the agent would have an abstract definition for concepts like rough, home, I, etc. I did build an early proof of concept here. Even if the database is not successfully built, this idea will be brought to mainstream machine learning (its starting to happen). If this database can be built, then for all intents and purposes, it will look like AGI (philosophical zombies). A database that can move from concrete to abstract concepts like “twitch <-> move <-> walk <-> exercise <-> activity to do with friend” will able to move between “object <-> world <-> me” and any other abstract relationship humans can think about.
My interest of research are grounded language learning, generalized knowledge representations, composable systems, and sensorimotor systems.
There are related fields as well (but most of them can be included under my main research areas): unsupervised learning(sensorimotor systems), predictive processing(sensorimotor systems), language of thought (grounded language learning), and abstraction mechanisms(composability).
My direction is different from mainstream machine learning in that I do not want to improve current machine learning systems. I am interested in combining new research with engineering to create a new paradigm that is scalable (although this is similar to large orgs like OpenAI and DeepMind).
My direction is also different from mainstream machine learning research in that I want to pull as much research from neuroscience as possible. Human brains are the only source of general intelligence that we know of, so it does not make sense to me to focus solely on traditional machine learning research (which is usually about abstract programs.)
My focus is different rom mainstream machine learning in that I am absolutely not interested in supervised systems.
I have a devops, software development(I write bad code…), engineering management, and production engineering background. Hopefully I can take those skills and use them for development and research.
I believe in fast, small iterations as a means to bring vision to reality.