So I stumbled upon a paper which apparently is common in transhumanist and AI theory research circles titled: "A theory of universal artificial intelligence based on algorithmic complexity.".
A video of the concept:
And to fully grasp this concept this video on the foundation of intelligent agents is necessary:
Upon review of this work I noticed that it's not computable and therefore not practical in the real world. It is useful because it presents a formalization of generalized AI as a theoretical framework. It may be possible to take this paper as a basis for building something practical and some other papers have come out since which seem to be going in this direction. What does this actually mean?
Conclusion
So far it looks like a viable approach for AI but because it's so generalized I would expect the learning rate to be very very slow compared to the neural net approaches (uses lots of data) or the symbolic logic approach (requires huge knowledge base). The best performance mentioned in the video is tic tac toe which is an achievement but it's not a life changing breakthrough in AI. If somehow someday they can find a way to improve the performance and increase the learning rate then things could get very interesting.
References
Hutter, M. (2000). A theory of universal artificial intelligence based on algorithmic complexity. arXiv preprint cs/0004001.
Hutter, M. (2007). Universal Algorithmic Intelligence: A Mathematical Top→ Down Approach.
Veness, J., Ng, K. S., Hutter, M., Uther, W., & Silver, D. (2011). A monte-carlo aixi approximation. Journal of Artificial Intelligence Research, 40(1), 95-142.