The main problem of modern science is pointed out: reality of specific three-type systems (TTS) that are usually presented as complexity, and simultaneously impossibility of a description of such systems by a traditional modern reducing approach. Studying properties of system elements cannot help in a description of a complex system itself – complexity (three-type systems, living systems). Hence there arises an acute need in creation of new theories which would operate with maximum uncertainty and unpredictability and provide modeling TTS. The first step in this direction was taken on the grounds of a creation of theory of chaos and self-organization (TCS), according to which complexity cannot repeat an initial state of a system (or a vector parameters x(t0)), measures are not invariant, autocorrelation functions do not converge to zero and Lyapunov exponents are not positive. Chaos of TTS differs from a deterministic chaos and statistical distribution functions f(x) are not appropriate to describe it, because they continuously change. Deterministic, stochastic and chaotic models cannot describe TTS. This is the main property of emergent systems (complexity, TTS), therefore they are described by quasi-attractors.