Highly simplified bayesian analysis. The universe of causal relationships is a subset of the universe of predictive relationships. The universe of predictive relationship (correlations) tends towards infinity. The universe of causal relationships is finite. Thus the likelihood that a predictive relationship is also a causal one is rather small and most correlations therefore are *probably* non-causal. Think of it as a Venn diagram where casual relationships are a tiny circle inside a much much larger circle of predictive relationships. Randomly throwing darts at the circle, you're much more likely to land outside the inner circle than inside it.
I would think that a "predictive" relationship requires a predictor. If roosters crowed only on uninhabited planets, no one would claim that they cause the sun to rise. Similarly, if roosters crowed only at noon, no one would claim that they cause the sun to rise. Not every imaginable relationship is necessarily predictive.
Think of it as a Venn diagram where casual relationships are a tiny circle inside a much much larger circle of predictive relationships. Randomly throwing darts at the circle, you're much more likely to land outside the inner circle than inside it.
You can't "randomly" throw darts at a circle. If you are throwing "at" the circle, you are aiming for it and there is nothing "random" about the deliberate act of aiming.