What the Self Sampling Assumption (SSA), an alternative principle to SIA, says depends on the reference class. If the reference class includes AIs, then we should strongly not anticipate such an AI explosion. If it does not, then we strongly should (by the doomsday argument). These are both basically due to the Doomsday Argument.
The Self Indication Assumption (SIA) is a compelling topic because it's a tool we can use to think about questions like the Fermi Paradox, not to mention our own future.
As a general rule (read: admittedly sloppy heuristic) - when an outcome of a chain of reasoning depends strongly on how we categorize the involved entities, that reasoning is suspect. This would seem to be a general problem for all SIA/SSA reasoning, although since Katja just finished her degree on exactly this topic, I would refer you to her if you want counterarguments to this suspicion.