There are many things we know to worry about. Some are very likely events, but by themselves not existential threats to civilization. Others could easily destroy civilization and even life on earth—but the chances of such disasters occurring in the near historical future seem to be vanishingly small.
There is a known possibility that stands out for being both likely in the next few decades and capable of destroying our civilization. It's prosaic and banal, something dismissed by many as a danger that the twentieth-century confronted and definitively rejected: That is war between great nations, especially when fought under a doctrine of Mutually Assured Destruction (MAD).
Arguments against the plausibility of MAD warfare are especially believable these days: MAD war benefits no one. Twentieth century USA and USSR, even in the depths of the MAD years, were sincerely desperate to avoid tipping over into MAD warfare. That sincerity is a big reason why humanity got through the century without general nuclear war.
Unfortunately, the twentieth century is our only test case, and the MAD warfare threat has characteristics that made surviving the twentieth century more a matter of luck than wisdom.
MAD involves very long time scales and very short ones. At the long end, the threat is driven by social and geopolitical issues in much the same way as with unintended wars of the past. At the other extreme, MAD involves complex automation controlling large systems, operating faster than any real-time human response, much less careful judgment.
Breakers (vandals, griefers) have more leverage than Makers (builders, creators), even though the Makers far outnumber the Breakers. This is the source of some of our greatest fears about technology—that if weapons of mass destruction are cheap enough, then the relatively small percentage of Breakers will be sufficient to destroy civilization. If that possibility is scary, then the MAD threat should be terrifying. For with MAD planning, it is hundreds of thousands of creative and ingenious people in the most powerful societies—many of the best of the Makers, powered by the riches of the planet—who work to create a mutually unsurvivable outcome! In the most extreme case, the resulting weapon systems must function on the shortest of time scales, thus moving the threat into the realm of thermodynamic inevitability.
For the time (decades?) in which we and our interests are undefendable and still confined to a volume smaller than the scope of our weapons, the threat of MAD warfare will be the winner in rankings of likelihood*destructiveness.
There's a lot we can do to mitigate the threat of MADness:
A resurrection of full-blown MAD planning will probably be visible to the general public. We should resist arguments that MAD doctrine is a safe strategy with regard to weapons of mass destruction.
We should study the dynamics of the beginning of unintended wars of the past, in particular World War I. There are plenty of similarities between our time and the first few years of the last century. We have much optimism, the feeling that our era is different ... And what about entangling alliances? Are there small players with the ability to bring heavyweights into the action? How does the possibility of n-way MADness affect these risks?
With all the things we have to worry about, there is also an overwhelmingly positive counterweight: billions of good, smart people and the databases and networks that now empower them. This is an intellectual force that trumps all institutions of the past. Humanity plus its automation is quite capable of anticipating and countering myriad possible calamities. If we can avoid blowing ourselves up, we will have time to create things so marvelous that their upside is (worrisomely!) beyond imagination.