Well then, we’ve now been told there’s a one in ten chance that AI could kill us all within the next decade. The maths of this is interesting to say the least, though the credibility of predictive stats is not just an issue with this particular statement. The reason for saying that the probability of a coin toss coming down heads is 50% is easy enough to explain. But when it comes to predicting events that have never happened in the history of humanity, how do you come up with a probability figure?
But you’d like to think that the figure came from somewhere, even if Anthropic’s safety researcher was no more forthcoming on that than he was on the method that AI was putatively going to use for its genocidal attack.
On this maths, the odds of extinction within seventy years slightly exceed the chances of survival, which is pretty alarming to say the least.
I’m not trying to downplay the danger here, but I think it’s fair to say that the figure is more an expression of how worried the commentator is: they are simply using statistics to try to convey the risk.
But what nobody seems to be highlighting is that the dangers are, yes, in part due to the rapid and uncontrolled development in AI, but without the internet this AI would be powerless. How credible would this prediction have been in 1996 when less than 1% of the world’s population used the internet?
If you examined the disaster recovery plans of top companies and other large organisations, how many of them have plans that mitigate for total internet failure, either because access becomes impossible or it simply becomes unsafe to use? My suspicion is that many security officers rule this out as either unrealistic or take the view that if we were to lose the whole internet there would be bigger problems to worry about than could be mitigated against through any individual disaster plan.
And, until now, there was probably a reason for being reasonably confident that extended loss of continuity of internet service probably dropped into the category of “unrealistic” disasters, alongside meteor strikes and invasion from outer space. But think of this. In the year 2000 we knew exactly what to look out for, and by and large avoided a potential problem. Our faith in the continuity of a reliable internet may just be a failure of imagination over where the risk is coming from.
This warning over AI should give us a clue. In our determination to improve efficiency, often driven by the “imperatives” of our current economic system, we have adopted and promoted a communications monoculture. Too many organisation rely on just the one way of working. Much as the loss of biodiversity makes us more exposed to disease, or the loss through disease of the food and materials on which we depend, so this communications monoculture has left us exposed.
If we had lost the internet in 1996 most of life would have carried on as usual. We know that’s not the case now. How did we come to this point?







