A technological singularity is a hypothetical event occurring when technological progress becomes so extremely rapid, due to positive feedback, that it makes the future after the Singularity qualitatively different and harder to predict. It has been suggested that a singularity will occur during the 21st century, and there are several mechanisms by which a singularity could occur.
That at a certain point intelligence will become self-recursive and improvements to intelligence will lead to improvements to intelligence and eventually we'll have growing infinite intelligence, and the world will be a completely different place.
This is stupid.
It's based on a flawed assumption that you can tweak up intelligence and get greater amounts of intelligence. That just isn't the case. You can't just gather more and more intelligence and reach a point where you can use intelligence to make intelligence and then make more intelligence. Because, intelligence isn't a thing, it's a process and it takes real work to do. It's equivalent to evolution (I also mean this literally) and you don't get something for nothing. You can't have something that evolves to the point where it goes into superevolution. Evolution takes real work. This is the insight of Paley and the legacy of Darwin. You can't get it for free. You don't get solutions out of the blue like gifts from on high granted by your possession of intelligence as a skyhook. Intelligence is only and can only be a crane. You can only build from the ground up, and improve upon what you've already built. And while we're building better and faster, you don't at some point turn into a rocket ship and fly to the end of the universe abandoning your foundations.
There is an older and clearly example of this idea at work with regard to the idea of perpetual motion. That at some point with growing technological improvements the work out keeps improving and the energy in keeps decreasing and at some point these lines are going to cross and the energy in is going to exceed the energy out and we'll have a brave new world. Because once that happens, even if the amount is very slight we can use that energy to make more energy and more energy and more energy and soon we'll have limitless supplies of ever increasing energy.
If you apply computer power to intelligent pursuits the increases in computer power is going to make it much faster and easier to test more and more ideas and predictions and see if they are accurate or useful, but this is just going to make intelligent faster not more intelligent. You can have an intelligent computer that knows everything I know, and give it the same problem I'm given. And it might come to a consensus answer within a few milliseconds whereas it might take me a minute. And in some distant future, it might come to the answer I would come to in microseconds and it would still take me the minute. But that isn't making it any more intelligent.
I could live in super-slow motion for and exist for a thousand years, I wouldn't be any smarter. I could get to the answers faster relative to you, but my answers wouldn't be any better. I could do more research and spend more time on the problems, but I wouldn't transcend human knowledge or intellect and I wouldn't suddenly be excluded from science to determine truth from falsehood. I could understand the entire corpus of human understanding, and arrive at answers in mere seconds, or fractions thereof. But I wouldn't become magic. And even if I could modify what I am and make time go even slower for me and my life even longer, I wouldn't suddenly become magical or stop needing to do actual work to come to actual explanations. You cannot do experiments in your own head and come to true and impressive conclusions, you need to actually do the experiment and look at the real world to figure out how things work out. And while you could do this process hugely faster relative to others, it wouldn't matter how fast your processor was. You could be know everything I know and be as intelligent as I am, but run on an old 100 megahertz processor and come to my answer that takes a minute in five minutes, but five minutes or five nanoseconds, it's the same damned answer.
How much better is a chess computer that makes the best move it can in 1 minute than one that does it in 1 second? Would it be proper to argue that a faster chess computer is a "smarter" chess computer?