Deep learning is underway. Already, computers can beat champions at go, an incredibly high level game. Neural networks are being evolved that can do high level things that humans can without the massive neurological architecture. Clearly, the method being used is already better than traditional evolution. We have biological presets at our disposal, are reverse engineering the human brain to the point where we can simulate it well on a supercomputer (like we did in 2014). All we need is bigger neural nets and we'll develop neural networks that are constructed with a connection cost and hyperNEAT, will be more efficient than the human brain and will have a billion times the processing power of the human brain by 2045. The software won't need to be designed by humans; it will be evolving and self improving at exponentially faster rates than humans could design it. So, with all of this in mind, when will the singularity happen?
"2100+"
You're high. That's like saying computers were invented in 3,000 AD. And then there's the knock on effect of the singularity on other predictions. For example, when the singularity happens, in around 2045 (as predicted by an expert, Ray Kurzweil, who's always right about his predictions. Like in chess, you listen to an expert with a fantastic track record, not 500 million disagreeing novices and intermediates. What's right isn't a popularity contest.) and then you can expect a software and design explosion from the intelligent machines, making every other engineering-determined prediction arrive sooner.
Please fix your predictions' timing, future timeline. Throw away popular belief in favour of expert calculations.