Kurzweil argues that there is exponential progress in software as well as hardware. I’ve always wondered about whether this was true because it is so hard to quantify. I’ve also wondered how hard the AI software problem will be since we are getting close to having the hardware we need. Musk recently stated that he thinks software is accelerating because timelines on AI projects keep compressing, I.e. he said 5 years ago he thought self driving cars were 10 years away then 2 years later he thought they were 3 etc.
Machine learning itself could be said to greatly accelerate progress in software development but this project linked to below is analogous to how the “intelligence explosion” is hypothetically supposed to happen. If this machine mass fixes open source bugs then that makes everyone more productive and accelerates things even faster and so on.