How do new speech recognition companies operate with so many patents in this space?
> Disclaimer: I’m not a venture capitalist (but I play one on TV).
People patent things they can’t make, and people make things they can’t patent. Acquisitions love patents as evidence of intellectual property; but, there is no substitute for original code. If you don’t aim to be acquired, and if you don’t sell a physical product, then for services the point may be moot. I also believe a lot of speech recognition has been based on re-writing open source programs. Today, people can train speech recognition on neural networks themselves, which I believe can’t be patented. Otherwise, Nuance Communications seems to enjoy the lion’s share of speech services, and is used by many if not most of the major players, at least initially.
See also my Quora answers to: