Is speech recognition really that close to being perfected? I haven’t used anything like Dragon, but I know Siri and Windows Speech Recognition, two of what should be the highest quality pieces of consumer speech recognition, still struggle with me constantly. I would love to dictate my notes or documentation, but trying to do any formatting, let alone speaking uncommon words, ends up making a mess of things. Even things like “text my wife ‘I just got on I-90’” is tricky, because Siri doesn’t recognize that I’m talking about the road. I would think 3-d printing is much further along, unless we’re calling speech recognition solved well before it can understand what we’re saying.
[Comment removed by author]
On their hype cycle graph, Gartner puts speech recognition at the “plateau of productivity” stage, natural-language question answering at the “peak of inflated expectations” stage, and virtual personal assistants at the “innovation trigger” stage. I guess “innovation trigger” refers to a well-received movie that came out recently?
It does seem that, by speech recognition, Gartner only means translating sounds received through a microphone into specific words in a text format.
As far as understanding those words, Siri, for example, won’t parse anything longer than a sentence. Short, iOS specific commands work well enough, but still, like your example shows, Siri lacks enough context to translate flawlessly.