But no one has yet to solve the two problems plaguing SR: First, putting a whole SR engine on a machine let alone on a pair of embedded chips on a piece of eyewear for instant SR. Even with miniaturization and speed/power improvements I just don't see it happening mainly because chip makers have decided to just add more cores. I may be wrong but it seems difficult to parallelize this with a large corpus.
Obviously the solution is to go the other way and have super fast guaranteed wireless connection at all times. This will still produce lag but probably acceptable levels for most people.
The second problem is one that most people encounter with SR which is when the bloody thing doesn't translate what you said. Either because of an accent or because you're saying something unique or difficult to parse. I tried asking directions to a Mexican place in OC and the results were: Q Cortes, cute protcullis, cute Portos, Q Portales. Can you figure out what I was trying to say? That is the inherent difficulty of the "last mile" problems that SR has and I haven't read or heard anything about anyone solving this anytime soon. SR tech will always be just "good enough" where they can show it off at a meeting/conference but never good enough to be like Star Trek (which unfortunately has set a incredibly huge bar for people's expectations).
Even though the Nokia video is less ambitious they seem to have decided to not include any talking in their future tech videos which seems like a smart idea to me.