I’m convinced that in a decade or so every bit of technology will be based around vocal commands.
And I’m going to name all of them HAL.
Real translation in real time. Genuine, Hal type, AI. This is mainly software.
The Dec. Scientific American has an article about how a certain breed of wasps (Polistes fuscatus) can recognize individuals by their distinctive facial features. Most other breeds, even in the same genus, cannot do this. So their pea brains are routinely carrying out a task that modern very large and very fast computers can do only poorly. There is a lot of room for software improvements. Incidentally, the article mentioned software facial recognition only in passing, but I wonder if genetic algorithms (where you let solutions “evolve”) aren’t the way to go on this.