A conversation is not a product. Conversations aren't tied to practical concerns or engineering constraints. Conversations result in interfaces like the one in the demo: there isn't even recognizable input triggering most of the actions.
I mean, do we really think Google will be able to pull off a context-aware "hmmm" as a vocal trigger signifying desired interaction with a pop-up alert?
Microsoft has more processing power available to the Kinect and I can't even hold a conversation in the same room as a Kinect-aware game without accidentally triggering commands. [1] Becoming notably more accurate would require far more local processing, or shipping vocal commands off to the cloud for processing,introducing a lag that's wholly unacceptable for UI navigation. [2] Particularly for UI navigation in a product that's usefulness is predicted on it being quicker and easier than just pulling out your smartphone.
And if you don't know how people interact with the product, how can you usefully investigate what they can or will want to do with it?
If it turns out that the only dependable ways to interact with such devices involve socially dubious practices [3], they may only find popular use passively displaying various augmented reality data layers over identifiable objects.
That truly may be enough. But the result will look nothing like that concept video and no-one will have profited from Google having produced that video.
Don't get me wrong: fiction and futurist musing is great. But it matters who's having the conversation. When a corporation speaks about a product or technology, expectations are inevitably set. And Google's first pass at Glasses is bound to fail the concept video. Successful heads-up devices are bound to be nothing like that video. [4]
But instead of weighing the product on its merits and driving the product in the directions that work, people are going to be discussing the ways in which it fell short of the video and quite possibly trying to force those interactions onto the product, even where they're not working.
Consider Microsoft's Tablet excursion. (If nothing else, Google's Glasses project strikes me as a direct parallel to that.) Microsoft tried to force a full version of Windows onto the Tablet, despite it being unsuited to the form factor, despite the interaction method being wrong, despite the technical strain it put on the already-limited resources available to a mobile device. And they tried to force that to work for a decade -- long after it was obviously not going to work -- because they'd built up expectations based on a flawed conversation at the outset. [5] Expectations from management, developers, users and even shareholders had been set and the result was ten years spinning their wheels.
[1] Mass Effect 3 sure thought I asked it to throw a grenade, even though the conversation I was having didn't even include the word grenade. I can only imagine how annoying always-on voice control would make a head's up device in public.
[2] Local processing that seems increasingly unlikely for mobile projects, given that everything is shipping voice commands off to the cloud these days. And Apple's Siri is only borderline-tolerable for higher-level requests; it would be a bag of fail if it were asked to process 'open email' sorts of commands.
[3] Speaking to yourself, or donning a finger-sock and waving your arm about like a mad-man will naturally lead to people trying to do that as little as possible. (said finger-sock simplifying a digit-tracking mechanism)
[4] The only plausible features from that video are simplified camera access, hands' free calling and turn-by-turn navigation. And two of those are doable with existing bluetooth devices, while the third is only useful if Google can deliver a massive service upgrade to Maps, to include up-to-date coverage of building interiors.
The rest of that video are things people are increasingly dubious of now (always-on alerts and check-in based applications) or simply don't include a plausible interaction method.
[5] A conversation wherein "Windows everywhere" became the goal, despite any (in)applicability to the product.