Hacker News Reader

Comment by blowski | Hacker News Reader

If that's the main criticism of Alexa, then I'm rushing out to buy one immediately.

I'm not interested in an always-on voice activated device so that I can ask it rarely asked complicated questions. I'm interested in how well it understands my voice, and how many day-to-day chores it can help with.

View on HN

The ability to understand you involves understanding language. You have to understand the world to understand language. Google built a 70 node knowledge graph that drives their understanding of the world which enabled them to create a intelligent device. Or something that is more revolutionary versus Echo is more evolutionary.

I have both and purchased the Echo when first launched. The Google Home amazes me every day.

So with Home you say play Madonna song Sean Penn movie and live to tell starts playing.

Echo has a command of "goes like this". Never used it. But everyday Google figures out something.

Our understanding of the world allow us to talk in a condensed form as humans can infer a lot.

Google is doing the same so more and more I talk to Google Home in condensed format.

It is so early with all of this but Google has a incredible foundation. The future is going to be incredible.

Should have been 70 billion. :)

Seconding this, Google voice search (on my Nexus phone) is incredibly good. If you've spent the last 18 years expecting valid results from Google searches even when you dump a bunch of vaguely related words into the text box then Google's voice search gives you the same experience.

I could be wrong, but isn't the voice search merely a voice-to-text layer across the top of the normal text search? My old BlackBerry had a search feature and voice command. To this day I think it is slightly below the quality of Google Home (played with it at a family member's home; don't own one myself).

My daughter does voice searches, asking it to display pictures of various pokemon. "Show me pictures of the pokemon XXXX". These things have crazy names, and she doesn't have very clear speech, but it always gets them right. It is changing it's dictionary of probable words you will say, based on what words you have already said.

If she unclearly said, "Show me pictures of Absol", it would fail miserably. But if she says "Show me pictures of the pokemon Absol", it kicks butt.

So I would say it's more than a voice to text layer on top of the normal text search.

Reply on news.ycombinator.com