Hacker News Reader

Comment by drats | Hacker News Reader

I fairly certain Google can leapfrog Apple here, it's a machine learning company with big data built into its DNA vs. a computer company with Siri grafted on. Google simply has more training data and more expertise to build this on top of their already impressive voice recognition (which like their translation facilities just smoothly gets better and better).

I am also certain people will scream that Apple "invented" this despite the decades of public and private research on the problem and the fact that IBM's Watson (far more impressive than Siri) was released and won Jeopardy before Siri was even launched.

View on HN

"it's a machine learning company with big data built into its DNA"

Is that really an overwhelming advantage? On the one hand, "data" definitely helped Google build solid speech recognition, on the other, Google has shown us time and time again that big data isn't an advantage in creating graphical user interfaces. My sense is that the problems in an AI interface involve just as much design and interface expertise as they do data expertise.

Siri has pretty much no interface. The AI/data part of the problem is much, much harder.

The way Siri converses & jokes with the user is the interface. Not all interfaces are graphical.

saturdaysaint said GUI.

I mentioned GUI as an example of an analogous domain where Google's strengths weren't necessarily of any advantage.

No, he did not. He mentioned GUIs in another part of his comment --and then went on to comment on the AI interface:

"""Google has shown us time and time again that big data isn't an advantage in creating graphical user interfaces. My sense is that the problems in an AI interface involve just as much design and interface expertise as they do data expertise""".

Making a UI disappear is much harder than it seems.

(and I don't mean to compare it to the data mining problem, because you just can't)

Siri very much has an interface -- it's just not the graphical part. The voice interface is still an interface, and still needs to be designed as naturally and nicely as possibly.

I'd hope nobody screams that Apple invented this, since it's a matter of public record that Siri started out as a DARPA project and was purchased by Apple a mere 18 months ago.

Very few people say that "Apple invented X", especially in geek sites. What we DO say however, and are most often correct, is that "Apple popularized X", "Apple finally made the first version of X that masses of people want to use and use", "Apple finally got X right", etc...

Laypeople do it all the time.

So what? They also make worse mistakes that that harmless one (it's not like who made the original invention matters much, except for proper historical attribution: I think of it in the way we usually think about idea vs execution. The "original invention" is closer to the "idea". It's who put it on our households that did the actual "execution". I can use Linux GUIs, OS X, Windows but could care less about Xerox's original UI).

Also, for some profession or other, we're all laypeople, and make equally bad or even graver mistakes. But we don't care that much about those. A computer geek feels superior to other people for knowing that Xerox invented the modern GUI, but he doesn't feel stupid to believe any number of BS about medicine, economics, literature, etc.

Data? What data? Siri is AI which tries to understand what I said and then executes, it's intelligent command line processor. All the words in English and their meanings are very small data, nothing that Google has and nobody else not.

Data? What data? Siri is AI which tries to understand what I said

By using a vast amount of data. Words, phrases, accents... it all involves crunching a lot of data. That's why Siri isn't actually intelligent- if you ask it a question it hasn't been programmed to answer it doesn't work out what you're asking for- it just tells you that it doesn't understand.

So what specifically in this crunching is so specific that Google has an advantage?

Google has had billions of people asking billions of questions to it, and then has seen what answers (webpages) they find most satisfactory (i.e the navigate to from the search results).

It doesn't have to be specific. Google has extensive experience in crunching incomprehensibly huge datasets into meaningful, end-user responses. Apple does not.

Reply on news.ycombinator.com