Fast Company

3 Ways Google's Majel May Out-Chat Apple's Siri (And 1 Big Way It Could Remain Speechless)

Siri is amazing. Let's not pull any punches--ignore the critcs, as it's probably the best mobile voice recognition device that any user has ever encountered. Above all else, it's a fabulous PR tool, it works, and it threatens Google ad revenues. Google and its Android army know this. Why else would Samsung lampoon it and Google execs poo-pooh it? Especially when you consider that Google is desperately working on a rival to Siri. It's codenamed Majel, and in several ways it's got a serious chance to outsmart Apple's smart assistant right from the get-go.

Great Implementation

According to leaked information that reached the site AndroidAndMe, Google's project to rival Siri (being cooked up in the top-shelf division Google X) was supposed to be ready before the end of 2011, and it's dubbed Majel. Though it looks like Google may miss its timing window, the goal was smart: By launching a Siri rival soon Google would be seizing on the pro-Siri publicity, preventing Apple from marching far ahead in a new paradigm.

And even the name is clever: Majel is named for Majel Barrett-Roddenberry, wife of Star Trek creator Gene Roddenberry. This is enough to warm the cockles of hardcore sci-fi fans. The Star Trek computer reacted to natural language, was incredibly smart at returning all sorts of information--scientific, historic, reference-based, inferred and real-time-tactical--and it worked in that classic sci-fi way, only returning a response after, typically, Captain Picard started a sentence with: "Computer..."

To some minds this is how a smart digital assistant should work--less slick and shiny than Apple's Siri, with a more toned-down personality and a more structured dialog format. That way there's less of the supposed creepy or weird factor of talking to one's smartphone rather than talking through it to someone else (although I personally admit to confusion on this point--you're still chatting to a plastic slab held to your face, so how can a passerby know any different?). If the codename is a sign that Google is trying to make its Majel a more digital, less personal assistant--perhaps even responding with tones instead of voice cues in some instances, then this way of interacting with the service may well appeal to many folk who dislike the idea of chatting with Siri.

Android Integration

The other thing that's hinted at in the leaked information is a deeper integration of Majel into Android than Apple's allowed with its first beta release of Siri in iOS. Majel seems to be a smarter version of the existing Voice Actions system in Android, a service that's meant to be threaded throughout the mobile operating system to bring voice commands to more sophisticated uses than merely "call Mom"--something that many phones already offer, and that Apple's had for a while in iOS (we guessed a while back that Voice Actions may be one of Android's secret weapons, though it's never quite happened that way).

Where Siri can't send a tweet, do complex navigational tasks, or operate inside apps (with app-specific vocabularies), it would seem that Majel is designed to do this right from the start. That would instantly one-up Siri, even if at first this implementation was limited to Google-specific apps, like Gmail or Voice or Google+ or a limited range of systems that include social networks like Twitter. It's also plausible that smartphone carriers could enable specific powers via Majel--since Google permits carriers to fiddle with the services Android offers on the smartphones they sell, and for some offerings from carriers this could be a good thing.

As such, this integration would be a great PR coup--aceing Siri straight away, since many frustrated users (perhaps miffed by Apple's rare use of a "beta" release) complain Siri can't do this stuff. Apple has itself said that Siri will rapidly evolve, and it's clear from the powers Siri had when it was a standalone app that Apple probably will integrate these kind of smarts eventually. According to the leaked information, Android's engineers are working so speedily that at first this kind of integration may not be enabled, but it would likely follow soon after.

Google Integration

One thing Google has at its advantage is a host of different search services that it could blend right into the output of Majel. Where Siri stutters, then recommends "Would you like to try a web search?" Majel could probably just seamlessly wallow in Google's sea of different data and come up with an answer that matches the kind of information that the user is asking it for. This is for all sorts of services, but one immediately stands out--translation. Ask Siri how to say "hello" in French and she can't answer, directing you to that ubiquitous option to search the web. Majel could probably work out your query at first stab.

Google may even be acting to shore up its resources for Majel--aiming particularly at the kind of data that users may use it to look for. Just last week Google bought Clever Sense, the team behind the Alfred personal assistant app for finding good restaurants and bars ... and it's not too much of a stretch to see how that would work within Majel: "Computer: Find me a great sushi bar nearby." Back in September it also bought Zagat, which gives Google access to high-quality crowdsourced reviews, which its algorithms could crawl all over to return a smart response to this sort of Majel search. With the company, by its own admission, acquiring roughly one firm a week it's plausible that many other recent buy-ups are designed to support Majel.

Think of Majel as a voice-control portal to all of Google's different services, from patents to web searches to Google+ to Voice to Translate to Maps and so on ... and you see how powerful it can be. Of course much of this is possible with Google's Voice Search anyway, but to have it all delivered via one interface and with a smarter, more natural language-friendly front end would help. If Majel manages to deliver the most relevant information to a voice query, sampling it in a context-aware way from among Google's huge search resources and answering in a concise and simple way (without requiring further user screening, which you typically have to do to a Google search, and which Siri avoids) then it could easily champion its powers as beating Siri's.

Google-Centricity

But Google's vast search database is also potentially Majel's Achilles heel. 

If Google closes its shutters, and keeps Majel's information searching skills confined to Google property--for technical reasons, perhaps, and because it guarantees more user-eyes-on-adverts time--it is actually a limiting rather than an expanding maneuver. Part of Siri's strength is that she acts as a high-level filter for search--being designed by the team at Apple to look at the user's query, then try to respond from a more precise and relevant data source than simply resorting to a web query through Google or Bing. It's why integrating Wolfram Alpha was a genius move, and it's also why Google hates Siri--because it gives Apple control over where search queries go, and that could mean diverting some away from Google. 

But if Google keeps Majel centered on its own properties then potentially it means users could miss out on richer, more relevant data sources that Google doesn't yet quite rival--firms that Google hasn't bought, or can't acquire, and which it could thus find difficult to data-mine to feed data into Majel responses.

And while Apple got into a spot of political bother over seeming search biases in Siri--actually based on a misunderstanding of both the readiness of the system, as well as how it and its various data sources work--Google could get into much more legal hot water if it championed a premiere new search function of its market share-leading Android phones...and it turned out the system was mainly pushing search queries to Google. Any number of anticompetitive bells could ring out.

Oh, and let's hope Google doesn't make Majel too U.S.-centric. That's a mistake Apple slightly made with Siri, which is why non-U.S. users (who make up the greater majority of iPhone owners, historically) seem less pleased with it. 

Ultimately then, Majel has a shot at having a different character and behavior to Siri and this, coupled with Google's simpler access to the search database, means it could beat Apple's voice assistant in many ways. It'll all be about the finesse of how Majel interacts with the public. Get that right, and dodge legal and user-experience issues about keeping searches within Google, and Larry Page's company might have a killer product on their hands.

[Image: Flickr user telstar]

Chat about this news with Kit Eaton on Twitter and Fast Company too.

Add New Comment

8 Comments

  • Heaven

    i am no expert but there are lots of thing that you missed that will help Google outgrow Apple, things like ADK and also that Android is open source, i think Google's 'Majel' will be more or less the biggest competition to Siri undoubtedly. 
    http://bit.ly/Googlemajel

  • aquatiger

    Two things that I noticed in this article.  There are way too many coulds, shoulds, and mights, and there is no release date yet.  I love Android, but the problem is that if they plan to match or barely beat Siri, by the time it comes out they will be behind again.  Guaranteed.  We need a leapfrog feature that puts Android one step ahead, and let Apple play some catch up then.  

    And if you think that Apple is just sitting around patting themselves on the back about Siri, I'd think twice.  You know those guys are brewing something else that will just make us question Android.

    And Kyle, have you ever used Google Voice?  If Google has a team with such deep experience with speech recognition, they need to be fired.  I use Google Voice for my voicemail, and it gets about 5% of the words right on transcription.  Best case scenario is I can kind of maybe get the gist of what the voicemail is about.  It's almost comedy.  They have a boatload of work to do.

  • Kyle Lyles

    Yes, I have been using Google Voice since Google acquired GrandCentral.
    I work in speech recognition, so have insights into what they are doing with it.  

    Note that every single voicemail you received transcribed to text has two ways to help Google tune the hundreds of millions of VMs they process: Over on the right bottom of the VM is "Transcript Useful?" with a positive checkbox or a negative checkbox. Then by Call Text, is more....  Click there and then edit transcript.  You can then tick the "Donate this Voicemail" once you've corrected it.  The update then goes into the tuned grammars to improve the offering.  This is crowd sourcing their speech tuning rather than paying $150,000-200,000/yr each to speech scientists to tune this loss leader.

    Google is not focused on improving the VM function of Google Voice through expensive tuning of caller utterances.  They are however, very focused on tuning the grammars in their ASR offering that is accessed through their API.

    So you are comparing two totally different services, one that is a profit loser (Voice) and the other which is key part of their mobile program.

    Note again - Apple IS sitting on their backsides about speech reco.  They have ZERO.  They are buying it from Nuance.  Apple has their family jewels laying on a stump and has handed an ax to Microsoft and Google.  Who would you bet on right now?  It sure isn't Apple.

  • aquatiger

    Unfortunately people have been saying that for years.  Every single time Apple comes out with something it's their "last" good thing.  Yet the facts remain the same.  Google gives Android away for free, and makes exactly zero dollars per app in their App store.  I don't buy for one second that they are running two different voice groups.  They've axed too many things this past year for them not to have consolidated.  Again, I love Android, but it makes peanuts compared to Apple in terms of dollars.

    Click on the 10 year data:  http://www.google.com/finance?...

  • Kyle Lyles

    One point not made here is that Apple has ZERO speech recognition expertise.  Google and Microsoft both have deep experience with speech recognition and have their own teams.  Apple contracts with Nuance Communications NASDAQ: NUAN for these services. 

    Nuance is heavily involved in the healthcare and automotive industries - both areas of Microsoft growth. They could easily acquire NUAN for ~$10B with no risk of anti-trust interference, and Apple would be MSFT's customer. 

    GOOG has the original "Blue" Nuance team (current Nuance is formerly SpeechWorks morphed into ScanSoft, morphed into the "Green" Nuance) that came out of SRI Int'l.  Their acquisition of NUAN would provide them considerable patent protection around the work of their speech recognition team, and complement their recent Motorola acquisition.

    If Apple wants to stay competitive, they better get their checkbook out and add a few $B before they are beat to it by either MSFT or GOOG!

  • David Kaiser, PhD

    I'm looking forward to this. As a former computational linguist, I have been eagerly awaiting the day a machine can handle a natural language transaction cleanly. We're getting close.

    David Kaiser, PhD
    Executive Coach & CEO
    www.DarkMatterConsulting.com