Samsung Galaxy S3's S Voice vs Siri on iPhone 4S

When Apple unboxed the iPhone 4S it also uncorked Siri --- the sometimes smart-arsed robotic genie/butler that can fetch stuff from your calendar, tell you what the weather's like or do a web search for Kevin Costner's age. Now it has a challenger -- the Galaxy S3's S Voice.

Siri was not the first voice-control app on a phone -- limited use of voice recognition has been kicking around for ages, such as BlackBerry's 'say a command' feature. Apple made some big claims for Siri, though -- this was apparently not a dumb interface which only understood a very simple set of barked commands. Instead you were encouraged to natter away to Siri like it was a real flesh and blood buddy.

The big problem for UK iPhone owners is that Apple still hasn't got a deal in place with a local information provider -- so if you ask Siri to find a restaurant, direct you home, tell you how bad the traffic is or ask it what's on at your local cinema, you'll get the same, apologetic response: "Sorry, I can only look for businesses in the United States... " Lame.

Now, with the launch of the Galaxy S3, Samsung is lining up its own voice assistant app as a Siri rival. S Voice builds on the Vlingo Voice Talk app it baked into the Galaxy S2. So how does S Voice measure up against Siri? I took both phones off to a nice quiet interrogation room and got barking.

NB: Neither phone had become accustomed to my voice prior to this comparison.

Round one: Call Zack

This is a really basic question. It's akin to a voice command of old -- especially as the contact I've chosen has a short, distinct-sounding name. It should be easy peasy lemon squeezy for our intrepid robotic butlers with their mad natural language processing skills.

S Voice: And Samsung's showpony stumbles at the first hurdle -- braying out, "I didn't catch that. Please try again." But on my second attempt, S Voice picks up the baton and gets Zack on the phone.

Siri: Apple's posh-toned one has no trouble understanding my intentions and calls Zack without delay.

The first round goes to Siri.

Round two: What's the weather like today?

Again, keeping things simple, this query should be bread and butter for our mobile secretaries. So how do they get on?

S Voice: Once again, Samsung's vocal one fails to understand the first time I ask my question, and asks me to repeat it. On the next pass it gets as far as "what is the weather" but misses that I'm asking about today's weather, so doesn't understand my question -- and instead asks if I'd like to do a web search for 'what the weather'. Useful, not.

I try again, and more misunderstanding follows. Eventually S Voice pins me down to 'what the weather like today'. Things look promising and it does return weather info -- but for New York, not London (despite location settings, GPS and Wi-Fi all being turned on). Fail.

Siri: On my first attempt to ask what the weather's like today, Siri helpfully informs me that in order to do this I need to turn on location in the settings. Once I've switched this on I ask again and Siri returns in a jiffy with a forecast. At this point, as S Voice flounders, Siri is not even breaking a sweat.

The second round goes to Siri.

Round three: Do I need an umbrella today?

A slightly more complex query, this requires our virtual assistants to think more like us fleshbag humans and work out that the moistness (or otherwise) of the weather is what's at stake here.

S Voice: Oh deary me. It fails repeatedly to understand this question -- after five attempts, I give up. During this time it offers to web search 'I bring my friend today'. The concept of an umbrella appears utterly alien to it. I try to make things easier for it by asking, "Is it going to rain today in London?" but even this gives the S3 a headache. "I'm not sure what you mean by E in London tube," it moans cryptically. Useless!

Siri: My first attempt to ask about the need for an umbrella also foxes the iPhone. Siri returns one of its stock responses: "Sorry, I can only look for businesses in the United States, and when you're using US English." Since an umbrella is not exactly a business I can only assume my British English accent is upsetting it. I try a second time and Siri redeems itself, with a confident prediction: "It sure looks like rain today," accompanied by a mini weather forecast. Spiffing!

The third round goes to Siri.

Round four: Text Zack

It's time to ask my robot pals to pass on a message for me. This could get complicated...

S Voice: It starts well, understanding my command to text Zack. It asks me, "What is your message?" So far, so good, but the message content proves to be an intractable hurdle. Instead of faithfully transcribing, "How's it going?" it wants to deliver, "Is it going?" which while not a million miles away from what I said, remains nonsense. I cancel that message and try again, but it just gets worse -- now it wants to send Zack a motivational fragment that I simply didn't utter: "You can!" I cancel that message too. After a third unsuccessful attempt I cut my losses and give up.

Siri: "Text Zack!" I bark, and Siri asks me what I want to text. I decide to make things a little harder for the iPhone, since it's been doing so well so far, so I ask it to text, "Bring an umbrella today, it's going to rain." Siri takes this down as, "Bring an umbrella today is going to rain." Not word perfect, but the sense is clearly there.

The fourth round goes to Siri.

Round five: Find a good restaurant nearby

S Voice: This should be a simple query -- it's certainly not particularly complex, but instead of understanding my request to recommend some positively reviewed eateries nearby, S Voice gets in a right muddle. After a few repeats of "I didn't catch that. Please try again", it renders my question as a single word: "Bed". I suppose 'rest' might be construed as 'bed' but only if you stop listening halfway through the word 'restaurant'.

S Voice follows up its bed-based revelation by saying, rather ironically it must be said, "I'm not sure what you mean by bed." You tell me, S Voice, you tell me. More failure to understand me follows, but finally it successfully grasps my question. Can it now find a good restaurant nearby? It cannot. "I'm sorry I don't have the answer. Would you like me to search the web?" is all it offers.

When I -- rather foolishly -- respond verbally by saying "yes, search the web", rather than just tapping the 'web search' button, I inadvertently send S Voice off round the houses. Now it thinks I want to text someone so it asks me who. I try to cancel this phantom message and it feigns deafness before trying to send a message to a contact called "sucks". Followed by a contact called "hello". And then a contact called "cancel".

As I repeatedly beat my head against the desk in frustration, I finally resort to sending a faux message (to Zack) just to quit out of this phantom message cycle that I never wanted to get into in the first place. What message does Zack get? "Directors meeting". Oh my days.

In a last-ditch effort, I try another tack, searching for "cafes in Southwark". S Voice turns this into 'cafes in Suffolk' but actually returns listing for Virginia Beach, US. Massive, massive fail.

Siri: This query runs straight into Siri's UK brick wall. "Sorry, I can only look for businesses in the United States..." Shame, but at least it doesn't take 10 minutes of toing and froing to work out that Siri can't do this. It's to the point, even where its limitations are concerned.

The fifth round has no winner.

Round six: Direct me to Tate Modern

Providing directions to a major tourist attraction 5 minutes' walk from the CNET office should be child's play for these high-end smart phones, packed to the gunnels with location-determining tech such as GPS, Wi-Fi and cellular radios. It should be, but is it?

S Voice: Samsung's S Voice help lists 'direct' as a command this robo-concierge can handle, so I'm hopeful it won't have too much trouble. And sure enough it confidently fires up Google Maps. So far, so good. But then it slaps an overlay atop the map -- which asks: "Did you mean for your end point: Mile End, The Mad Hatter Hotel, Hotel All Seasons London Southwark?"

These are UK locations, just not the UK location I was after. Another attempt and my question is transmogrified into: "Direct me a mile". Another go and the S3 returns the weather for Southwark... In short: frustrating chaos.

Siri: Asked to direct me to Tate Modern, Siri becomes confused -- almost charmingly so. "I don't understand 'Directly detainment'," it opens cautiously. On my next attempt it turns to philosophical conundrum: "Sorry, I don't understand Direct Nietzsche Tate Modern." A third try yields the nonsense-string: "Direct media take modern". I change tack and ask Siri, "Where is Tate Modern?" which finally results in the stock US-centric service limitation apology. 'Direct' is clearly not a command Siri understands.

The sixth round has no winner.

Round seven: Set alarm for...

I decide it's time to give the inhuman butlers a bit of a confidence boost after their recent struggles, so I return to a basic voice command. Can they set an alarm?

S Voice: I ask the S3 to set an alarm for 10.23 and the phone promptly responds with the correct alarm prompt. Success at last!

Siri: Apple's mouthy one also has no trouble tackling alarm setting. I ask it to set an alarm for 11.25 and that's exactly what Siri does.

The seventh round is a tie.

Round eight: How high is Mount Everest?

Okay, enough with the super easy stuff. It's time to test our software servants on their trivia-retrieval skills.

S Voice: Samsung uses the Mount Everest example in its S Voice help list, so I'm confident the S3 won't have any trouble answering this one.

But, oh dear, I'm starting to suspect S Voice is a little deaf. My first attempt sees, "How high is Mount Everest?" turned into "hi" -- which understandably confuses proceedings. I try again, and it hears "Are now". Finally it hears "Mount Everest" but that's all it's heard -- turning my question into a meaningless statement. After five attempts I throw in the towel.

Siri: I bark at Siri and Apple's digi-butler doesn't flinch. With a nonchalance boarding on the smug it says, "This may answer your question" and delivers a data sheet that tells me the height of Everest is 29,035ft. Slick.

The eighth round goes to Siri.

Round nine: Tell me a joke

After all this ceaseless barking at slabs of glass and plastic I'm feeling in need of some light relief. Can my robot friends tell me a joke?

S Voice: Moments after asking the S3 to tell me a joke, it seems like something is happening. An app is being fired up. The Samsung logo comes in to view and music starts playing. You could say this is a joke, but only if the joke is on Samsung. I switch back to S Voice and it all becomes clear: the phone thought I was asking it to play some music. Awkward to say the least.

Siri: Can Siri do any better? Siri has clearly been asked this before. Smug levels are at maximum as it responds with a quip about not understanding "tongue-in-cheek". I push my case for some better light relief than that once more and get: "Two iPhones walk into a bar... I forget the rest." It's pretty weak sauce I grant, and my sides remain unsplit, but it is technically a joke.

The ninth round goes to Siri.

Round 10: Search bonobo apes

S Voice is definitely having the tougher time of it so again, I opt for another question that is given as an example query in the S Voice help list. Search plus the search term takes us back into voice command territory, but perhaps the S3 can make up some ground here?

S Voice: The S3 steps up to the plate, but instead of searching for bonobo apes, it hears "send sa", followed by "send message me" -- and once again we're back on the entirely unwanted phantom message treadmill. Now it tells me there's no one called 'me' in my contacts. Not. This. Again.

Siri: By contrast, Apple's butler is mockingly slick -- rattling off: "If you like, I can search the web for 'bonobo apes'" -- picking up the ball S Voice so clumsily dropped and throwing it back without hesitation.

The tenth round goes to Siri.

Round 11: Make me a diary appointment

What use is a robot assistant if it can't keep on top of your calendar?

S Voice: The S3's wordy one clearly prefers rather staid commands, so I follow the help advice and use the phrase "new event" -- making up an imaginary lunch date with Zack on 21 July 21 at 1pm. I'm hopeful that S Voice can handle this, but at first it returns 'network error'. I put this down to Wi-Fi and/or 3G gremlins and try again, and after thinking about it for an insolent amount of time, S Voice successfully schedules a meeting -- albeit without a title or a reference to who the meeting is with.

An appointment without info about who I'm supposed to be meeting is not a proper appointment in my book, but I'll give S Voice the benefit of the doubt and chalk that up as a (tepid) win.

Siri: I ditch the 'new event' wording and ask Siri to "schedule an appointment with Zack at 1pm on July 21st". Its response is fast and slick: "Okay, I set up your meeting with Zack... " I know which assistant would certainly not be getting fired today.

The eleventh round is a tie (just).

Round 12: What's your favourite colour?

I feel like I should get to know a little more about my not-always-trusty assistants, seeing as they are working for me. Can they tell me what their favourite colour is without getting their knickers in a twist? This question doesn't help me in any way, but if they can understand queries like this it might convince me they're more than just a disembodied robot voice. Or not.

S Voice: Finally the S3 doesn't hesitate and ploughs right on in to an intelligible answer. Or it would be intelligible if S Voice didn't dictate symbols as if they were words. So the resulting answer sounds a little something like this: "Blue. No yel hyphen hyphen auuuuuuuugh!" In the battle to convince me of its human credentials, S Voice just fell on its hyphenated face.

Siri: In a typically slick display, Apple's virtual Jeeves responds on cue with the following: "My favorite colour is... Well, I don't know how to say it in your language. It's sort of greenish, but with more dimensions." I'm not sure Siri has made me think of it as any more human, but at least it paused in the middle of its answer, rather than articulating 'dot dot dot' or the word 'ellipsis'.

The twelfth round goes to Siri.

Round 13: What's the best smart phone?

I'm giving the robot butlers a chance to boast -- if nothing else this should be entertaining...

S Voice: Instead of "what is the best smart phone", the S3 hears "what the best smart phone" so consequently returns with: "I'm not sure what you mean by what the best smart phone". This is just getting embarrassing.

Siri: Of course Siri has no trouble answering this -- or rather not answering it, and thus answering it by implication. It returns: "You're kidding right?" Full marks for savvy and sass to Apple then.

The thirteenth round goes to Siri.

Round 14: Update Twitter

Relying on a voice assistant as a quick way to update Twitter could be pretty handy -- assuming it works. Let's give it a shot.

S Voice: The S3 grasps what I'm up to first time and handily tells me I need to authorise the Vlingo for Android app so it can post tweets on my behalf. After authorising the app, I try again and one "I didn't catch that" later, it's offering to post a tweet for me. The wording isn't 100 per cent perfect -- I said 'testing S Voice' it heard 'testing voice' but it's not a bad effort. Huzzah!

Siri: For Siri, it's another service limitation brick wall as Apple has not authorised Siri to access third-party apps such as Twitter. "I can't send tweets for you. Sorry about that!" is the response.

The fourteenth round goes to S Voice.

Round 15: Speed

Throughout this test it's been noticeable how much quicker Siri is than S Voice. Take the relatively straightforward trivia query: "What's the capital of Iraq?" -- how fast do they each respond?

S Voice: This mundane query sends the S3 off into a 20-second reverie. S Voice does return the correct answer -- Baghdad -- but it sure makes me wait for it.

Siri: Siri also returns with the correct answer, but it takes around 5 seconds -- a considerable margin faster than Samsung's robo butler.

The fifteenth round goes to Siri.

Overall voice champ

S Voice takes a proper pounding from Siri in this comparison, losing by a margin of 10 rounds to one. Two rounds were tied, and two rounds had no overall winner.

Despite Siri giving S Voice a verbal beatdown, neither voice assistant comes out of this test with a great deal to recommend it. S Voice's usefulness is constrained by what a poor listener it is, while Siri is severely limited by not being able to respond intelligently to Brits' surroundings. So both apps are gimmicky -- albeit Siri is the slickest gimmick of the two.