backtop


Print 6 comment(s) - last by jharper12.. on Nov 14 at 6:02 PM

Want to know how you'll be talking to your car in 2 years? Check out what we learned

DailyTech on Tuesday attended an interesting roundtable discussion hosted by Nuance Communications, Inc. (NUAN) at the Westin hotel in Detroit, Mich.  Co-sponsored by Ford Motor Comp. (F) and Pandora Media Inc. (P), the event included a four speaker panel, which included Pandora Automotive Business Development director Geoff Snyder, Nuance Senior Product Manager Ed Chrumkra, Gracenote (a subsidiary of Japan's Sony Corp. (TYO:6758)) VP Vadim Brenner, and Ford voice control engineer Briggite Richardson.

At the panel and in the demos that followed a number of interesting insights were revealed, which are broken down below:

When will cars talk naturally?

One commenter at the event pertinently pointed out, that even the most advanced voice control systems on the market primarily use stilted commands.  

It was a good point. (The event was, after all, titled "Let's talk about talking to your car".)  Let's face it -- you may talk to your car, but it sure doesn't sound like a typical conversation between humans.

Ms. Richardson with Ford responded that infotainment systems tend to do this to improve accuracy -- making sure the system really catches what you said.  She pointed out that Ford does offer some natural-speech aliases, but that the emphasis is still on a command language.

Nuance's representative Mr. Chrumka added that a big part of natural speech recognition is that it requires either a powerful processor locally or data exchanges with a powerful cloud server.  Nuance and other companies are developing some pretty impressive natural speech solutions on the PC.  

Cars and smartphones currently lack enough high speed throughput to make a solution that's as strong as what you could have on a fast PC.  And these devices also lack the local processing power, he argues, to accomplish moderately capable natural speech, so they must try to leverage the cloud.  And these efforts are just starting, he adds, with commercial (mass-market) solutions a year or two out still -- at least when it comes to strong natural speech.

You'll see an increased amount of natural speech creep into luxury systems like the Cadillac Cue, which launches in 2012, in the meantime.  But even these systems will be a work in progress.

On smartphones versus discrete modems for the car...

Will "connected" cars eventually receive data from built-in modems, like cell phones and tablets, or will they receive data through a tethered connection, like a smartphone's Bluetooth connection?  

Ford smartphone tethering
[Source: Laptop Magazine]

The panelists seemed to primarily feel that tethering to smartphones was the way to go.  Mr. Snyder comments "We definitely leverage the smartphone as the vehicle to get into the vehicle, so to speak. [laughter]"

Will data caps impact in-car tethering?  How will 4G services impact in-car tethering?

These questions provoked a lot of vaguely worded nebulous talking.  The answers told little, but the way they were told did give some hints.  

For the capping, the nervous responses which mostly summed up to "we're going to wait and see where the market goes", seemed to indicate this is something connected car folks don't know quite how to deal with.  Remember, the top connected car players view the smartphone as the penultimate data source, and when your data source hits you with big fines, that's a big problem.

With regard to 4G, the nonresponses seemed to suggest that the answer of added benefits is "not much".  Cellular data coverage, at this point, is arguably far more important to the connected car than a mild increase to speed in select markets.  And even coverage can be mitigated to a degree, as Mr. Snyder pointed out. (Pandora uses local buffering to overcome temporary zones of no-coverage.)

Will Ad-Supported Infotainment (e.g. "Free") Systems Become Reality?

Today even the biggest newspapers like The Wall Street Journal offer much of their content for free, if not from their homepage, through services like Google.  This content could have cost you hundreds of dollars a year, in today's money, to obtain on a daily basis over the course of a year.  Thus it's not infeasible to imagine that advertising could eventually be used to pay for or partially pay for an in-car infotainment system.

Could car buyers one day have an option of picking an infotainment system (like Sync, Entune, etc.) for free that is paid for via ad-revenue?  (Buyers would likely also have the option of ponying up cash for an ad-free version.)

Here's the idea.  You embed advertising at a level beneath the core commands, so it doesn't distract, but over top of the radio and entertainment features.  You can choose between an ad-free version or a discounted version, which occasionally interrupts your media and plays you a locally-targetted ad.

targeted ad
Nuance feels that targeted advertising has the potential to make infotainment systems free or greatly discounted.  Ford's representative voiced skepticism. [Source: On the Way]

Given the data mining prospects, Mr. Chrumka seemed very enthusiastic about this concept.  He said, "We absolutely see that coming to the market."

He adds, "I absolutely think those things are being examined.  [In fact] I know they are."

Ms. Richardson with Ford wasn't so enthusiastic.  She said Ford's current perspective was that advertising could be used to offer "free apps maybe", but that it would be tough to ever "justify the cost of a hardware unit."

No Siri? No Problem?

It's sort of a poorly kept secret that Apple, Inc.'s (AAPL) Siri -- the company's talkative iPhone 4S assistant -- is essentially a Nuance application which Apple has slightly modified and rebranded. 

One Nuance engineer during the Q&A even outright said that Nuance's property was at the heart of Siri.  Others were more politically correct and quoted the official party line -- that Apple licenses technology from Nuance for unspecified purposes.

So the real question is -- "What do iPhone 4 and 3GS owners do now that Apple has yanked Siri from the App Store, making it an iPhone 4S exclusive?"  

A competing (or complimentary) solution is Dragon Dictation and Dragon Go (the new version of the older Dragon Search app).  Both apps are in their second generation on the iOS platform and have matured greatly in capabilities.  

Nuance's first mobile app solution, Dictation allows you to dictate messages, which are interpreted via cloud processing and neatly displayed.  You can then copy the blurb, send it as an SMS, send it as an email, or more.  

Dragon Go is much more Siri-like.  It allows you to speak a natural language query or command (e.g. "What's the Lakers score?") and the app will parse the question/command and use it populate searches or forms in a host of partner websites.  For example sports questions would take you to various sports sites, asking for a concert ticket would tak you StubHub, asking for a dinner reservation would take you to OpenTable, etc.

Dragon Go, in particular, is very cool.  So cool in fact, that not having Siri suddenly doesn't seem like much of a problem.

Dragon Dictation

The only baffling thing?  The slick Dragon Go is only available for the iPhone, iPad, and iPod Touch.  Why no support for Android, by far the world's most used platform?  

Nuance representatives responses seemed to indicate a general cluelessness as to why support hadn't yet been added.  About the only thing we can think of is that Apple paid Nuance an undisclosed sum to stay off Android for a specified time.  The alternative -- that Nuance can't program a decent Android app like the rest of the big players just seems too farfetched.

Regardless, the prospect of more downloads, and thus more ad-revenue from the company's website partners, is dangling in front of Nuance's face and will be hard to resist in the long run.  Indeed Nuance's representatives promise that Dragon Go is being ported to Android as stand alone apps and will launch sometime in the future (though they refused to say exactly when).

Dragon Dictation is already available packaged within the framework of some Android devices, but Dragon Go is not available yet on any Android device for whatever reason (possibly exclusivity).

Texting and driving...

Nuance has a plan to make the much loathed texting finally safe for the car.  It's leveraging an in-car display, a connection to the cloud (e.g. a smartphone), and a control knob to allow for full dictation of outgoing messages, while on the go.

Texting while driving
This is not Nuance's texting and driving solution. [Source: Alabama Injury Attorneys]

To start, you merely click the knob in.  While in the demo this knob was discrete, we could easily imagine the radio knob doubling as this control.

Once the software started listening, it gave you a cue to speak your phrase.  It then displayed what it thought it heard, as processed in and returned from the cloud.  

To edit the message, you used the knob to scroll between words.  With a click you selected a word, and yet again received an audio cue giving you the signal to speak the correct word or phrase to replace the selected word with.  

Sending the finished message was as simple as a voice command.  Nuance didn't explicitly say that you would be able to dictate, edit, and send messages on the go in the eventual commercial form, but that seemed to be the implied message here.

Clearly the technology is almost here.  Expect to see this technology pop up on Ford's Sync platform sometime around 2013, if not 2012.  The competition will likely be just a bit behind, deploying solutions around 2014.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Texting and driving...
By vanionBB on 11/10/2011 1:22:01 PM , Rating: 5
quote:
Once it started listening, you said you phrase. It then displayed the results. You could use the knob to scroll between words. You could then click to edit the words, reading the correct word or phrase to replace it with.


Why not install a keyboard on the steering wheel, or better yet, lets use the steering wheel as the knob and the horn to click and edit words! It looks like their ideas are made of fail.

I think at this point it would be safer for a computer AI to do the driving while the chick behind the wheel is "distracted."




RE: Texting and driving...
By Astral Abyss on 11/10/2011 5:47:48 PM , Rating: 2
It appears you have the same concerns I do. Anything that makes texting while driving EASIER yet does nothing to remove the distraction of looking at another device while driving is not solving the problem, it's making it worse.


RE: Texting and driving...
By JasonMick (blog) on 11/10/2011 5:58:44 PM , Rating: 3
quote:
Why not install a keyboard on the steering wheel, or better yet, lets use the steering wheel as the knob and the horn to click and edit words! It looks like their ideas are made of fail.

I think at this point it would be safer for a computer AI to do the driving while the chick behind the wheel is "distracted."

From an informed analyst perspective I have to take you to task slightly on the idea that all input forms are equal.

Studies have generally shown voice input to be minimally distracting. As far as text message dictation, all voice is obviously the optimal system, but it makes editing difficult.

Visual systems that maintain your eyes at a <30 incident angle off the road view (e.g. a high console touch screen) are less distracting than lower positioned devices (e.g. the Blackberry in that chick's lap). This seems common sense when you think of it and studies support it. They think part of the cause of this correlation with decreasing distraction-related accidents as the viewing angle decreases is because when you're viewing at a small off-road angle your peripheral vision can still operate normally and alert you to sudden movement or bright brake lights.

Ultimately the self driving car is the perfect solution, but we're still pretty far from having the AI sophisticated enough to navigate the busy city. And then there's the not insubstantial issue of cost of hardware.

One big barrier is that the quickest path to the self-driving car is inter-vehicle and vehicle-traffic control device communication, but this requires a broad preparatory period of infrastructure deployment. One big problem is that this means that road system modifications are also necessary, e.g. adding antennas to lights, etc. Given the anti-government spending sentiment of the U.S. public, the next several years will not be a good time to pull that out of the hat. And the private sector is unwilling to have the funds or interest to fund that kind of mass infrastructure rollout.

Financial and infrastructure issues considered, as well as the general state of the art, I think we're maybe 15-20 years from commercialization of mild self driving in ordered settings (e.g. highways, well maintained country streets) and at least 30 years from full self driving.

I say that because I've seen what some of the automakers are planning for 5 years from now and it's not self-driving.

You'll likely see bits of self-driving AI (e.g. lane keeping tech) creep in over the next 5 years, followed by progressively more in the decade that follows.

Until then canned text messages are the safest solution and a sophisticated dictation approach using familiar instruments (e.g. your radio knob) is a close second. Let's face it, you could say "just don't text", but if most young professionals get a text from their boss, they will reply. Without safe texting tech, they will simply try to disguise their action (if it's illegal), which typically involves lowering your eyes very far off the road (a very LARGE incident angle). That's by far the biggest danger.

In summary I agree with your second paragraph, disagree with your sarcastic sentiment in the first one. You gotta work with the technology you have today.


novel concepts that just wont work
By tastyratz on 11/11/2011 10:19:21 AM , Rating: 2
for 1, stop calling it "the cloud" people. it fing sever based and it's nothing new. This ambiguous generalization of some revolutionary concept almost all computer systems have been pivoting on since inception is boggling.

now the rest of my point:
"cloud" supported interface system? at what point does massive web traffic seem like a good idea? what about something obvious like... RECEPTION. if I drive through an area with poor coverage can I still change the radio station? no does it fall back to a second level of more basic controls? how agitating will it be to check your signal level to perform basic tasks and have to remember two different ways to interact on the fly.

How about putting an expiration date on your car, what happens when you become "unsupported" in 5 or 10 years? Are you expected to just buy a new car? Today's cars last longer and people drive them longer. You can not reasonably expect their wifi system to support old flavors for long. think about data speeds 5 or 10 years ago and computer speeds.

The concept is flawed, the system needs to be able to independently operate and function fluently.I don't and won't buy this connection dependent crap not for a second. Utilization of my smartphone should be complimentary to the functionality and extend capabilities.




By jharper12 on 11/14/2011 6:02:27 PM , Rating: 2
Your phone uses a .6 watt transceiver, OnStar, which operates off of Verizon's towers, uses a 3 watt transceiver. There's no reason to have a power sipping cell connection in a car... so there's really no reason for you to lose your data connection in the car.

I do agree with your general sentiment though. Nuance should pair up with QNX and an ARM licencee to build a local solution. That, or the automakers should start lobbying to bring back unlimited data connections on smart phones. ;-D.


It works pretty well
By Solandri on 11/10/2011 5:46:20 PM , Rating: 2
I was driving home from a job when my sister called asking me to join her for lunch. I asked her for the address of the restaurant. Then I hung up, held down the search button on my Android phone (activates voice commands), and said "navigate to restaurant_name, city_name".

A few seconds later it popped up the address which I confirmed by touching it (I did have to glance off the road for a split second to do this), and it fired up the GPS navigation app with directions on how to get there. This was a lot safer and quicker than how one normally uses (or is not supposed to use) a GPS. I can see this being a major selling point for an in-car GPS, assuming phone GPSes don't take over. (And before anyone asks, I've done this many times now and it's understood the correct address over 90% of the time.)




"If you can find a PS3 anywhere in North America that's been on shelves for more than five minutes, I'll give you 1,200 bucks for it." -- SCEA President Jack Tretton














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki