4 ms·
The author brings up an interesting example where Google Now provides him a tool to automatically identify and describe a TV show he is watching at night. When
by scottfr 13y ago
The author brings up an interesting example where Google Now provides him a tool to automatically identify and describe a TV show he is watching at night. When he later tries to demo this feature to his friend during the day, it simply does not appear.
The author takes this as an example of Google Now being very smart and learning from his behavior. It knows he watches TV at night so it provides him the identification feature during that time period. It doesn't provide it to him during the day, because it knows he isn't usually watching TV during the day (at least this is the author's assumption).
I take away a different message from this example and it is part of the reason I am very skeptical of software like Siri or Google Now.
Fundamentally I believe the one of the most important features of a software application is that it is predictable. When I do 'X', I expect the program to do 'Y' always. If it can't do 'Y' because of certain situations, I should get an explicit, informative error message.
The author's TV example is an illustration of deep unpredictability. It is quite possible that he is right and there are powerful learning algorithms driving the Google Now's behavior in this case. But all I see is unpredictability.
Yes, I want to customize and shape software to my own identity and patterns. But unless I control that process, I am likely to become very frustrated, very fast as a software application changes under my feet.
- deong 13y agoThis is definitely a big issue with Google Now. It's not that it's unpredictable though, at least in my opinion. It's that it's unpredictable combined with no way to deal with the consequences of that. It does what it thinks it should, and you have no knobs to turn. If it fails, it just fails. The first year of Google Now, it would often do exactly that. It popped up the card that it claims will appear at the end of the month with information about how far you've walked or biked maybe twice? The rest of the time, it just didn't appear, and there's no way to call it up manually. I noticed in KitKat that it now offers to let you tell it whether you typically walk, bike, drive, etc. to particular destinations. It seems like in a number of areas, they've chosen to add some manual control to the automated process. I think that's going to be good enough. I'm fine with software that tries to intelligently guess my needs as long as it provides me with a way to override incorrect guesses.
- minor_nitwit 13y agoI think the watching TV thing has to do with a TV connected to your WIFI network. I've experienced what you're talking about though with the flight cards. I'm anxious so I want to know that my flight information is there and ready before I get to the airport, but Google only shows it to me 'when I need it.'
- xur17 13y agoI just recently got a card that entitled 'Watch now' that has a list of shows to watch, and has a 'Listen to TV' button in the top corner. It appears that my phone can listen to the tv, and figure out what I am watching. http://www.youtube.com/watch?v=04dpHTxqI6U http://www.youtube.com/watch?v=04dpHTxqI6U - apparently this has been around for a while.
- darkmighty 13y agoI agree! This kind of learning system will never be able to hit 100% accuracy, and even if it hits 95%, sometimes the 5% it misses are unusually important. It has to have a way to deal promptly with unexpected circumstances. I don't think this requires manual user customization though. I think it's quite reasonable to expect the software itself to probe you for further info on the fly in case of failure.
- Houshalter 13y agoRemoving features unpredictably would be annoying. But merely suggesting a certain feature based on it's predictions would be helpful and useful I think. The feature should still be able to be found in the menus or through a search.
- bilalq 13y agoWhat the author is described is simply Google Now's contextual awareness that a feature could be useful in a given situation. Even if the Listen to TV card doesn't appear in Google Now, you can simply say "Ok Google, listen to my tv," and it will start listening.
- andrewflnr 13y agoIs there a reference for all the phrases it recognizes?
- nirvanatikku 13y agoThe best list I've seen to date: http://trendblog.net/list-of-google-now-voice-commands-infographic/ http://trendblog.net/list-of-google-now-voice-commands-infog...
- jayd16 13y agoYou're looking at the glass half empty aren't you? Instead of never getting the right answer, (in this case) you'll now get it at night. How is that not an improvement?
- DerpDerpDerp 13y agoIf we suppose that it's not going to get /every/ thing wrong or right, but about an even mix between them, then I'd rather have the distribution of half always working and half always not working as opposed to all of them working half of the time.
- lkozma 13y agoA similar point was described here, using Wolfram Alpha as a running example: http://unqualified-reservations.blogspot.de/2009/07/wolfram-alpha-and-hubristic-user.html http://unqualified-reservations.blogspot.de/2009/07/wolfram-...
- ronaldx 13y agoAgreed - if I had asked for the weather in Sao Paulo, and it gave me the weather for Sao Paulo on Sunday night, I would be frustrated rather than impressed. The app is making the implicit assumption that I didn't know what I really wanted, and correcting me. Rather than desirable, I find this creepy, incorrect behaviour and potentially offensive.
- puzzlingcaptcha 13y agoWhile I agree with you on all parts, a cynical side of me tells me that assuming that a majority of your users doesn't really know what they want (or less condescendingly: make assumptions they are not fully aware of) is a viable strategy.
- TheLoneWolfling 13y agoSo then setup the service to suggest the alternative interpretation.