Piper Jaffray analyst Gene Munster, who has regularly assessed Siri's accuracy in terms of correctly interpreting and answering queries, has issued the latest version of his Siri report card, noting that Siri has continued to improve under iOS 7, particularly in terms of being able to properly interpret questions being asked.
The primary improvement between Siri in iOS 6 and iOS 7 came from its performance in noisier areas, where Siri was able to translate better through a noisy environment (94% vs. 83% correct in Aug iOS 6). Based on our testing, iOS 7 Siri's net ability to answer questions (factors in questions not correctly understood) improved to 79% from 77% in August with iOS 6.
The improvements in Siri's understanding come alongside but separate from Apple's continued efforts to reduce its reliance on Google, as evidenced by Siri's shift to emphasize Bing, Wikipedia, and WolframAlpha results. In Munster's most recent testing, only 4 percent of Siri's answers relied on Google, compared to 17 percent in August testing and 27 percent in December 2012.
Click for larger
Munster also compares Siri to Google Now, finding that the two services are now essentially tied in effectiveness with a "C+" score after Siri held a substantial lead in the year-ago testing. In the latest testing, Google Now held small leads in content areas such as commerce and information, while Siri performed significantly better for operating system controls.
There is no way these charts are accurate. Even when I speak as clearly as possible and enunciate every syllable, Siri understands me about 40% of the time.
I get 90% accuracy, I don't know what's going on with you.
There is no way these charts are accurate. Even when I speak as clearly as possible and enunciate every syllable, Siri understands me about 40% of the time.
The one thing I will greatly defend Apple and Siri for is trying to understand people's voices. People either have a strong accent, speak with a lisp, speak too fast, speak too slow, speak with a low register voice, have an annoying voice, mumble their words, garble their words and most of all they try and use Siri in a noisy atmosphere along with any of the above.
No matter what, it's still an electronic device and people just have to learn to speak properly based on the source or just don't use as you've appeared to do. You say you have no need for voice control, it's not just about voice control. Many times I need to look up sports scores, it's much faster and easier just to ask Siri the game stats than typing them in or opening an app and searching.
Never defend a technology! Technology is meant to develop to serve the requirements/limitations of existing human behaviour; human behaviour shouldn't be expected to adapt to meet the requirements/limitations of technology.
I have had the opposite experience. After seeing steady improvement in Siri, it has recently gotten worse and become almost unusable for me.
Specifically, I use Siri a lot to play music in the car. I will say "Play Song X" or "Play Album Y". Siri used to be pretty good at hearing the name of the song and finding it in my music library. Now it frequently hear the names incorrectly. Even when it does hear the correct name, it will play a song with a completely different name or simply say "Sorry, I couldn't find that in your music". What do you mean you couldn't find David Bowie in my music??? It's right there!
This regression is quite infuriating. Also, Siri still often fails to respond ("Sorry, there's something wrong") or take 20+ seconds to respond with the wrong answer. I believe we need to give all of Apple's web services time to gather data and make iterative improvements, but I have very little tolerance when they actually get worse over time.
I don't know, that net percentage increase (4% better) year over year seems kind of low to me. Am I the only one who feels that way? Not my field of study, so I'm not entirely sure if this is supposed to be impressive.
Apple has announced it will be holding a special event on Tuesday, May 7 at 7 a.m. Pacific Time (10 a.m. Eastern Time), with a live stream to be available on Apple.com and on YouTube as usual. The event invitation has a tagline of "Let Loose" and shows an artistic render of an Apple Pencil, suggesting that iPads will be a focus of the event. Subscribe to the MacRumors YouTube channel for more ...
Apple has dropped the number of Vision Pro units that it plans to ship in 2024, going from an expected 700 to 800k units to just 400k to 450k units, according to Apple analyst Ming-Chi Kuo. Orders have been scaled back before the Vision Pro has launched in markets outside of the United States, which Kuo says is a sign that demand in the U.S. has "fallen sharply beyond expectations." As a...
Apple is finally planning a Calculator app for the iPad, over 14 years after launching the device, according to a source familiar with the matter. iPadOS 18 will include a built-in Calculator app for all iPad models that are compatible with the software update, which is expected to be unveiled during the opening keynote of Apple's annual developers conference WWDC on June 10. AppleInsider...
The upcoming iOS 17.5 update for the iPhone includes only a few new user-facing features, but hidden code changes reveal some additional possibilities. Below, we have recapped everything new in the iOS 17.5 and iPadOS 17.5 beta so far. Web Distribution Starting with the second beta of iOS 17.5, eligible developers are able to distribute their iOS apps to iPhone users located in the EU...
Wednesday April 24, 2024 3:39 pm PDT by Juli Clover
Apple today released several open source large language models (LLMs) that are designed to run on-device rather than through cloud servers. Called OpenELM (Open-source Efficient Language Models), the LLMs are available on the Hugging Face Hub, a community for sharing AI code. As outlined in a white paper [PDF], there are eight total OpenELM models, four of which were pre-trained using the...
Top Rated Comments
I get 90% accuracy, I don't know what's going on with you.
Never defend a technology! Technology is meant to develop to serve the requirements/limitations of existing human behaviour; human behaviour shouldn't be expected to adapt to meet the requirements/limitations of technology.
Specifically, I use Siri a lot to play music in the car. I will say "Play Song X" or "Play Album Y". Siri used to be pretty good at hearing the name of the song and finding it in my music library. Now it frequently hear the names incorrectly. Even when it does hear the correct name, it will play a song with a completely different name or simply say "Sorry, I couldn't find that in your music". What do you mean you couldn't find David Bowie in my music??? It's right there!
This regression is quite infuriating. Also, Siri still often fails to respond ("Sorry, there's something wrong") or take 20+ seconds to respond with the wrong answer. I believe we need to give all of Apple's web services time to gather data and make iterative improvements, but I have very little tolerance when they actually get worse over time.