I’m not the only one asking why Siri is so bad.
I thought Siri would be better on the HomePod, with AC power and all those microphones. It isn’t. And I want to know why.
I get that my $349 HomePod makes a better music speaker than my $99.99 Amazon Echo or my $129 Google Home. What I don’t get is why Apple hasn’t caught up to Amazon or Google in terms of answering simple, direct questions. See John Gruber’s much re-tweeted How many quarts in a gallon?
Apple had a head start. It has more cash than Scrooge McDuck. It’s dedicated, as Tim Cook never tires of reminding us, to making great products that improve people’s lives. Where did it go wrong with Siri?
Here are the theories I’ve seen so far. The first came from clicking “I’m feeling lucky” on the Google Search above.
- Sean O’Kane, The Verge: Apple has lagged behind in building out its artificial intelligence efforts. Apple’s tendency for secrecy has reportedly scared away some of the best minds in AI out of fear that there’s no chance for [voice] recognition in Cupertino.
- Tripp Mickle, Wall Street Journal: Progress has been slowed by a failure to set ambitious goals, shifting strategies and a culture that prioritizes user privacy—making it difficult to personalize and improve the product. The project also has suffered from the departures of key team members, some of whom went to competitors.
- Walt Mossberg, The Verge: There are three reasons for my doubts: first, Apple’s history with cloud-based services in general has been weak and inconsistent. Second, Apple has done shockingly little to capitalize on its lead with Siri. And third, Apple’s steadfast devotion to privacy and lack of a search service or social network means it doesn’t have the range and volume of data its competitors hope to use to power personalized, actionable AI capabilities.
- Eddy Cue, Apple SVP whose portfolio until last Fall includes Siri: “I don’t think anyone does an A+ on conversation. It’s a challenging problem and there’s a lot of work to be done in that area. It will get a lot better and needs to get a lot better. [From WSJ interview.]
My take: The stories of leadership dysfunction in Mickle’s piece may be reason enough, but I wonder if there might be a deeper, structural problem, a legacy issue, maybe, dating back to the original SRI International implementation? I’m open to ideas.
UPDATE: Speaking of deep legacy issues, a friend-of-the-blog at iPad Insight forwarded this link:
- Andrew Tarantola, Engadget: Fast forward to 2014. Apple is at the end of its rope with Siri’s listening and comprehension issues. The company realizes that minor tweaks to Siri’s processes can’t fix its underlying problems and a full reboot is required. So that’s exactly what they did. The original Siri relied on hidden Markov models — a statistical tool used to model time series data (essentially reconstructing the sequence of states in a system based only on the output data) — to recognize temporal patterns in handwriting and speech recognition. The company replaced and supplemented these models with a variety of machine learning techniques including Deep Neural Networks and “long short-term memory networks” (LSTMNs). These neural networks are effectively more generalized versions of the Markov model. However, because they posses memory and can track context — as opposed to simply learning patterns as Markov models do — they’re better equipped to understand nuances like grammar and punctuation to return a result closer to what the user really intended.
My question: What models are Amazon and Google using?