Have you seen this video? It’s probably the most impressive thing I’ve seen on the internet all week.
What you’re seeing here is Keyvan Mohajer, founder and CEO of SoundHound Inc., posing queries to his company’s new voice assistant called Hound. Mohajer demonstrates Hound’s ability to listen, handle context, and crunch data at a mind-blowing clip, which happen to be particular weaknesses of Siri, Apple’s voice assistant.
Siri is incredibly important to Apple. It’s now available on all the company’s computers, phones, tablets, music players, and now the Watch. It helps people do whatever they’re doing, whether it’s research, dictation, keeping up with sports and movies, or communicating with others.
But as much as it’s used, data shows Siri isn’t as accurate as Google Now, its prime competitor, when it comes to listening, comprehension, and giving the right answers.
Hound would make Siri considerably better — particularly at listening and conversing in a series of context-sensitive questions. Mind you, I went through Mohajer’s video and asked Siri all the same questions. Surprisingly, Siri gave me many of the same responses from early on in the video, albeit with some extra wait time. But Siri could not handle the context-sensitive dialogue displayed towards the end of the video, like when Mohajer asks Hound to calculate monthly mortgage payments. Siri simply pointed me to a web search.
Addressing Siri’s weaknesses
Building voice assistants is hard. You need good voice recognition technology, natural language understanding, and the ability to read back information in a natural-sounding way.
Siri is a big achievement for Apple, but other voice assistants, like Google Now and Hound, are getting more nuanced by getting better at conversations and dialogue, as well as specific tasks. They’re better at managing the state of conversations, maintaining knowledge between questions to inform future answers, and analysing the past actions of its users to predict and proactively suggest information.
Right now, though, if you ask Siri a question, it won’t remember the previous question. There are a few exceptions, like when you’re changing the time on a reminder or alarm before confirming it, but for the most part, Siri is a glorified question-and-answer robot with a few handy tricks, like the ability to handle reminders and alarms. But it’s not a true digital assistant, not yet.
Adding SoundHound’s expertise in this area could be worthwhile for Siri. Of course, working with SoundHound has other benefits, too: The company’s “Houndify” software could help Apple’s iOS developers voice-enable all their third-party applications for cross-platform functionality, and its “SoundHound” software, which offers song lyrics and a music discovery engine that can even identify songs you’re singing or humming, could help Apple’s various music efforts, namely iTunes and its new streaming music service to be announced Monday.
But Siri is arguably the most important digital service in Apple’s repertoire, since it crosses all product lines, and it’s also the one thing that stands to benefit the most from working with SoundHound. Apple has always been about making computers more approachable, and so a human-sounding voice assistant is central to that idea.
Siri is getting better all the time, but it’s still needs to be a better listener, and Hound clearly has a few legs up in the dialogue and context departments as well. As the voice of every Apple device, Siri needs to be the best it can be — it needs Hound’s special sauce, however it can get it.