CNET News Video
What Alexa's laughing means for voice recognition's futureA few laughing Amazon Echo and Dot speakers give us insight into a bigger story: voice recognition is still very much in its nascency. We still have a ways to go before it can breach the point of understanding tone and context.
So, last week was weird. There was reports of Amazon's Alexa kinda laughing by itself. Randomly. Randomly. What was that about, A? And B, what does this kind of teach us about the entire digital voice assistant And the technology and where we're headed with that. Yeah, so just to catch people off on this. Alexa, we're just starting to like Cackle, or people thought for no reason. Amazon came out afterwards and said, if you say, Alexa, laugh [MUSIC] That was it, and it was in rare circumstances where it thought it heard you say Alexa laugh. Even though maybe you didn't, It was just- it would just start laughing. It didn't like a false positive. Just unprompted it would just unprompted say Alexa laugh, so Amazon changed that, where now you have to say Alexa can you laugh, and then she'll say yeah I'll laugh. Here's my laugh. Whatever. Is it now they're going to make it preface the laugh with actual English? Right. Because by itself, I mean, yeah, we laugh at it. It was kinda creepy and kinda funny. By itself it was sort of like, my God, something's laughing, this is- Right, so the engineers were probably looking at this, and the writers. Amazon has thousands of people working on Alexa. They were thinking, how do we humanize Alexa more thoroughly? Well what do people do all the time that's actually fun and interesting and quirky? Well, people laugh. Yeah. What would be more interesting than trying to get Alexa to cry, which by the way, I should try to do. I've never asked Alexa to cry before. Cry for me. It's pretty strange isn't it? It's bazarre. It obviously blew up in their face. They were trying to do this humanizing thing with Alexa. And this is the kernal year is that this really conveys how complicated it is to try to get a machine to interact with humans and to create human conversation. And that's, people tend to forget. That Alexia is really only about three years old. Yeah. And we have a ways to go for these machines to actually fully realize and fully understand what it's like to communicate with people. So, the story that happened last week, it was a one off. It's funny. Nobody got hurt. Nobody got hurt. It's funny. It's creepy. But I think it speaks to a much bigger degree of what is going on with the [UNKNOWN] of voice recognition and where it's sort of heading. I think people look at voice recognition and they think it's the future and we have this amazing technology from the future. But in reality, we're sort of just in the [UNKNOWN] of it THis. Yeah, absolutely. The thing that's fascinating to me about writing about this stuff is just thinking about this conversation. Yeah. Regular conversation you have with your mom where you're not even paying attention. The level of complexity and the nuances of the conversation, the dialect that you use, the words you use. Right. The tone. The tone, all that stuff is so difficult to actually train a machine on. They're just two people regularly talking. It's way more complicated than you could ever actually realize. And cramming that into code and shoving into a machine. It could take years and years for it to actually work. And that's why it's this great scientific effort. Yeah. That a lot of these major companies are trying to do right now. Now. All right. Go a little more into that because right now, Alexa and Google Home, and all these assistance, they can really only detect voice like actuall voice matching. They don't understand any sort of sarcasm or obviously, any sort of tone or the pitch in which you're talking. I mean, maybe surface level sort of stuff. But what does that mean for the challenges ahead, and what are they trying to do ultimately? Because right now it's sort of like okay, I could buy more diapers using my voice. Mm-hm. But where are we really headed with this sort of stuff? I haven't seen a lot of this out of Amazon or Google but for someone really small, some smaller companies. They really are trying to add emotional recognition and emotional understanding into voice assistance. Because when you say something, whether you say shut up, or I can't deal with this, or that's really funny, your tone has so much to do with it. And for a machine to fully understand you, fully has to understand your tone. Again, to layer that on top of everything else is gonna take a long time. Yeah, we got a way to go. Yeah. All right, cool man.