I heard Alexa speak like a trusted friend and I don’t know how to feel
It’s a future we’ve imagined for a long time, but when you’re faced with an Alexa that doesn’t need a password to activate it and can hold a conversation as well as your colleague, friend or a simple acquaintance, it’s shocking.
True voice assistance for conversations is about to become a reality. Not today, but in the very near future thanks, of course, to generative AI and large language models (LLM). Amazon showed us on Wednesday (September 20) a new experimental Alexa that goes miles beyond its first stilted response days and easily surpasses the Alexa of 2023.
Admittedly, my experience with this powerful new Alexa was limited to what I could see and hear in both a canned video and a live demo that outgoing Amazon VP of Devices Dame Limp told us showed.
Even so, this isn’t the Alexa you or I know. He speaks clearly and without that jerky manner that we expect and grudgingly accept. He knows the context and has a memory.
To activate the new Alexa, Dave Limp started with “Alexa, let’s chat,” and from that point on, Limp had an open conversation with Alexa where he didn’t need to repeat “Alexa.”
Alexa responded, “Okay, how can I help you”?
Limp sort of redirected himself and said, “How are you?”
It’s the kind of response that might normally throw Alexa off balance, sending the voice assistant into a dead end of confusion.
This time, however, the experimental Alexa immediately responded: “I’m fine, I’m very happy that I can help so many people every day by chatting with them.”
The conversation continued like this for a few more moments, with Limp asking him about his favorite football team and how the quarterback performed in the last game. It was an unexceptional exchange, a normal conversation with an absolutely not normal interlocutor named Alexa.
Alexa’s voice doesn’t sound entirely human, but closer than ever. Think about the movie Her but without the sexual innuendo. This ten-year-old film, released a year before Alexa, tells the disturbing story of a man played by Joaquin Phoenix who falls in love with an AI voiced by Scarlett Johansson. It’s a disturbing romance that probably paints too bleak a picture of a future that now more closely resembles our reality.
To be clear, even this experimental Alexa isn’t quite at that level yet.
However, since Limp did not need to repeat the command word, the conversation continued as if Alexa were almost human. Improved text-to-speech should radically change the quality of Alexa’s voice. Ultimately, I bet this will fool a lot of people in blind audio tests.
Humanity comes not only from the sound of Alexa’s voice but also from emotion. Alexa now seems connected to the present moment.
Alexa’s voice doesn’t sound entirely human, but closer than ever. Think of the movie Her but without the sexual innuendo.
It wasn’t a perfect demo. Limp had to repeat himself several times for Alexa to respond, which is what you’d expect from experimental technology.
Alexa’s more personality-filled voice and contextual awareness are only part of this coming transformation. Alexa will eventually recognize you when you approach an Echo Show, then pick up a conversation where you left off.
Yes, I know, that sounds excellent and maybe a little scary. Yes, for this to work you will need to opt for visual identification so that the Echo Show camera can see you coming and access your last conversation data.
I actually like this idea because it’s not much different than when my wife starts talking to me about something we were talking about yesterday. His reasonable expectation is that I remember and can continue the conversation. Granted, I sometimes fail in this regard, but why should a computer fail in this way?
The scenario is that you walk up to the Echo and start talking. If you say “What time is the game?” » Alexa will remember the team you like and that it’s probably related to today’s game and not next week’s. The conversation could flow from there.
We don’t know much about how Amazon formed its LLM, although it did mention “responsible AI.” Let’s just hope it’s not just lip service and that Amazon trained the AI on various voices and not on snippets of recorded customer statements.
We probably won’t see this new Alexa until 2024, which will not coincidentally mark the 10th anniversary of Alexa and the Echo. What better time to release an all-new Echo and this revolutionary Alexa voice assistant?
I have a feeling this is the update that warns Siri and Google Assistant. Google may be further along in this area, but Apple is well behind in making Siri truly familiar. If the next Alexa is as good as I hope, Apple may need to subject Siri to emergency brain surgery.
Even if at first glance this LLM-powered Alexa doesn’t live up to today’s promise, I have no doubt that it will soon. Of course, our enthusiasm for generative AI and LLM is mixed with a generous dose of fear about what the future might bring, but it’s just another signal that the AI train is coming with great strides and that it will never stop.