Amazon’s generative AI vision for Alexa is appealing, but unproven
Amazon’s long-awaited update to its assistant is almost here. About 18 months after the company first previewed the “next-gen Alexa” built with generative AI, it unveiled Alexa+, and early access will be available starting in March. Alexa+ will exist alongside the older Alexa and will cost $20 a month, unless you have a Prime membership, which will make it free to use. The new assistant will come with all the modern upgrades that its contemporaries like the redesigned Siri or Gemini offer, like more conversational interaction, better contextual understanding and the ability to “summarize complex topics” and “make suggestions based on your interests.” But it does one thing differently, and it’s the way Amazon purports to integrate with third-party apps and the rest of the internet that could set it apart.
At the presentation, vice president of Alexa and Fire TV Daniel Rausch outlined three ways the new assistant can integrate with other services you use. Firstly, it already works with “tens of thousands” of integrations already available, with Uber, Sonos, Samsung and Xbox being a few of the many logos that were displayed when Rausch said this. Presumably, that means Amazon worked with these partners to get their apps to play nicely with Alexa+ through their APIs.
Secondly, for the large swath of the online world that doesn’t have apps or the resources to code an API just for Alexa+, the assistant should be able to scour the internet for their website and navigate it on your behalf. During the presentation, Rausch demonstrated how Alexa+ was able to go to the Thumbtack website to hire a professional to fix his oven. Instead of having to whip out a phone or laptop to click through menus himself, Rausch could just verbally tell Alexa+ what he needed and what times he was available, and the AI did the rest.
This example in particular struck me as very similar to Google’s restaurant-reservation system Duplex, which, since way back in 2018, could call businesses on your behalf to book a table. All you had to do was tell Google what date and times you were considering, how many people were in your party and it would make the call for you, even speaking in a human-sounding voice to the restaurant. The way Alexa+ would click around websites on your behalf seems like Duplex on steroids.
Finally, Alexa+ can work with the “AI agents” on other sites, so you can talk to just one assistant instead of dozens of chatbots. In the presentation, Rausch got Alexa+ to work with the AI music generator Suno to come up with a country-style song about bodega cats.
After the company’s presentation, I spoke with Rausch to get more clarity on how Alexa+ will work with the rest of the online world. For one, I wanted to know for sure if that third method would work with customer service chatbots from companies like Capital One, United Airlines and more. While Rausch said that he had no details to share on specific names or services, he did confirm that “the SDK is all about integrations like that,” saying it’s for people to use Alexa to connect with agents on their behalf to complete tasks.
Rausch is aware that any friction at all in the adoption and setup process can turn people away. To that end, upgrading to Alexa+ should not require any additional sign-ins or authentication. You should be able to have all your connected appliances, security cameras and home routines carry over without any effort.
“The reason that customers love Alexa is it takes away all the complexity,” Rausch said. People don’t need to remember the brand of the WeMo plug they bought, for example, to be able to tell Alexa to turn off a lamp. “We would never take that away,” he said. When you’re adding new services after upgrading to Alexa+, Rausch said you’ll either do so “in the ways that you do it today” or that it’ll get even easier, since “Alexa can walk you through those setups in many more cases.”
According to Rausch, instead of having to “dig around in the Alexa app, you just say ‘Alexa, I want to set up a streaming account with Hulu' or something.” You’ll be presented with a QR code on an Echo device with a screen to facilitate that, and the assistant should guide users along the way.
“We like to say Alexa is an expert, and now an expert on herself.”
Amazon isn't the only company that has made its assistant perform tech support for users. This approach is very similar to how Siri can teach you how to, say, shoot a video in Cinematic mode or create a Genmoji. Where better to get help about a product you’re using than the product itself? Why make me go to a different place for information?
“From a customer experience perspective, customers just want the thing done,” Rausch said. “They don’t have to be responsible or care about any of these things, right?” He believes people just want the plumber or a reservation booked, and don’t want to fuss around with websites and phone calls.
People do want their digital assistant to be helpful and easy to use, but isn’t it equally important that these AI services are accurate and reliable?
When I asked Rausch how Amazon worked around the tendency of generative AI to hallucinate and sometimes spread misinformation, he said “I actually think in the industry, there’s been a mistake of thinking a model is a product.” He also said “LLMs are at the foundation of the architecture, but they’re not the only thing answering the question.”
In other words, Alexa+ is using a combination of Amazon’s knowledge graphs, reliable sources on the internet and partnerships with authoritative outlets. “Other products will simply give you an answer out of the LLM. If you're asking for an authoritative answer, that's not actually sometimes the way to get one. and I think that's what you're highlighting,” he said.
Rausch added that Amazon has “taken great care” with Alexa. “Will it make mistakes? Every piece of software makes mistakes,” he said. “But we’re working hard to ground it in knowledge.”
I’ve yet to try out Alexa+ for myself, and everything we’ve seen so far has been in highly controlled demos, so it’s hard to tell what real world performance will be like. But if the new Alexa is truly able to work with third-party services in a way that’s seamless and effective, it would not only bring Amazon back into the consumer AI race, but could possibly give it an advantage over the likes of Google, Apple and OpenAI. Considering Amazon really brought the idea of a virtual assistant into homes around the world, this could also have an impact that goes beyond the tech-savvy users of today’s AI services.
This article originally appeared on Engadget at https://www.engadget.com/ai/amazons-generative-ai-vision-for-alexa-is-appealing-but-unproven-140014416.html?src=rss
Amazon’s long-awaited update to its assistant is almost here. About 18 months after the company first previewed the “next-gen Alexa” built with generative AI, it unveiled Alexa+, and early access will be available starting in March. Alexa+ will exist alongside the older Alexa and will cost $20 a month, unless…
