Tag:

assistant

Amazon’s AI assistant is smarter, but still struggles.
Product Reviews

Amazon’s AI assistant is smarter, but still struggles.

by admin


This week, Amazon will launch new Echo hardware designed to supercharge Alexa Plus, the AI-powered upgrade to its voice assistant. I’ve been using Alexa Plus for the last few months as part of its Early Access program, and while the new assistant is off to a promising start, it’s still clearly a work in progress.

To fix Alexa, Amazon had to break it apart and rebuild it. The result is a hybrid smart home assistant, personal assistant, and Amazon’s answer to ChatGPT. Right now, in its Beta phase, this new Alexa isn’t doing any of those things as well as I’d hoped.

This is most obvious in the smart home. Controlling my lights, locks, and robot vacuum with natural language rather than precise phrases is a huge improvement, as is not having to say “Alexa” repeatedly and being able to interrupt, um and er, and change my mind mid-thought. But we are still far from the dream of the ambient home that runs on a Star Trek–style “Computer.”

Today, running on what feels like underpowered hardware and with surface-level integrations into my smart home, Alexa Plus often leaves me frustrated. There’s power under that hood, but it feels largely inaccessible. The assistant desperately needs something to make it more compelling — and better hardware could be the answer.

Alexa Plus should make the smart home smarter

Generative AI is supposed to be a watershed moment for the smart home. By cutting through the complexity of programming your home and removing the frustrations of clunky commands, LLMs should make the smart home more accessible. And in many ways, Alexa Plus delivers.

I can now say, “Alexa, dim the lights in here, adjust the thermostat down a few degrees, lock the front door, and turn the upstairs lights off. Oh, and remind me to take the trash out in the morning,” and it all happens. This kind of easy, hands-free convenience is exactly what the smart home has promised for years.

The old Echo Show UI (left), compared to the new Alexa Plus UI, which offers more control and a more intuitive interface. Photo by Jennifer Pattison Tuohy / The Verge and Photo by Chris Welch / The Verge

Managing my devices is also simpler. Instead of scrolling through thumbnail clips in the Ring app, I can ask an Echo Show 21 when the cat was last on the porch and instantly see a full-screen video. The updated UI on the Show 15 and 21 is a big improvement, with larger widgets, customizable layouts, and easier access to smart home controls.

Recently, Alexa and I chatted about the best ways to use my smart home gadgets to their full potential. It suggested possible routines, built the automations, tweaked them based on my feedback, and tested them — all in minutes, with no fiddling in the (still clunky) Alexa app. It even helped me set up a new air purifier and folded it into one of those routines.

But there are issues. Alexa Plus is noticeably slower, with some requests taking up to 15 seconds for a response. While turning on lights or adjusting a thermostat is fast enough (I assume due to using local connections over Matter), waiting for over 10 seconds for the weather or a song to play is tiresome.

Some basic features that used to work reliably now don’t or require new phrasing every time. My struggles to control my Alexa-enabled coffee machine persist, and I can’t get Alexa to consistently turn on my bathroom fan for a set period of time.

The Echo Show 21 can display live feeds from up to four Ring cameras, as well as pull up specific events using voice commands. Photo by Jennifer Pattison Tuohy / The Verge

I used to say, “Turn on the bathroom fan for 15 minutes,” and it did it. Now, Alexa Plus tells me it has to create a routine to do that, and then doesn’t run that routine. Or it says, “Sure,” turns the fan on, but never turns it off. I’ve tried this a dozen times and haven’t had a consistent response yet.

One surreal moment: after weeks of Alexa Plus’ new voice, the old Alexa suddenly surfaced when the system hit a snag. “Sorry, something went wrong,” it said in that stiff, familiar tone. For a second, I wondered — is the old Alexa still in there, trying to get out?

Building the old on top of the new

The old Alexa — the deterministic model built on rigid command-and-control rules — is essentially gone. According to Panos Panay, head of Amazon’s devices and service division, whom I spoke with in February, Alexa Plus runs on an entirely new architecture. One that, based on my testing, feels much more powerful than the old Alexa, but also less reliable.

That’s the paradox of LLMs: they’re excellent at parsing human language, but they’re not designed for consistency. Ask ChatGPT the same question twice, and you’ll get different answers. The unpredictability of LLMs, known as nondeterminism, is a poor fit for smart home control, where reliability and repeatability are crucial. It’s great when you’re brainstorming, but frustrating when you just want your morning coffee.

Amazon’s workaround has been to use its LLM models as a kind of translator. It interprets what you say, then hands off the request to deterministic systems — APIs, device controllers, or local Matter connections.

The unpredictability of LLMs is a poor fit for smart home control, where reliability and repeatability are crucial.

I’ve found this works most of the time, but if the LLM translates a request incorrectly or there’s a gap in the API, it appears that handoff can fail. I assume that’s why my bathroom fan sometimes turns on as requested and why Alexa sometimes insists on creating a routine but then forgets to finish the job.

This is the problem every company with a voice assistant in the smart home is dealing with — merging the old and predictable with the new and exciting. LLMs aren’t designed to be predictable, and what you want when controlling your home is predictability.

Panay says they’ve worked hard to bring predictability to Alexa Plus and to ensure it won’t hallucinate in your smart home. While the former still needs work, so far my smart home has been hallucination-free. There have been no bizarre behaviors such as unlocking doors or cranking up the heat unbidden, or doing something different from what was requested.

However, this tightly controlled structure has resulted in an Alexa Plus that is not the paradigm shift I was hoping for. Of course, it’s still early days, but the promise of LLMs is that they will unlock the potential of technology within our homes — and that hasn’t happened yet.

Alexa Plus hasn’t changed anything for me; it’s just made my smart home (mostly) easier to manage. It still feels like pieces and parts, not a cohesive whole being run by an intelligent machine.

Hardware could hold the key

Many of my frustrations with Alexa Plus are connected to the hardware, and changes here could make a big difference.

The current Show devices are the flagship Alexa Plus interfaces, specifically the Show 21 and 15. But the interplay between voice and screen is still lacking; the hardware remains voice-first.

For example, I’ll ask Alexa to show me the recipe I was just using, and instead, it will read out the directions. With hardware that synchronizes voice and visuals seamlessly, Alexa Plus would be very compelling. (Also, the Shows are the worst Echo devices at hearing commands, and Amazon really needs to fix that.)

Andy Jassy has promised us “beautiful” new hardware for Alexa. As the first products fully designed under Panos Panay, who told me he believes in screens, we have some idea of what’s coming. But ultimately, it will be about how well the hardware and software work together. The devices revealed this week will be Alexa Plus’s moment to prove it’s more than just potential.

Photos by Jennifer Pattison Tuohy / The Verge

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.

  • Close

    Jennifer Pattison Tuohy

    Senior Reviewer, Smart Home

    Posts from this author will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All by Jennifer Pattison Tuohy

  • AmazonClose

    Amazon

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Amazon

  • Amazon AlexaClose

    Amazon Alexa

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Amazon Alexa

  • AnalysisClose

    Analysis

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Analysis

  • ReportClose

    Report

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Report

  • ReviewsClose

    Reviews

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Reviews

  • Smart Home ReviewsClose

    Smart Home Reviews

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Smart Home Reviews

  • TechClose

    Tech

    Posts from this topic will be added to your daily email digest and your homepage feed.

    PlusFollow

    See All Tech



Source link

0 comments
0 FacebookTwitterPinterestEmail
A man wearing a suit kneels on the ground as a dog wearing a harness rushes toward the screen. Both characters are rendered in stylish silhouette.
Gaming Gear

4 years after launch, card-based brawler Fights in Tight Spaces gets an expansion that adds a canine companion and an ‘all new assistant system’

by admin



It’s been four years since Fights in Tight Spaces grabbed turn-based tactics by the collar and beat it about the head with a stack of collectible cards, so the fact that it’s getting an expansion this far out was a surprise in and of itself. That this DLC primarily adds a dog to help you batter baddies in cramped conditions is, if anything, a pleasant bonus.

For the uninitiated, Fights in Tight Spaces takes the tactical layer of games like X-COM and replaces the aliens with a heavy dose of John Wick. Playing as suave silhouette Agent 11, each level places you in an isometric, turn-based action scenario filled with monochromatic goons, then deals you a handful of martial-arts themed playing cards. You need to play the cards to take out your opponents in the most efficient way possible. It’s a straightforward premise, but one sleekly presented and with more depth than you might expect.

The DLC, named K9 Division, adds a helpful hound into the game’s card-based brawls. According to developer Ground Shatter, this involves more than just an extra character to play with. While Agent K can fight on their own terms using an array of dog specific moves such as biting enemies, intimidating them by barking, or distracting them to give Agent 11 more breathing room, the DLC also layers in an “all-new assistant system” that allows both agents to team-up for flashy combinations.


Related articles

Fights in Tight Spaces: K9 Division Trailer – YouTube

Watch On

You can experiment with this system via not one but two canine starter decks, each offering a slightly different hound-assisted fighting style through a total of 50 new cards. The DLC also brings a “remixed story mode” letting you experience the game afresh with Agent K’s assistance, as well as new enemies for you to batter/bite. Oh, and in case you’re wondering, yes, there is a card for petting the dog.

If you’re wondering why Fights in Tight Spaces is getting this DLC four years after launch, it seems K9 Division ports over some of the companion systems seen in Ground Shatter’s sequel Knights in Tight Spaces, which released earlier this year. While a decent follow-up, Knights doesn’t appear to have gone down as well as Fights did. Knights’ ‘Mostly Positive’ Steam reviews claim that it’s a richer experience all-told, but one that takes a while to show its strengths, lacking the more immediate appeal of its predecessor.

In any case, Fights in Tight Spaces: K9 Division is available now, with an RRP of $10 (£8.50) Ground Shatter is currently running a 10% launch discount, bringing that price down to $9 (£7.65) until October 2.



Source link

0 comments
0 FacebookTwitterPinterestEmail
DAAPrivacyRightIcon
Product Reviews

Facebook adds an AI assistant to its dating app

by admin


Facebook Dating has added two new AI tools, because clearly a large language model is what the search for love and companionship has been missing all this time. The social media platform introduced a chatbot called dating assistant that can help find prospective dates based on a user’s interests. In the blog post announcing the features, the example Meta provided was “Find me a Brooklyn girl in tech.” The chatbot can also “provide dating ideas or help you level up your profile.” Dating assistant will start a gradual rollout to the Matches tab for users in the US and Canada. And surely everyone will use it in a mature, responsible, not-at-all-creepy fashion.

The other AI addition is Meet Cute, which uses a “personalized matching algorithm” to deliver a surprise candidate that it determines you might like. There’s no explanation in the blog post about how Meta’s algorithm will be assessing potential dates. If you don’t want to see who Meta’s AI thinks would be a compatible match each week, you can opt out of Meet Cute at any time. Both these features are aimed at combatting “swipe fatigue,” so if you’re 1) using Facebook, 2) using Facebook Dating, and 3) are really that tired of swiping, maybe this is the solution you need.



Source link

0 comments
0 FacebookTwitterPinterestEmail
Smart Glasses, Buggy Voice Assistant
Product Reviews

Smart Glasses, Buggy Voice Assistant

by admin


Smart glasses are an exciting idea right now. In theory, they’re a new gadget that does lots of the stuff that our phones do, but in an always-there form factor. They can take pictures, make calls, translate menus, and—if the tech and the investment get there—they might slap a screen right onto eyeballs for notifications, navigation, and maybe even augmented reality à la Pokémon Go.

I say “in theory” because just because smart glasses can do all of those things on paper doesn’t mean they can do them well, and if they can’t do them well… they may as well not do them at all. We’re still in the early stages of the grand ascension of smart glasses as a device category, but a brave few are venturing to do it all right now, and one of those few (at least in the U.S.) is a company based in China called Rokid (pronounced rock-id).

See Rokid Glasses at Amazon

I got a chance to try Rokid’s plainly named Rokid Glasses, and while there was a lot that intrigued me, I can say for certain that the kinks are still being worked out. One thing that these smart glasses have that big-time entrants from the likes of Meta and its Ray-Bans don’t have is a screen. That screen is a very simple dual-micro LED display that only shows things in a very Matrix-style green. I got to use the Rokid Glasses for 15 minutes and was surprised at how sharp the screen was, even if the display functions were fairly basic. And look, you can see the screen from outside the glasses. And that’s good because sharpness is crucial for some of the things that make the Rokid Glasses unique.

© James Pero / Gizmodo

One of those distinct capabilities is a teleprompter feature that displays a presentation in front of your eyes, so you can read along and not sound like a total moron during your big keynote. A thing that I found very cool is the fact that the Rokid Glasses actually use the onboard microphone to listen to your words and scroll the prompter in stride with the words. Even in a crowded room with lots of noise, the feature worked smoothly, which is no small feat.

Another screen-centric feature I got to try was translation, which—though my conversation was fairly brief—seemed to work better than you’d expect. My demo companion spoke to me in Mandarin Chinese, and the Rokid Glasses were able to translate his speech in small snippets and slap them onto the screen. Again, the microphone did all of this in a noisy room, which was legitimately impressive. The microphones on Rokid’s smart glasses work so well that I’m pretty sure you could use them for spying—it picked up bits of conversations across the room that I wasn’t even able to make out with my own ears. Cool! Also scary!

© James Pero / Gizmodo

To use all of this stuff, it’s best to couple the Rokid Glasses with an app (Android-only right now) where everything you’re doing is displayed. As sharp as the screen is, it’s also quite small, and words are pushed off when new information arrives at a fairly quick pace. If you need to see something, it’s best to have the app ready, lest you ask someone to repeat themselves multiple times. And in case you’re wondering, you can control the display from the smart glasses by swiping the right arm and using a tap to select things like settings, translation, and other stuff, but it’s not exactly the smoothest experience. That’s why voice assistants—for Rokid and any company making smart glasses right now—are also critical. That brings me to a not-so-bright spot.

The Rokid Glasses voice assistant, which is supposed to activate with the wake phrase “Hi, Rokid,” was basically broken. No matter how many times I screamed “Hi, Rokid” into the smart glasses, it wouldn’t answer my calls. Others around me were also having the same issue, which is not great from a UI perspective. The interesting thing is that when a native Mandarin-speaking represenative said  the phrase, it seemed to work every time. American English-speaking people, not so much. I thought maybe it was the loud, crowded room at first, but after noticing that strange quirk, I think it may be a problem with how the voice assistant is trained. I can’t say for sure without testing the Rokid Glasses more thoroughly, but it’s definitely a concern for anyone buying a pair in the U.S.

© James Pero / Gizmodo Just one lone sensor on these glasses.

Like Meta’s Ray-Bans, the Rokid Glasses can also use AI for computer vision-based tasks like asking your glasses to read a menu in a different language using the built-in camera. I wasn’t able to launch that task myself, given the aforementioned voice assistant issues, but when a Rokid representative asked the smart glasses to translate a menu in Finnish, it did so (at least I think) fairly well, displaying the translated Finnish words in the Rokid app. Again, I’d need to test this feature out more thoroughly in a better environment to verify the translation separately and determine how well (or terribly) it actually works on a consistent basis.

As long as we’re talking about computer vision, I was pleasantly surprised with the camera, which is a 12-megapixel sensor from Sony. Just one sensor, not two, though. I would say it’s on par with Meta’s Ray-Bans, but I didn’t get to test video recording out in my demo. I wouldn’t try and use the Rokid Glasses to win any photo contests, but then again, I wouldn’t do that with any pair of smart glasses.

I won’t know until I get to try Rokid Glasses for a longer period, but I get the sense that they’re smart glasses with some peaks and valleys. Translation could be impressive, as could computer vision, and they’re incredibly light (as light as Ray-Bans), but if there isn’t a functional, English-ready voice assistant to tie it all together, that’d be a big problem for anyone in the U.S. who wants to buy a pair. That’s potentially not the best news for smart glasses enthusiasts in America, but I assume Mark Zuckerberg would welcome that quirk with open arms.

See Rokid Glasses at Amazon



Source link

0 comments
0 FacebookTwitterPinterestEmail
Google’s Gemini Live AI assistant will show you what it’s talking about
Gaming Gear

Google’s Gemini Live AI assistant will show you what it’s talking about

by admin


Google is bringing a bundle of new features to Gemini Live, its AI assistant that you can have real-time conversations with. Next week, Gemini Live will be able to highlight things directly on your screen while sharing your camera, making it easier for the AI assistant to point out a specific item.

If you’re trying to find the right tool for a project, for example, you can point your smartphone’s camera at a collection of tools, and Gemini Live will highlight the correct one on your screen. This feature will be available on the newly announced Pixel 10 devices when they launch on August 28th. Google will begin rolling out visual guidance to other Android devices at the same time before expanding to iOS “in the coming weeks.”

Google is also launching new integrations that will soon allow Gemini Live to interact with more apps, including Messages, Phone, and Clock. Say you’re in the middle of a conversation with Gemini about directions to your destination, but you realize you’re running late. Google says you’ll be able to interrupt the chatbot with something like: “This route looks good. Now, send a message to Alex that I’m running about 10 minutes late.” From there, Google can draft a text to your friend for you.

Lastly, Google is launching an updated audio model for Gemini Live that the company says will “dramatically improve” how the chatbot “uses the key elements of human speech, like intonation, rhythm and pitch.” Soon, Gemini will change its tone based on what you’re speaking about, such as using a calmer voice if you’re asking about a stressful topic.

You’ll also be able to change how fast — or slow — Gemini talks, which sounds a bit similar to how users can now tweak the style of ChatGPT’s voice mode. And, if you ask Gemini for a dramatic retelling of a story from the perspective of a particular character or historical figure, the chatbot may adopt an accent for a “rich, engaging narrative.”



Source link

0 comments
0 FacebookTwitterPinterestEmail
Close