Group items tagged

"hen you sign up to use Alexa, you agree to be responsible for any purchases made on the device by you, your resident parrot, a mischievous friend or relative, or an unwelcome burglar. It doesn't matter whether you intended the purchase or not."

"Nadella [le patron de microsoft] considère qu'à long terme, le choix de l'assistant vocal n'aura plus d'importance. Il deviendrait une simple porte d'entrée pour lancer des logiciels de toutes sortes."

"A voice-controlled virtual assistant-Siri, Alexa, Cortana, or Google Home-is only as good as the data that powers it. Training these programs to understand what you are saying requires a whole lot of real-world examples of human speech." "Developers generally need access to hundreds or thousands of hours of audio, says Alexander Rudnicky, a research professor at Carnegie Mellon University" "All of that is privately held, and generally unavailable to academics, researchers, or would-be competitors. That's why Mozilla decided to launch its Common Voice project." "Common Voice invites anyone with an internet connection and a microphone to submit brief recordings of themselves reading particular sentences, all through a couple of clicks or taps on a web browser. " "Mozilla aims to release a version of the dataset later this year and is hoping to have 10,000 hours of audio by that time, the quantity it estimates is enough to train a modern, production-quality system. "

Synthèse du colloque Vox Machine (1/2)

"Duplex - Google's artificially intelligent chat agent that can arrange appointments over the phone - has expanded from a "set of trusted tester users" earlier this year to a "small group" of Google Pixel phone owners, who can now use Duplex via the Google Assistant to secure restaurant reservations in "select cities." [...] Notice that Duplex begins the conversation by saying "Hi, I'm calling to make a reservation for a client. I'm calling from Google, so the call may be recorded." Google received a ton of criticism after its initial Duplex demo in May - many were not amused that Google Assistant mimicked a human so well. In June, the company promised that Google Assistant using Duplex would first introduce itself."

"Google started an unusual artificial intelligence experiment this month. If you instruct its Siri-style virtual assistant to "talk like a Legend," it will speak in a simulacrum of the smooth sound of Grammy-winning crooner John Legend. The singer helped demonstrate a promising, but contentious, use case for AI. Software that can impersonate people's voices can make computers more fun to talk with, but in the wrong hands might be used to make so-called "deepfakes" intended to deceive. How good is voice cloning technology now? Google's project provides a snapshot."

How Voice Computing Will Transform the Way We Live, Work, and Think By James Vlahos "When his father was diagnosed with cancer, Vlahos decided to program a chatbot - which he named "Dadbot" - with his father's life stories, jokes and some of his mannerisms and personality quirks. Vlahos begins by gathering source material, interviewing his father about his childhood, family memories, school, work and favorite sports teams. He records the interviews, and their transcriptions fill a series of giant binders. As his father's health fails, the Dadbot gets better and better. Vlahos knows the Dadbot "will no doubt be a paltry low-resolution representation of the flesh-and-blood man," but he's still reasonably certain he can teach it to mimic his father's charm and humor. Most of "Talk to Me" is only tangentially related to the making of Dadbot; it's actually a deep history of the technology that made such a robot father possible, how we came to be on the cusp of truly conversational, natural-language technology that can hear and understand us, and talk back. "

"In particular, Google is making a big push to shift machine learning operations from the cloud onto peoples' mobile devices. This should allow ML-powered applications to be faster, more private, and able to operate offline."

"Female owners of smart speakers are more likely than men to report that their device fails to understand their commands, according to a recent study of 1000 British smart speaker owners by YouGov." "The researchers also found that women tend to speak more politely to their smart speakers, with "45% saying they "always" or "often" say 'please' and 'thank you', compared to only 30% of male owners"." "This discrepancy between male and female users could be a result of bias at the point of training AI assistants like Alexa or Siri; if programmers train the AI to respond to mainly male voices, it may have trouble recognizing female voices in the future. Not everyone believes this to be the case however. In its reporting of the study, the Evening Standard cites a blog post by founder and CEO of R7 Speech Sciences, Delip Rao, who believes that the discrepancy is down to technological issues rather than gender bias. "

"You can only understand the voice platform wars by first recognizing that voice assistants, specifically, represent both a platform and user interface (UI) shift comparable to the web and smartphones. The key difference is that these new platforms are neither based on open standards, nor on relatively open access to consumers. Voice assistants introduce a proprietary intermediary into all digital consumer interactions. This scenario both excites and frightens the leading tech companies that carved out enviable positions in the earlier web and smartphone platform wars." "Voicebot subsequently introduced the 5 A's framework for evaluating value segment dominance: Access, Acquisition, Authority, Attention, and Agency." "Each of these segments is again up for grabs in the new voice era. Voice assistants offer easy Access to content. They are being used to Acquire goods, and are a new source of Authority, as they answer billions of questions annually. They are also diverting Attention that previously went to smartphone interactions to new voice interaction properties. This means that each of the winners of the previous web and mobile platform wars has existing territory they must protect. " "Agency is what all of the big winners of the earlier tech platforms fear most. Voice assistants reserve agency for making choices about where answers are sourced (Authority), and can heavily influence content sourcing (Access), such as steering people toward Prime Video or YouTube. They also can order from multiple sources (Acquisition) that are not Amazon.com. And, they introduce new sources of interactions that displace consumer time with media (Attention). Voice assistants are an intermediary. " "Voice assistants are designed to help simplify users' lives. Over time, more and more agency will be granted to voice assistants to simply execute tasks on behalf of the user. The consumer will not necessarily care how the task is fulfilled, just that it gets done. " "If an agen

"Ce mois de mai aura lieu la conférence I/O de Google. Chaque année, c'est durant cette grande messe pour les développeurs que la firme de Mountain View lève le voile sur ses derniers projets." [...] "Michael Brenner, de Google, expliquera comment le projet Euphonia tire parti des technologies de Google pour redonner la parole aux personnes souffrant d'un trouble de la parole", lit-on dans cette description."

"Women Reclaiming AI (WRAI) is an expanding activist artwork, presented as a feminist AI voice assistant, programmed through participatory workshops by a growing community of self-identifying women. Through creating a platform for collective writing and editing, the project co-creates an AI that challenges prescribed gender roles. It is a response to the pervasive depiction of AI voice assistants gendered as women, subordinate and serving. Designed by development teams which lack diversity, these systems are embedded with unrepresentative world views and stereotype in ways that reinforce traditional gender roles. WRAI aims to reclaim female voices in the development of future AI systems by empowering self-identifying women to harness conversational AI as a medium for protest. This project is created by artists-technologists Coral Manton and Birgitte Aga in collaboration with an ever evolving community of self-identifying women."

"Selon Google, la nouvelle version de Google Assistant pourra répondre 10 fois plus vite à une commande vocale. Il sera même possible d'interagir avec l'assistant personnel tout en naviguant entre plusieurs applications, sans dire à chaque fois "OK Google". L'algorithme pourra plus facilement faire la différence entre l'action et le message que vous voulez dicter. [...] Les autres appareils Android munis de Google Assistant pourront, eux, attendre l'arrivée d'un Mode Conduite ainsi que des Références Personnelles. Cette dernière fonctionnalité permet à l'algorithme de mieux comprendre les requêtes du type "Que se passe là où habite maman ?". À noter que ces fonctionnalités seront prise en charge directement sur l'appareil et non les serveurs de Google." "Google utilise désormais sa marque Nest pour ses appareils domotiques. [...] Google a mis en place des mesures pour garantir votre vie privée. Selon les intervenants du Google I/O, une LED indique quand la caméra fonctionne. Aucun flux vidéo ne sera par ailleurs mis en ligne ou enregistré sans le consentement de l'utilisateur. Comme les autres produits connectés à Google Assistant, celui-ci proposera Voice Match : l'assistant vocal distingue les différentes personnes de votre maison selon leur voix. Avec la caméra embarquée, Google Assistant pourra également vous différencier selon vos visages. Google appelle cette nouvelle fonction Face Match. " "Pour améliorer l'accessibilité de ses produits, Google lance donc le projet Euphonia. Il vise à faciliter l'interaction des personnes sourdes, muettes ou souffrant d'un trouble du langage avec leurs smartphones ou enceintes connectées."

"At its annual developer conference in Mountain View, Google boasted of shrinking its speech recognition software to 1/25th of its prior size. CEO Sundar Pichai called that a "milestone" because it means software that traditionally lives in Google's cloud servers can be installed in Pixel smartphones Google will launch later this year, allowing the devices to respond to a person's voice much more quickly." "Werner Goertz, a research director at Gartner, calls the shift to on-device speech recognition "game changing" and potentially a significant challenge to Apple's and Amazon's more conventional speech systems. "Latency has always been an issue," he said, and most users have experienced the problem." "Processing speech on-device instead of transmitting it to the cloud can also be more private than the conventional model-although in some cases the transcribed text will be sent to Google anyway. Pichai and other executives made privacy a theme of Tuesday's event, gently trying to neutralize Google's reputation for data-slurping. The company showcased redesigned privacy settings, and a new "incognito mode" for Google Maps that pauses the service's default tracking of a device's movements."

"L'année dernière, Google a présenté Duplex, une intelligence artificielle qui permet aux utilisateurs de Google Assistant de réserver une table au restaurant ou de prendre un rendez-vous chez le coiffeur sans téléphoner. Comment ça marche ? Lorsque l'utilisateur demande à Assistant de faire la réservation, Duplex téléphone l'établissement et discute (grâce à l'intelligence artificielle) avec le responsable pour réserver. Une fois l'appel terminé, Google Assistant confirme la réservation à son utilisateur. CallJoy, un nouveau service lancé par Area120, l'incubateur interne de Google, fonctionne de manière similaire mais fait l'inverse. Contrairement à Google Duplex qui a été développé pour contacter les établissements à la place des clients, CallJoy est destiné aux entreprises et répond aux appels des clients."

"Microsoft ne délaisse pas Cortana, son assistant personnel. Si l'entreprise s'est plus ou moins détournée du marché grand public pour la voix, elle entend bien faire de Cortana un must en entreprise. Cela passe entre autres par la capacité de tenir des conversations plus longues et plus naturelles." "A sa conférence Build 2019, ce 6 mai, Microsoft a présenté un exemple impressionnant d'une cadre gérant tout son planning à l'oral de cette manière. Une performance inégalée à l'heure actuelle. Elle est rendue possible par la technologie de Semantic Machines, start-up rachetée en mai 2018."

La voix de djingo expliquée par Pascal Taillard designer sonore chez Orange : une voix avec des interjections pour plus d'humain mais pas trop avec des sons, pour rappeler la relation humain machine.

"Dr. Lam is collaborating with a group of Stanford faculty members and students to build a virtual assistant that would allow individuals and corporations to avoid surrendering personal information as well as retain a degree of independence from giant technology companies." ... "The Stanford researchers are hoping to gain support by making their software freely available to users of smartphones, computers and consumer appliances. They are encouraging makers of consumer products to connect their devices to the Almond virtual assistant through a Wikipedia-style service they call Thingpedia. It is a shared database in which any manufacturer or internet service could specify how its product or service would interact with the Almond virtual assistant."

"These findings reinforce past studies that showed Google Assistant as the clear leader in voice responses, helped by Google's decades of work sorting and organizing internet search results. These studies shows that Google may remain the top contender in voice results for years to come, despite Amazon's work to hire thousands of Alexa engineers and its huge push to expand its voice assistant's capabilities. While Alexa still dominates the US smart speaker market, with about 70% of the devices sold already, Google continues to gain ground and could use its superior tech to eventually beat out Alexa."