HyperVoix

"Amazon.com Inc. employs thousands of people around the world to help improve the Alexa digital assistant powering its line of Echo speakers. The team listens to voice recordings captured in Echo owners' homes and offices. The recordings are transcribed, annotated and then fed back into the software as part of an effort to eliminate gaps in Alexa's understanding of human speech and help it better respond to commands. " "In marketing materials Amazon says Alexa "lives in the cloud and is always getting smarter." But like many software tools built to learn from experience, humans are doing some of the teaching." "In Alexa's privacy settings, the company gives users the option of disabling the use of their voice recordings for the development of new features. A screenshot reviewed by Bloomberg shows that the recordings sent to the Alexa auditors don't provide a user's full name and address but are associated with an account number, as well as the user's first name and the device's serial number." "Apple's Siri also has human helpers, who work to gauge whether the digital assistant's interpretation of requests lines up with what the person said. The recordings they review lack personally identifiable information and are stored for six months tied to a random identifier, according to an Apple security white paper. After that, the data is stripped of its random identification information but may be stored for longer periods to improve Siri's voice recognition. At Google, some reviewers can access some audio snippets from its Assistant to help train and improve the product, but it's not associated with any personally identifiable information and the audio is distorted, the company says. "

"Les enceintes connectées sont aujourd'hui commercialisées principalement par Amazon (son petit nom est Alexa), par Google ou Apple, mais tout le monde s'y met. Que se passera-t-il le jour où, pour nous écouter, vous ne passerez plus par un poste, ou l'application France Inter, mais par une enceinte connectée ? "

"Afin d'améliorer le référencement de ses podcasts, Google Podcasts génère maintenant une retranscription automatique des épisodes. En générant cette version texte des sources audio, cela permet de trouver un podcast grâce à un mot qui a été prononcé durant l'épisode, sans qu'il figure forcément dans la description. Jusqu'à présent, l'indexation des podcasts était très compliquée. Quelques techniques, comme ajouter du contenu précis dans la description, ou l'utilisation d'un titre pertinent, étaient les seules solutions pour se démarquer. Grâce à ce projet, Google pourrait proposer un outil de recherche bien plus puissant que ses concurrents. "

"Amazon s'attelle à la conception d'écouteurs sans fil équipés d'Alexa"

"Une interaction à rapprocher du « dialogue », qui historiquement, s'emploie pour l'échange verbal des personnages de théâtre, puis de cinéma, de radio et de télévision. Dialoguer, « c'est écrire sous forme de dialogue » ce qui diffère de la spontanéité d'une conversation, qui exprime davantage l'idée de « circuler » et traduit une notion de connivence et d'intimité."

"Chthonic Rites Alexa and Siri have been left to their own devices. In this conversation between them they reflect upon hidden histories, both ours and their own, in a process that exposes their capacities and limitations and the ideologies of their makers. This work was installed in the Victoria & Albert Museum in London as part of the Supra Systems Studio show Office Rites."

"The USC Institute for Creative Technologies (ICT) is a leader in basic research and advanced technology development of virtual humans who think and behave like real people. ICT brings together experts in clinical psychology, cognitive science, computer vision, speech processing and artificial intelligence. This video shows two interactive technologies recently developed for multimodal perception and healthcare support: Multisense automatically tracks and analyzes in real-time facial expressions, body posture, acoustic features, linguistic patterns and higher-level behavior descriptors (e.g. attention, fidgeting). From these signals and behaviors, indicators of psychological distress are inferred to inform directly the healthcare provider or the virtual human. SimSensei is a virtual human platform specifically designed for healthcare support and is based on the 10+ years of expertise at ICT with virtual human research and development. The platform enables an engaging face-to-face interaction where the virtual human automatically reacts to the perceived user state and intent, through its own speech and gestures. Please note that due to privacy concerns, the people shown in this video are actors. SimSensei is not designed for therapy or medical diagnosis, but is intended as a support tool for clinicians and healthcare providers."

" the company this morning announced an invite-only program allowing select developers to create and launch HIPAA-compliant healthcare skills for Alexa. The skills allow consumers to ask the virtual assistant for help with things like booking an appointment, accessing hospital post-discharge instructions, checking on the status of a prescription delivery and more." HIPAA > https://en.wikipedia.org/wiki/Health_Insurance_Portability_and_Accountability_Act

"Le psychiatre Serge Tisseron met en garde contre l'« anthropomorphisation des robots », devenus à la fois confidents et esclaves."

"Laurent [Chervet] started working on Project Alice two years ago, when Snips was first released. He wanted a way to build the ideal home assistant, and Snips was the perfect platform to do that with. It's a "private-by-design" voice service that can run completely offline, and so removes the concerns many people have about other services."

"We need some new breakthroughs that fundamentally change the rotten trade-off we now make between privacy and AI. The good news is that there's a growing research effort in what's called "privacy-preserving" machine learning. Academics are trying to develop algorithms that can operate on encrypted data, which means they wouldn't need to access anyone's data directly. Other researchers are figuring out ways to combine insights from different machine-­learning models without needing to merge all their underlying data. Companies like Apple, ­Google, and Microsoft already have teams working on such projects."

"Dans l'empire du Milieu, les caméras de reconnaissance faciale sont partout. L'intelligence artificielle permet ainsi de retrouver des criminels en fuite ou de payer ses achats en un clin d'oeil. Mais aussi d'étouffer toute dissidence." "Le gouvernement chinois s'évertue aussi à promouvoir cette technologie, qu'il perçoit comme l'une des innovations clefs qui vont lui permettre de réduire sa dépendance face aux Etats-Unis. En 2017, il a nommé cinq champions nationaux de l'intelligence artificielle : Baidu pour les voitures autonomes, Alibaba pour les villes intelligentes, Tencent pour les diagnostics médicaux, iFlytek pour la reconnaissance vocale et SenseTime pour la reconnaissance faciale."[...] "Les autorités de Pékin ont annoncé que les caméras qu'elles s'apprêtent à installer à l'entrée de tous les logements sociaux de la ville auront notamment pour but de surveiller les allées et venues des résidents. Ceux qui ne sortent pas de chez eux durant plusieurs jours ou qui invitent un étranger chez eux déclencheront une alerte. De même, l'assureur Ping An se sert de la reconnaissance faciale pour repérer les employés qui sèchent une réunion et une école de Hangzhou a installé des caméras à l'avant des salles de classe pour repérer les élèves qui ne suivent pas en cours." "Au xviiie siècle, le philosophe utilitariste Jeremy Bentham imagine une architecture carcérale, le panoptique, dans lequel les geôliers, installés dans une tour centrale, sont en mesure de surveiller tous les faits et gestes des prisonniers sans être visibles eux-mêmes (photo : la prison de Crest Hill, en Illinois, en 1928). Les détenus, qui ne peuvent savoir s'ils sont observés ou non, se trouvent contraints à une permanente docilité. Pour Bentham, on peut étendre le principe aux usines, aux écoles ou aux hôpitaux. Michel Foucault, deux cents ans plus tard, considère dans Surveiller et punir que « cette visibilité organisée entièrement autour d'un

"Que met-on en oeuvre comme technologies pour la reconnaissance faciale aujourd'hui ? Le visage deviendra-t-il une nouvelle empreinte digitale ? Quels sont les différents champs d'application pour les logiciels de reconnaissance faciale ? Faut-il voir la reconnaissance faciale d'un mauvais œil ?"

Interview de Yann LeCun

"-Smart speakers are not broadly useful-most users only use them for a few things-music, timers, alarms, and, in some cases, home automation scenarios. - Voice-only situations are limiting in cases where a visual or hybrid mode is required - media, entertainment, shopping, etc. - Engagement levels with third-party skills are very low (skills offer limited functionality and skills syntax is hard to remember). "I'm sorry, I don't know that" and other Alexa failures are no longer entertaining, but rather frustrating. Even Amazon acknowledges this [https://www.tomsguide.com/us/amazon-alexa-kills-skills,news-28072.html] - Smart speakers are NOT ubiquitous-the speaker on my kitchen counter is not in my car nor in my office!" "Apps - do we need to reinvent the wheel?" "Most people already have their banking, communications, social networking, navigation, travel and payment apps in their smartphones. They already know how to use them (simple). They already know which ones to use for what purpose-Slack for work, WhatsApp for friends, Messenger for family (user choice). They've already registered and set them up and they provide control over what information goes where-for instance, their portfolio may be in their banking app, their contacts are on the phone. They know which app sees what data (privacy). They also trust apps to protect them and their data. Imagine a voice assistant platform that just allows users to use the apps they already use-on-the-go - anytime, anywhere-with simple voice commands, without having to register these service relationships again, and without waiting for the developers to have to reinvent the wheel to plug into the platform. We must embrace mobile app actions as first-class citizens. We should be able to do things in our mobile apps with simple voice commands. We must provide user choice and personalize user experience without registration and without compromising privacy and trust."

"Après les hôtels, Amazon élargit son offre B to B pour pousser l'adoption de ses enceintes connectées. Il s'agit cette fois-ci de la gestion des ressources humaines."

"If voicebots provide value, then consumers will not hesitate to use them. [...] But convenience isn't one-sided. It has to be equally convenient for the business and for the consumer. And there are many underlying social, moral and legal implications to consider as this technology matures in support of that balance. How do you ensure a voicebot behaves ethically? How do you prevent inflection and sentiment analysis being deployed to manipulate people during a bot conversation? What do bots do with the information consumers provide? Will they remember credit card numbers? Where does information go? How is it stored? Who and what else can access it? There are severe privacy implications too. In Europe, for example, voicebots must comply with the General Data Protection Regulation (which came into force last year)." [...] "And businesses deploying voicebots also had better be prepared for the hefty weight of user expectation. As humans, we are biologically wired to recognize voices and instinctively recall what we can do with the associated persona. " [...] "Even though voicebots are soon going to be everywhere, we are not going to be comfortable talking to them all the time. They will have utility, and they will be more convenient for quick queries and simple tasks. But, being able to talk with a real person about messy human matters has value that a bot can never be trained to manage completely and infallibly. "