We present a generative dialogue system capable of operating in a full-duplex manner, allowing for seamless interaction. It is based on a large language model (LLM) carefully aligned to be aware of a perception module, a motor function module, and the concept of a simple finite state machine (called neural FSM) with two states. The perception and motor function modules operate in tandem, allowing the system to simultaneously speak and listen to the user. The LLM generates textual tokens for inquiry responses and makes autonomous decisions to start responding to, wait for, or interrupt the user by emitting control tokens to the neural FSM. All these tasks of the LLM are carried out as next token prediction on a serialized view of the dialogue in real-time. In automatic quality evaluations simulating real-life interaction, the proposed system reduces the average conversation response latency by more than 3 folds compared with LLM-based half-duplex dialogue systems while responding within less than 500 milliseconds in more than 50% of evaluated interactions. Running a LLM with only 8 billion parameters, our system exhibits a 8% higher interruption precision rate than the best available commercial LLM for voice-based dialogue.

Perplexity has rolled out an update for its iOS app, giving iPhone users access to its AI voice assistant that was initially released for Android users earlier this year. Its voice assistant can perform tasks for the user by browsing the web or accessing other apps for them. If they ask the assistant to find them a table for a specific restaurant, for instance, Perplexity can launch the OpenTable app with the number of people, the date and the time already filled out. The user still has to perform the final action and book a reservation, but it's already laid out for them -- all they have to do is click the button. Users can also ask the assistant to draft emails for them for specific contacts, which they'll have to send themselves, and create reminders for them on the calendar.

Car manufacturers will be able to develop new AI voice assistants for their cars thanks to a new partnership with Qualcomm and Google. Qualcomm announced earlier today that it's working with Google on a new AI development system for carmakers. The new version is based on Android Automotive OS (AAOS), Google's infotainment platform for cars. Qualcomm is offering its Snapdragon Digital Chassis with Google Cloud and AAOS to generate new AI-powered digital cockpits for cars. Qualcomm also unveiled two new chips for powering driving systems including the Snapdragon Cockpit Elite for dashboards and the Snapdragon Ride Elite for self-driving features.

LG has introduced a smart home hub called ThinQ ON that has the technology to control not just LG-branded appliances but also other smart home devices. It comes with a built-in speaker that gives you a way to talk to LG's AI voice assistant, so you can use it to look up information, as well as to control your smart devices with spoken commands. LG says its technology can "understand the context of conversations" and can determine your preference for a specific device. It could, perhaps, tell your preferred temperature for the thermostat or the washer cycle you typically use. And it can notify you when a task is done, such as when the dryer cycle is finished.

We present a generative dialogue system capable of operating in a full-duplex manner, allowing for seamless interaction. It is based on a large language model (LLM) carefully aligned to be aware of a perception module, a motor function module, and the concept of a simple finite state machine (called neural FSM) with two states. The perception and motor function modules operate simultaneously, allowing the system to simultaneously speak and listen to the user. The LLM generates textual tokens for inquiry responses and makes autonomous decisions to start responding to, wait for, or interrupt the user by emitting control tokens to the neural FSM. All these tasks of the LLM are carried out as next token prediction on a serialized view of the dialogue in real-time. In automatic quality evaluations simulating real-life interaction, the proposed system reduces the average conversation response latency by more than 3 folds compared with LLM-based half-duplex dialogue systems while responding within less than 500 milliseconds in more than 50% of evaluated interactions. Running a LLM with only 8 billion parameters, our system exhibits a 8% higher interruption precision rate than the best available commercial LLM for voice-based dialogue.

This article was featured in the One Story to Read Today newsletter. Let's get this out of the way: OpenAI's voice assistant doesn't sound that much like Scarlett Johansson. The movie star has alleged that, though she rebuffed multiple attempts by Sam Altman, the company's CEO, to license her voice for the product that it demoed last week, the one it ended up using was "eerily similar" to her own. Not everyone finds the similarity so eerie--to my ear, it lacks her distinctive smoky rasp--but at the very least, the new AI does appear to imitate the playful lilts and cadences that Johansson used while playing Samantha, the digital assistant in the 2013 film Her. That's depressing--and not only because OpenAI may have run roughshod over Johansson's wishes, but because it has made such an unimaginative choice.

Arizona mother Jennifer DeStefano joins'America's Newsroom' to share her experience after being targeted by scammers who used artificial intelligence to clone her daughter's voice in an extortion plot. Voice cloning through artificial intelligence is quickly becoming more advanced, more accessible and more widely used. While using artificial intelligence to clone voices can be beneficial in making certain work more efficient and assisting those who have lost their ability to speak, there are also many problems that have, so far, been hard to solve. Anyone who has a phone, computer or similar technological device has access to voice cloning through a long list of available software. Even though some are more advanced and efficient than others, voice cloning as a whole is an extremely easy thing to do.

Care patients could see at home visits replaced by a call from an AI-powered voice assistant in a new British trial. Dubbed'Siri for care', a human-like virtual assistant will ring patients once a week to ask a list of automated questions. An algorithm will then analyse the answers and alert carers if there are any deteriorations in health so they can arrange a doctor's visit. Similar trials in Europe have reduced A&E visits by 55 per cent, according to the tech company behind it. The new technology will be tested out on patients in domiciliary care for those who are living independently but who rely on helpers to visit them regularly.

Today, artificial intelligence (AI) has become an irreplaceable part of how we shop and do business on the web. It's a key component of the underlying infrastructure that brands and retailers rely on to engage customers, track trends, make better business decisions, and provide the most optimal, personalized customer experiences possible. Here are the top nine trends in eCommerce AI that you can expect to see in 2022 and 2023. AI voice assistants like Amazon's Alexa, Apple's Siri, and Google Assistant have become household names used by millions of people worldwide. In fact, 27% of shoppers took advantage of voice assistants to make online purchases in 2020, accounting for $40 billion of revenue in the U.S. and the UK alone.