Talk, don’t type: Big Tech bets AI’s future will be spoken

{"text":[[{"start":8.18,"text":"Artificial intelligence groups are racing to make conversations with chatbots feel more natural, betting that talking will overtake typing as the dominant way people interact with AI agents despite the social awkwardness of speaking to machines."}],[{"start":21.32,"text":"OpenAI and Google have recently updated their voice offerings to make AI sound less like a robot reading from a screen and more like a colleague who understands instructions and context."}],[{"start":32.08,"text":"Their latest products come amid exploding demand from consumers and businesses and represent a significant break from earlier voice assistants, which converted speech into text, generated a text response and then read it back."}],[{"start":45.04,"text":"Newer models can process and produce speech directly, helping to reduce delays, awkward tone and missed context that made earlier products feel mechanical."}],[{"start":54.76,"text":"“People love using their voice to ask questions . . . It is just good for that humanlike back-and-forth conversation. Text is great too, but typing is a lot of effort,” said Atty Eleti, OpenAI’s product lead for ChatGPT voice."}],[{"start":68.6,"text":"The bet is that as AI systems move from answering questions to carrying out longer-running tasks, speaking to them will feel more natural than typing prompts into a chatbot window."}],[{"start":78.6,"text":"But the push is also exposing the old frictions of voice technology, from lag and background noise to the social discomfort of talking to machines in public."}],[{"start":87.32,"text":"AI groups say the shift is already changing user behaviour. Google said its live voice usage doubled in the year to April, with chatbot conversations on average five times longer than on text. At OpenAI, more than 150mn people use voice in ChatGPT each week."}],[{"start":104.16,"text":"Voice as a medium is “really starting to rise again with these long-running agents where it just feels very natural to give them tasks via voice”, said Sarah Bird, chief product officer of responsible AI at Microsoft."}],[{"start":116.08,"text":"Bird added that she often sees colleagues working on GitHub, the company’s coding platform, sending voice commands to their phones to code. “I think it’s going to be a new world soon on this,” she added."}],[{"start":125.96,"text":"OpenAI said voice technology was previously hindered by a “lag in intelligence”. Until the release of ChatGPT Live last week — its first voice model in two years — OpenAI’s voice model had been criticised for being slow and confusing different languages or accents."}],[{"start":142.06,"text":"“People are just starting to speak more because they know the model understands, and they’re now expecting they can do more [and] they want those devices to do more,” said Leland Rechis, who leads voice experiences at Google."}],[{"start":154.64,"text":"Rechis added that Google opted for a “far more spontaneous conversational style” after finding that users became more comfortable speaking to AI once large language models were “behind the microphone”."}],[{"start":165.08,"text":"One early use case is coding, where advocates say speech can help users describe intent and constraints more fully than a typed prompt."}],[{"start":174.04,"text":"Sandro Gianella, who works in international strategy and operations at OpenAI, said voice has become an important part of how he uses OpenAI’s coding tool Codex."}],[{"start":183.56,"text":"“When I type, I often compress the problem too early and jump straight to the task. When I speak, I tend to explain more of the why: what I am actually trying to achieve, what my intention is, which constraints matter, where I am uncertain, and what a good outcome would look like,” he wrote on LinkedIn this month."}],[{"start":200.92,"text":"The rise of voice tools has prompted some start-ups in Silicon Valley to invest in new microphones and office soundproofing to prevent systems from confusing audio with background noise. These include everything from lapel microphones to ones housed in a black tube and a mouthguard."}],[{"start":null,"text":"

"}],[{"start":217.24,"text":"The shift has opened a new front for hardware start-ups trying to make voice input less conspicuous and more reliable, such as small handheld dictation devices from Pocket and Plaud."}],[{"start":227.32,"text":"OpenAI is also expected to release the first in its family of devices early next year, which will be an ambient voice assistant with a camera, microphone and speaker, according to multiple people familiar with the project."}],[{"start":239.32,"text":"AI voice start-ups are also attracting more interest from investors. Venture capital investment rose to $7bn in the first quarter of this year, compared to just under $1bn in the same period last year, according to PitchBook data."}],[{"start":253,"text":"New York-based start-up Sandbar has designed a ring that you can speak into to give commands or record notes, which responds only to the user’s voice."}],[{"start":262.12,"text":"Chief executive and co-founder Mina Fahmi said he developed the product after finding existing products awkward to use. “Trying to use voice by shouting into my AirPods was very bad for everyone around me and very embarrassing personally.”"}],[{"start":276.48,"text":"Evan Goldschmidt, chief technology officer at Tolan, a voice start-up with an AI alien avatar, said the technology still had room for improvement with background noise, reliability and responding to the user with the right emotional tone and context."}],[{"start":291.04,"text":"“We’re going up the curve much more quickly, but definitely there are still those paper cuts that need to be worked through,” he added. “It is only a matter of time before [voice] is a dominant interaction paradigm.”"}],[{"start":306.64,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1786001473_4010.mp3"}