shut

ChatGPT voice mode has learned to listen, shut up and respond better

by

Greg Brockman started the OpenAI live doing something very simple: talking to ChatGPT and cut him off while he responded. What mattered was not the questions, but the system’s reaction. Instead of stopping with a sharp cut, as could happen in previous experiences, in the demo the voice seemed to adjust the interruption in a more natural way. It was a short demonstration, but it served to show the promise of GPT-Live– Let ChatGPT’s voice mode not only speak better, but know when to respond, when to wait, and when to shut up. GPT-Live is the name chosen by OpenAI for this new stage of ChatGPT’s voice mode. It’s not just a more polished sound layer, but a family of models designed to process speech differently. An architecture called full-duplex allows AI to listen while generating a response, something the company says reduces the rigidity of traditional shifts. Since 2022, OpenAI has been moving the relationship with its models from the keyboard to increasingly closer forms of interaction. ChatGPT marked a before and after in the massive relationship with text models; early voice features added a more natural layer; and GPT-4o took that ambition to a much more expressive experience, inevitably associated with ‘Her’. The GPT-Live announcement comes after that tour. Its promise is not just to sound more human, but to solve a less spectacular and more important friction: how to have a conversation. without it looking like a sequence of commands. ChatGPT voice no longer works only in turns To understand the change you have to look at how previous generations worked. OpenAI explains that ChatGPT’s first voice mode was a cascade system: one model transcribed the voice, another generated the response, and a third converted it back to audio. Advanced voice mode It reduced that friction by processing and generating audio within a single model, but still operated in turns. GPT-Live changes that logic: you no longer work just with separate messages, but with a continuous interaction that allows you to decide several times per second whether you should speak, listen, pause or use a tool. So what can we do in practice? Well, the company led by Sam Altman gives several examples. We can interrupt with a question, pause to think or ask the system to speak more slowly. The voice can also respond with brief signals such as “mmm” or “yes” to indicate that it is still listening, and the company says it has improved its ability to focus on the user’s voice when there is background noise. Added to this are the nine ChatGPT voices remastered for GPT-Live, visual responses in the form of cards for queries such as weather, sports, stocks and other quick data, as well as support for search, memory, images and file uploads. The other important piece is what happens when the conversation demands more than a quick response. OpenAI explains that GPT-Live can delegate web searches, reasoning or more complex tasks to its frontier models, while keeping the conversation with the user alive. At launch, that support layer will be GPT-5.5although the company says that it will update the model used in the background as it publishes new generations. ChatGPT’s voice mode will also allow you to choose between reasoning levels: Instant for quick responses, and Medium or High when you need to spend more time thinking. The company maintains that the jump is not limited to an impression of use. In their internal evaluations, the GPT-Live-1 and GPT-Live-1 mini (the smaller version of the model that lands on free accounts) appear above the advanced voice mode in comparative five- to ten-minute conversations, both in overall preference and fluency, interruptions, turns, and natural feel. OpenAI also cites advances in scientific reasoningagent search and simulated telephone support tasks. Of course, these results serve as an initial reference, but they come from OpenAI itself and we will have to see what impact they have in real life. GPT-Live starts rolling out today for ChatGPT users on iOS, Android and ChatGPT.com. OpenAI speaks of a global deployment and does not mention a specific exclusion for the European Union, although it is worth maintaining the nuance: in other deployments of AI functions there have been doubts or delays in the region. GPT-Live-1 will be the default model for Go, Plus and Pro users, while free accounts, as we say, will use GPT-Live-1 mini. The API will arrive later, without a specific date for now. Now, there is also an important limitation at launch: GPT-Live does not currently support voice with video or screen sharing in ChatGPT, although OpenAI says it is working to introduce those capabilities later. Images | OpenAI In Xataka | Meta already has its rival for Nano Banana 2. Its problem is the same as always: mercilessly invading our privacy

That Anthropic has shut down OpenClaw is understandable. That they do so confirms that they are becoming the Nintendo of AI

by

Peter Steinberger, the creator of AI agent OpenClawrose on Saturday to a ton of mentions on his Twitter account. In all of them they warned him of the same thing: Anthropic announced that Claude Code (Claude Pro/Max) accounts could not be used in OpenClaw. The decision left him anything but indifferent, and users of this agent have criticized a decision that, although reasonable, is, in a way, a disturbing tactic because of how and when it arrived. what has happened. OpenClaw is the AI ​​agent that, if you want, takes control of your machine and uses its apps to do for you everything you ask. Its operation is very powerful, especially if you use it with quality models such as Claude Opus 4.6 or Claude Sonnet 4.6. Many users were taking advantage of the Claude Pro and Claude Max plans to get the most out of OpenClaw, but Anthropic has said that that cannot be done. As explained, OpenClaw and other AI agents consume too many tokens and those plans are designed to be used in Claude Code for programming. If you want to use Claude with OpenClaw, pay. At Anthropic they do not prohibit the use of their AI models with OpenClaw, but they make it clear that if you want to use them you must use them with their API. It’s as if you bought the monthly transport pass for 20 euros to travel unlimited on the subway: it works perfectly when you go to and from work or university, but Anthropic says that you cannot use that pass to use it in your courier company that makes hundreds of trips a day. The token consumption in Claude Code is manageable, but in OpenClaw that consumption skyrockets and in Anthropic they want you to pay per use, not take advantage of the “flat rate” (with limits) of their Pro/Max plans. It’s understandable… Many users have attacked Anthropic and criticized that decision. Boris Cherny, one of the top managers of Claude Code, answered in X to a user who told him that decision “sucked”: “I know it sucks. At its core, engineering is about making hard decisions, and one of the things we do to serve a lot of customers is optimize how subscriptions work to reach as many people as possible with the best model. Third-party services are not optimized in this way, so it is very difficult for us to maintain it in the long term.” It is true that the massive use of Claude in OpenClaw raises the internal costs of Anthropic’s infrastructure: it does not pay off for them that so many instances of OpenClaw are being used with Claude, at least not if they are not used with the API. It is reasonable because these plans effectively “cheat” by being able to be used with this and other AI agents. But… …the moment is curious. This decision comes shortly after Anthropic has started to “copy” some of the OpenClaw features in their products, something we also expected. Claude Cowork, Dispatch and Remote Control have become the “official” ways to be able to do some of what OpenClaw does directly with Anthropic tools, and shortly after releasing them is when they begin to cover the way in which users can use their monthly plans. For Peter Steinberger, creator of OpenClaw, Anthropic’s decision comes now It is significant: “It’s funny how the timing coincides: first they copy the most popular features of our tool to their own closed product, and then they block access to open source.” Anthropic doesn’t lie. The technical argument that Cherny mentions is real, and the truth is that there is a capacity problem. Claude models are expensive to run, demand grows faster than infrastructure, and users of AI agents like OpenClaw consume resources much more intensively than conventional chat or Claude Code users. This is not sustainable, but it is also true that this decision comes three weeks after Steinberger “sold” OpenClaw to OpenAIAnthropic’s nemesis. Anthropic in Nintendo mode. This is the classic walled garden pattern: you see what works on another platform or rival product, absorb it, and then close the door. Nintendo has done this for decades with its platform developers, and Apple has perfected it with the App Store. The difference is that Nintendo and Apple had that walled garden from the beginning, and Anthropic is building it now. Although it’s not exactly the same. It should be noted that Nintendo is protecting an ecosystem with decades of irreplaceable IPs (Intellectual Properties): Mario, Zelda, or Metroid. It is normal that there is an access cost. Anthropic is doing that right now with Claude as the star product, but obviously it doesn’t have anything comparable (at the moment) to the IPs that Nintendo has. Here is another disturbing comparison: Apple or Nintendo charge to enter the ecosystem but it does not keep the meter running. Anthropic does: it has an increasingly closed garden, but it also forces the use of the API to use OpenClaw, with a pay-per-use model that is reasonable given Claude’s demand. But the rest do “leave”. What Anthropic has done clashes with what other AI companies are doing, especially when we talk about Chinese startups. The creators of Kimi, Minimax, GLM or the recent Xiaomi MiMo They do not have these policies: you can contract their monthly plans, very cheap, and take advantage of their models for OpenClaw without problems and without (barely) limits. It is true that these models are not as capable as Claude, but the way they act is still striking. In Xataka | OpenClaw changed the rules of the AI ​​race. Technology companies already have their answer: copy it

We knew that the Spaniards did not shut up or under water. Now a study has measured it: 6.9 seconds

by

It has happened to us all. You get into An elevator With four other strangers to climb to the ninth floor of a building and when you go for the second you already feel how a leading silence takes over the cabin, of those who seem to be cut with a knife. Uncomfortable. Annoying. Almost, almost pasty. The same when they introduce you to someone and nobody knows what to say or you just reached a first date. The silent silence. But … does everyone bother equally? A study He has just revealed no. A figure: 6.8 seconds. Yeah You feel uncomfortable When you are surrounded by strangers in a closed and small space and suddenly the silence becomes quiet: you are not alone. A study Prepared by the Online Platform for Language Learning PREPY He has confirmed that this sense of irritation before silence is common, very common. It is so common, in fact, that the vast majority of the 26,700 people from 21 different countries that have surveyed PREPLY in Your study They recognize sharing it. And not just that. Its report concludes that this feeling of restlessness does not take long to generate. On average, it occurs when we have 6.8 seconds without pronouncing or listening to a single word. In the case of the Spaniards they are 6.9. A percentage: 77%. That is the percentage of Spaniards who claim to feel uncomfortable when a conversation dies and no one is able to resume it: 77%almost eight out of ten. There are many, but much less than in other societies even more ‘allergic’ to silence. The palm in that aspect is taken by Brazil, where 85% feel uneasiness to mutism. In Italy they are 75%, in Colombia, 80%, in the US 82%and our French and Portuguese neighbors move between a fork of approximately 70 and 75%. A country: Thailand. The survey De Preply shows that discomfort in the face of prolonged and unwanted silences is a universal feeling. Of course, the key is what we understand by “prolonged silence.” Depending on Culturehabits and customs of each country The span of mutism that we are able to tolerate varies considerably. Two data arrives to check it: in Brazil on average they feel uncomfortable after 5.5 seconds of silence, while in Japan that does not happen until past 8.1. It is not easy PREPLY STUDY It is that our relationship with him is complex. For example, Thailand is the country of the list that to tolerates the most seconds and secondly is also an Asian nation, Japan, with 7.8 seconds. But the third and fourth positions are for European countries: Netherlands, where the brand is 7.4; and Germany (7,3). If we cross the data to 6.2. City Second average before feeling uncomfortable Saragossa 5.79 Valladolid 5.95 Murcia 6.05 Barcelona 6.21 Valencia 6.28 Bilbao 6.29 Palma de Mallorca 6.5 A Coruña 6.55 Las Palmas 6.68 Gijón 6.75 Malaga 6.84 Cordova 6.86 Madrid 7.01 Seville 7.17 Alicante 7.32 Grenade 7.51 Vitoria 7.66 Santa Cruz de Tenerife 8.44 Vigo 8.64 San Sebastián 8.67 A city: San Sebastián. The report is not limited to analyzing countries. It also compares cities. And in the case of Spain it reveals some striking contrasts. According to the data collected by PREPLY, the most tolerant Spaniards to involuntary silences are donostiarras. In San Sebastián they do not feel uncomfortable until past 8.67 seconds, more or less like the vigueses (8,64) and Tenerife (8,44). In the opposite pole are the Zaragozanos, who worry at 5.79 seconds, and the Pucelanos (5.95). A place: the elevators. Not everyone reacts to mutism. And not all places awaken the same sensations. 79% of respondents recognized that the site where they have been found more commonly with pasty and heavy silences are closed spaces, such as elevators. Other situations in which that same sensation abounds are the breakups (73%) and when you maintain a first date with someone (72%). In the list there are other scenarios equally predictable, such as casual talks with strangers, funerals or conflict situations. A situation: the presentations. There is something that seems to be especially uncomfortable: the awkward silences during the Public presentations. 36% of respondents by PREPY recognized that this is the situation in which mutism fear the most, even more than in the first appointments, fights, elevator trips, couple discussions or family gatherings. Where we have more assumed are in the talks with co -workers or when we contact someone online. Images | Baruk Granda In Xataka | This is the most silent room in the world. No one is able to endure an hour in it

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.