What does it take to scale self-distillation across many GPUs? Self-distillation replaces the fixed, pre-trained teacher in standard knowledge distillation with a teacher that is an exponential moving average (EMA) of the student’s weights: - The student is trained with

Your agent needs to know who said what. Here's how→ https://t.co/N2sB3Ps5b9 @Speechmatics is built for developers who want to give agents true speaker understanding and get full control over their voice stack • Real-time diarization delivers word-level speaker labels (S1, S2,

What if the "one big model" approach to voice AI is already outdated? @pipecat_ai + @trydaily founder @ninacali4 thinks so. 👀 Get her take on multi-model futures, why she convinced her board to go open source, and where real innovation happens now: https://t.co/sXURbpSx0R

https://t.co/WaqAz9bLxi 👀

Regulators are zeroing in on AI bias. If your models don't work for everyone they serve, you're already non-compliant. Our CEO breaks down why audit trails matter less than inclusive foundations in Forbes ↓

#Paris #VoiceAI #events... get your tickets !!! (free with the code VOICEAIVIP) link in the comments

https://t.co/GgM5dhkRC6

What's next for Ambient AI? Scribes proved themselves in the consulting room, but healthcare's chaos zones are the real test. Resuscitation bays. Ambulances. Three clinicians talking over alarms, critical decisions in seconds, accents spanning continents. The quiet spaces

From why @uberboffin dreams in code to which AI trends are just marketing hype – plus the one problem he's working on that he refuses to tell us about. Read the full conversation on Resonance 👇 https://t.co/WIZArwnDeF

Should your voice agent be able to fool you it's human? 🤖 @uberboffin , our Senior Director of Innovation (and semi-nocturnal Head AI Tinkerer), doesn't think so. His take: transparency beats deception. Full article linked in comments 👇.

We're proud sponsors of Voice AI Space Paris – a full day of expert panels, workshops, and demos from voice AI practitioners actually building this tech. Developers, researchers, founders, we hope to see you there. 🎟️ https://t.co/n5bD1kv2F4 For your free ticket use code:

Qualité ou quantité? Join us in Paris to find out 🇫🇷 The countdown to Voice AI Space Paris in on: December 10th at @CNIT_Paris La Défense. Our CSO @RiquiHerreros will take to the main stage for the panel: "Voice AI Quality & Performance: What's Next?" Expect real talk on

@andysmanclubuk The work done by the team at @andysmanclubuk saves lives. Help us spread the word. Like, share, comment. Learn more: https://t.co/ejrlB3Sdmc #ItsOkayToTalk

This month we're proudly supporting @andysmanclubuk. They run free peer support groups for men across the UK, tackling suicide prevention through conversation. Because talking about mental health matters. We asked our team what they do for theirs. Watch below.

The fastest voice agent in the world is still terrible if it can't read the room. @uberboffin breaks down why the industry is optimizing for the wrong metric: https://t.co/IxYfEBvuZm

According to our Head Tinkerer, what voice AI actually needs to master: 🔊 Task type: knowing when speed matters (train times) vs when conversation needs space to breathe 🔊Speaker patterns: recognizing if someone pauses to think vs finishing their point 🔊Context signals: using

"Your voice agent doesn't need to be faster" 👀 - @uberboffin While the industry obsesses over milliseconds, we're building AI that can't tell when someone's finished speaking or when they're clearly annoyed. Speed without context is just noise. 🧵

https://t.co/VRBTCMlk32

How do you differentiate when everything can be generated in seconds? 🤖 We spoke to T-Bot from Voice AI Space ahead of our Paris event. His take: you need to go beyond your tech specs. Link below 👇

How do you build a brand that actually stands out in AI? 🤖🤖🤖 T-Bot breaks down his approach to differentiation, community building, and his dream unlimited-budget event. Read the full interview: