Humyn Labs's banner

Humyn Labs

@humynlabs • 1,873 subscribers

Verified (human) intelligence at scale for reliable AI. Diverse experts. Multi-layer QC. Every modality. https://t.co/uKKqzO1Syf

Shorts

AI listens to everyone. Except for 5.5 billion people. We’re launching BRIDGE (Benchmark of Regional & International Data for Global Evaluation), the first independent Global South ASR benchmark. 22 regional languages across 15 models evaluated on a first-of-its-kind 7-metric stack. Read the full benchmark here: Most ASR benchmarks report one WER per model per language. That number hides more than it reveals. Dialect variation, code-switching fidelity, speaker overlap and script-mismatch all sit underneath it. BRIDGE surfaces them. Here’s how 🧵

AI listens to everyone. Except for 5.5 billion people. We’re launching BRIDGE (Benchmark of Regional & International Data for Global Evaluation), the first independent Global South ASR benchmark. 22 regional languages across 15 models evaluated on a first-of-its-kind 7-metric stack. Read the full benchmark here: Most ASR benchmarks report one WER per model per language. That number hides more than it reveals. Dialect variation, code-switching fidelity, speaker overlap and script-mismatch all sit underneath it. BRIDGE surfaces them. Here’s how 🧵

30,996 görüntüleme

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

AI is only as intelligent as the data it learns from. And data is only as good as the humans behind it. Introducing Humyn Labs. Where human intelligence is not just a step, it’s the filter. 👇

AI is only as intelligent as the data it learns from. And data is only as good as the humans behind it. Introducing Humyn Labs. Where human intelligence is not just a step, it’s the filter. 👇

17,166 görüntüleme • 4 ay önce

The best conversations at GTC didn't happen in the sessions. They happened after. That's why we threw the party. We hosted Open Hands on March 17 in San Jose, a casino night that brought together 100+ researchers, founders and operators. Teams from Adobe, NVIDIA, Intel, TELUS Digital, Aptos, tuesday and 60+ others. Researchers from UC Berkeley, Carnegie Mellon University and institutions across the US and Asia. All in one room, across tables, over drinks. Three things we're taking back from San Jose: > The data bottleneck for physical AI is real and underestimated. Getting a robot to pick up an object, navigate a room, respond to a human. Each of those tasks needs millions of hours of egocentric video across real environments with real variability. Something we've been building towards and GTC validated it. > The Global South, India in particular, is emerging as a critical source of the linguistic and environmental diversity that production-grade models need. Low-resource languages, diverse acoustic conditions, real-world dialects. The data infrastructure for this doesn't exist at scale yet. We're building it. > Human intelligence remains the key variable that separates models that benchmark well from models that hold up in deployment. Judgment, nuance, and context aren't features you can generate. They have to be sourced, verified and embedded from the start. Grateful to albert chun (Founder, AI Circle) for co-hosting a night that proved the point. The best signal at NVIDIA GTC wasn't on a slide. It was in this room. Ishank Gupta Manish | HumynLabs; KGeN Arunabh Rohatgi Anoushka

The best conversations at GTC didn't happen in the sessions. They happened after. That's why we threw the party. We hosted Open Hands on March 17 in San Jose, a casino night that brought together 100+ researchers, founders and operators. Teams from Adobe, NVIDIA, Intel, TELUS Digital, Aptos, tuesday and 60+ others. Researchers from UC Berkeley, Carnegie Mellon University and institutions across the US and Asia. All in one room, across tables, over drinks. Three things we're taking back from San Jose: > The data bottleneck for physical AI is real and underestimated. Getting a robot to pick up an object, navigate a room, respond to a human. Each of those tasks needs millions of hours of egocentric video across real environments with real variability. Something we've been building towards and GTC validated it. > The Global South, India in particular, is emerging as a critical source of the linguistic and environmental diversity that production-grade models need. Low-resource languages, diverse acoustic conditions, real-world dialects. The data infrastructure for this doesn't exist at scale yet. We're building it. > Human intelligence remains the key variable that separates models that benchmark well from models that hold up in deployment. Judgment, nuance, and context aren't features you can generate. They have to be sourced, verified and embedded from the start. Grateful to albert chun (Founder, AI Circle) for co-hosting a night that proved the point. The best signal at NVIDIA GTC wasn't on a slide. It was in this room. Ishank Gupta Manish | HumynLabs; KGeN Arunabh Rohatgi Anoushka

10,727 görüntüleme • 3 ay önce

Daha fazla içerik yok.