Nathan Roll @ EMNLP 2026's banner

Nathan Roll @ EMNLP 2026

@nrol_ling • 1,941 subscribers

CEO @OrukLabs | a16z Speedrun A001/SR007 | PhD student @Stanford/@stanfordnlp

Shorts

humans are NOT big fruit flies. Here's how they process emotion differently. Oruk

humans are NOT big fruit flies. Here's how they process emotion differently. Oruk

755,452 次观看

we trained a fruit fly to understand emotion We created a neural network from the male drosophila connectome, then trained it on human emotions. It was shocking to see how good it was at identification (better than some humans I know)

we trained a fruit fly to understand emotion We created a neural network from the male drosophila connectome, then trained it on human emotions. It was shocking to see how good it was at identification (better than some humans I know)

613,864 次观看

AI doesn't hear the world like people do. In fact, AI doesn't hear the world like other AI. Across 24 speech recognizers, phonemes were easiest to read out at 21% of encoder depth in Conformers on average, vs 49% in Transformers. Catch my presentation tomorrow at EMNLP.

AI doesn't hear the world like people do. In fact, AI doesn't hear the world like other AI. Across 24 speech recognizers, phonemes were easiest to read out at 21% of encoder depth in Conformers on average, vs 49% in Transformers. Catch my presentation tomorrow at EMNLP.

30,323 次观看

we gave our fruit fly brain damage (so you dont have to) we lesioned one neuron in our drosophila model and it starting misidentifying emotions. should we try this in people? Oruk is building AI speech models that actually understand emotion/intent/delivery.

we gave our fruit fly brain damage (so you dont have to) we lesioned one neuron in our drosophila model and it starting misidentifying emotions. should we try this in people? Oruk is building AI speech models that actually understand emotion/intent/delivery.

229,287 次观看

we made a speech model 13× smaller by making it repeat itself. 52 layers → 2 shared blocks. 2.04B parameters → 153M. The encoder loops 24 times. The decoder loops 28. Smaller weights, roughly the same compute. Oruk

we made a speech model 13× smaller by making it repeat itself. 52 layers → 2 shared blocks. 2.04B parameters → 153M. The encoder loops 24 times. The decoder loops 28. Smaller weights, roughly the same compute. Oruk

53,634 次观看

Not all models listen in the same way! We probed 24 speech encoders. Speech-sound categories were easiest to read out early in Conformers and much later in Transformers. Catch our presentation on Monday @ EMNLP 2026 in the 9:00–10:30 AM CET session in Room F9–10.

Not all models listen in the same way! We probed 24 speech encoders. Speech-sound categories were easiest to read out early in Conformers and much later in Transformers. Catch our presentation on Monday @ EMNLP 2026 in the 9:00–10:30 AM CET session in Room F9–10.

10,161 次观看

Voice AI is getting the words right, but it still doesn't understand tone. We paused our PhDs at Stanford to build Oruk, backed by a16z speedrun. We train speech models to understand what you say AND how you say it. DM me for preview access to our most powerful model yet

Voice AI is getting the words right, but it still doesn't understand tone. We paused our PhDs at Stanford to build Oruk, backed by a16z speedrun. We train speech models to understand what you say AND how you say it. DM me for preview access to our most powerful model yet

13,971 次观看

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

Speech can survive being flipped backward every twenty milliseconds. The same speech in backward pieces of 200 then 100 then 50 then 20 milliseconds. The original comes last. When do the words become clear? A listening demo from Oruk.

Speech can survive being flipped backward every twenty milliseconds. The same speech in backward pieces of 200 then 100 then 50 then 20 milliseconds. The original comes last. When do the words become clear? A listening demo from Oruk.

Nathan Roll @ EMNLP 2026

634,702 次观看 • 13 天前

Voice AI is getting the words right, but it still doesn't understand tone. We paused our PhDs at Stanford to build Oruk, backed by a16z speedrun. We train speech models to understand what you say AND how you say it. DM me for preview access to our most powerful model yet

Voice AI is getting the words right, but it still doesn't understand tone. We paused our PhDs at Stanford to build Oruk, backed by a16z speedrun. We train speech models to understand what you say AND how you say it. DM me for preview access to our most powerful model yet

13,971 次观看 • 1 个月前

没有更多内容可加载