“Normally you would take 40 to 100 diffusion steps, you can reduce it to four… flow matching… makes the transformation much simpler… works just as well”
♫
Said on Latent Space: The AI Engineer Podcast
“Under the hood, it’s a diffusion transformer… transformer ingests the most amount of data… diffusion helps you to refine the outcome to get to the smooth, precise, fine-grained motion”
Latent Space: The AI Engineer Podcast · Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy - Jason Ma, Dyna Robotics
Play from 4:10
4:10
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
What was described
used a diffusion-transformer architecture (diffusion + transformer) to predict future video frames and robot actions
Said nearby on the same subject
“I use a simple formula called story action, story action, story action, meaning I tell a story and then I give you the action you can do.”
JA
Jon Acuff
The Think Media Podcast ·
“We're probably going to be bringing someone onto our team for social video editing and someone else to help us with this building, this proprietary data model to predict crowd flows”
♫
Said on This Week in Startups
“We’re predicting the next optimal pedagogical action… sometimes text, sometimes silence, sometimes draw a diagram… difficult for language models”
♫
Said on This Week in Startups
Moments like this, every Monday
The strongest new moments on your topics, in the speakers' own words. Free for three topics.
No card. No password. One-click unsubscribe.
Track a competitor?
Pro watches up to 25 names or phrases and emails you the morning after they are said.
See Pro, $29 a monthVerbatim quotes a short extract and links to the episode; the episode belongs to its publisher. Remove a show.