Savlar4 sources healthy · latest episode 15 Oct 2026
Speaker

Jason Ma

Guest on Latent Space: The AI Engineer Podcast. 268 claims across 1 episode, Oct 2026.

268claims, each timestamped
1 of 12signals on the brief
0verified changes of mind
29open predictions

Positions over time

Every pair of statements that was compared. Only a verified change counts as a change of mind; rewording, added nuance and different questions do not.

No two statements by this speaker have been compared yet.

Predictions

As stated. A window appears only where the speaker gave one.

ExpectsBySaidStatus
Training a robot only on human data is unlikely to enable it to directly execute robot actions. no date given13 Oct 2026pending
Jason Ma expects synthetic environments, synthetic data, and simulation data to become a major next frontier in robotics. no date given13 Oct 2026pending
Jason Ma expects static data to add a new dimension to robotics scaling. no date given13 Oct 2026pending
Frontier-model technology may be at a tipping point and could substantially accelerate robotics. no date given13 Oct 2026pending
Jason Ma believes that a model generating its own training data could create a powerful self-reinforcing flywheel. no date given13 Oct 2026pending
Jason Ma believes current robotics progress is only the beginning of what is yet to come, implying that further robotics progress is expected. no date given13 Oct 2026pending
People will adopt the practice of training large robotics models and distilling them into smaller models. no date given13 Oct 2026pending
Dyna Robotics aims to develop in-house the components it considers critical to overall system performance. no date given13 Oct 2026pending
Dyna Robotics is developing actuators in-house and intends to own them. no date given13 Oct 2026pending
The speaker expects legged robots to be substantially less stable, and says guaranteeing their safety would be extremely difficult. In a cluster of 3 →no date given13 Oct 2026pending
Dyna Robotics aims to make its robot a platform that anyone can use to train tasks. no date given13 Oct 2026pending
Some robotic tasks may require a dexterous, human-like hand rather than a parallel-jaw gripper. no date given13 Oct 2026pending
Without capturing human sensory dimensions beyond current data, scaling to millions of hours might still omit critical task information and prevent robots from solving some tasks. no date given13 Oct 2026pending
Robotics will scale in the same way as other domains solved by machine learning. no date given13 Oct 2026pending
Model performance will improve predictably if robotics uses an appropriate set of sensors and scales the resulting data. no date given13 Oct 2026pending
Scaling will ultimately work in robotics. no date given13 Oct 2026pending
Starting with a robot's final form factor risks producing a design that is inadequate or short-lived and must be discarded. no date given13 Oct 2026pending
An iterative robotics-development approach will reach a final robot sooner than a non-iterative approach. no date given13 Oct 2026pending
Dyna Robotics aims to scale up its robots and model and deploy them over the next six months to one year. 13 Oct 2026 – 13 Oct 202713 Oct 2026pending
Dyna Robotics aims to build a general-purpose model or platform. no date given13 Oct 2026pending
Dyna Robotics aims to deploy hundreds of robots in total next year. 1 Jan 2027 – 31 Dec 202713 Oct 2026pending
The speaker thinks a video-only world model could be powerful, yet may lack sufficient richness for robotics. no date given13 Oct 2026pending
A vision-and-tactile world model could achieve 100× the data efficiency of a vision-only model. no date given13 Oct 2026pending
Tactile sensing or another modality may carry denser information than vision. no date given13 Oct 2026pending
Predicting tactile information could reduce the vision, compute, and data needed to reach the same performance. no date given13 Oct 2026pending
The speaker believes robotics companies would outperform video-generation companies that pivot to robotics at developing the eventual solution, if that solution requires a vision-tactile or vision-tactile-audio world model. no date given13 Oct 2026pending
If video models prove insufficient for robotics, video-model companies may lack the tactile-sensing expertise that the speaker's company has. no date given13 Oct 2026pending
The speaker believed simulation was not ready in 2024, though new frontier models and internal developments might change the situation. no date given13 Oct 2026pending
Dyna Robotics aims to bring robots and intelligence into the physical world, ultimately deploying robots to thousands of customers and reaching millions of robots. no date given13 Oct 2026pending

Appearances

Claims per episode