Zhipeng Yan

ML Infra@Dyna

[Zhipeng Yan @ZhipengYan7]

@JasonMa2020 I think they meant this not in a bad way, you can tell by how they went through the tech stack in the article. It calls for a reset on the expectation of demos after a whole year of non-stop stream of them.

[Justin Kerr @justkerrding]

Almost all bimanual robots include wrist cameras. They add bulk, cost, and complexity, but are critical for performance. Are they really needed?

Excited to finally share EyeRobot 2.0, where we use active gaze from an ego stereo camera instead 🧵 (with @kushtimusprime)

[Zhipeng Yan @ZhipengYan7]

This reminds me of the first hub-to-hub autonomous truck run by TuSimple in 2018, when the majority of the AV industry was still demoing on highway segments or controlled areas. Everybody knows that demos don’t generate commercial value. But what does is not always obvious. By deploying early to the customers, collecting feedback, and solving real problems with frontier research, we have found our path to product-market-fit and this is just the beginning.

We are delivering real world impact and building a cracked team with like-minded superstars. Come join us on this journey!

[Dyna Robotics @DynaRobotics]

We are releasing Dyna-2.1, the first Physical Agent that achieves reliable super long-horizon whole-body autonomy. It combines our brand-new semi-humanoid hardware with an agentic system built around Dyna-2 to handle ultra-long real-world workflows.

Here is an uncut footage of Dyna-2.1 completing an entire hour-long laundry room workflow, just like a human does.

[isabel @artficialisabel]

the openai huggingface incident, from an agents pov.

(part 1)

[Monica Xie @immonicax]

I’m at #Actuate26 today and tmr. If you want to talk about data, partnership, deployment, my DM is open!

Ping @ZhipengYan7 if you want to scale the infra to train the next 10-100x robot data 🙌

[Zhipeng Yan @ZhipengYan7]

@yangcheng Training jobs are not cross cloud and metrics are very straightforward; cluster level metrics such as GPU util can be collected through services like Datadog.

[Zhipeng Yan @ZhipengYan7]

The most amazing part of the story is, this was done by a team of less than 5 ppl within a span of several months, without dropping support on other parts of the company. If you want to be part of a cracked team, DM me or check https://t.co/ZkCAws3lr8

[Dyna Robotics @DynaRobotics]

When people talk about robotics, they usually talk about models, data, or hardware. Few people talk about the infrastructure that lets you iterate on all three quickly. Today we're publishing how we trained Dyna-2 on over 1,000,000 hours of egocentric video, repeatably. At this scale, most of what worked at ten thousand hours did not hold up:

• ingestion throughput was capped at 14,000 episode-hours per week — a million hours would have taken over a year
• building a training manifest took 48 hours before a run could even start
• reading a petabyte from cloud storage during training left GPUs exposed to latency and packet loss

🧵

[Zhipeng Yan @ZhipengYan7]

I’ll attend #actuate26 with the #dyna team. If you are building in the physical AI domain, let’s chat!

[Jason Ma @JasonMa2020]

I will be at #Actuate26 giving a talk on the model and infrastructure behind dyna-2. Excited to chat with everyone about scaling robot foundation models and deploying them in the real world! Please reach out if you'd like to chat!

[Dyna Robotics @DynaRobotics]

Today we are introducing Dyna-2, a world-action model pre-trained on one million hours of human video. At this scale, for the first time, we discovered several new scaling laws:

• world-action models exhibit scaling law on human data across four orders of magnitude, from 1000 to 1,000,000 hours,
• this human data scaling law implied a scaling law on never seen robot data,
• both data and objective matter; world modeling and scaling on video data are essential for cross-embodiment scaling transfer to emerge

🧵

[Zhipeng Yan @ZhipengYan7]

@DynaRobotics Exciting! An industry-first scaling law connecting human video to general robot capabilities.

Video pre-training 🤝 robotics foundation models.

More on how we made this possible on the infra side — stay tuned!

[Dyna Robotics @DynaRobotics]

A new breakthrough is coming. Stay tuned.

[Zhipeng Yan @ZhipengYan7]

Same here at Dyna. The demand is huge!

[Jon Miller Schwartz @JonMSchwartz]

I'm honestly shocked by the demand we're seeing for our robots. We're now having 30min intro calls with customers, and signing contracts right after. Soon, it'll happen directly though our website.

Two thoughts:
1) Seeing (in the real world) is believing. The more you can show, the more trust you'll be given.
2) There are going to be so many robots.

[avi @avizurlo]

https://t.co/8aO7KXyKpe

[Zhipeng Yan @ZhipengYan7]

Can I get a refund

[Will Smith @WillSmith2real]

This is getting out of hand!

- Will Smith