The funding and customers, as the company stated them
On October 10 Latent Space interviewed Evan Beard, co-founder and CEO of the industrial-robot company Standard Bots, and Leif Jentoft, its head of AI.
The article says Standard Bots claims to be America's largest AI-native industrial robot manufacturer. It recently raised $200 million at a $1 billion valuation in a Series C led by General Catalyst and RoboStrategy. The article says customers include NASA, Amazon, and Lockheed Martin. Those figures and names are the company's account in this interview, not an independent audit.
[1]
The model sees the part, and code moves the arm
Beard says the company has a shared base model that customers adapt through demonstrations and fine-tuning. Its largest model is in the low billions of parameters, which he says is not large by frontier-lab standards.
Jentoft uses machine tending as the example. A user tells the robot which parts to look for. The model handles perception: locating and identifying the parts. Conventional programming handles the motion and the cell logic. He says the backbone for this task is trained on over a billion images, so the robot can tell lighting, materials, and background apart.
Beard says training happens in the cloud and inference runs locally. The company keeps tasks short-horizon to meet cycle time and reliability. Parts of a workflow that do not need learned behavior stay in conventional programming.
[1]Not every deployment sends data back
Jentoft says many defense customers deploy in air-gapped environments, where nothing comes back. Customers who are not air-gapped may trade failure records and human corrections for later performance. That is his account.
The company already offers StandardOS APIs and SDKs to outside developers, including bringing their own models. Beard says developers still have to write their own integration code today. Making collection, training, and deployment easier is a plan, not a finished feature.
[1]