Anthropic发布模型硬件标准，让AI直接操控实验设备

Now, experiments that took weeks take days. That can enable scientific breakthroughs [music] much quicker.

For the past year, we've been putting AI to work inside labs and manufacturing facilities. We kept hitting the same wall. There's no common way to connect a model to physical equipment. The Model Hardware Standard [music] changes that.

MHS gives any AI model one standard way to connect with and operate devices. If you wanted your instruments [music] to work together, say a camera and a stage of a microscope, you had to build a custom software integration between them. Every device you add complicates the picture. For complicated experiments, that's weeks of work. With MHS, each device connects once through one interface.

[music]

The setup is accelerated by the agent. Any device that speaks MHS can talk to anything else that does. Devices communicate at bare metal speed. The agent has access to this context and can control operations.

Here we're seeing Claude controlling a Leica microscope on its own. It's focusing the microscope, looking for a bacteria, and deciding what to capture next. At Genentech, a scientist designed an experiment in a PDF, dropped that PDF into Claude, then Claude autonomously executed that experiment. [music] When things went wrong, it was able to recover even overnight.

The iteration cycle is much faster and the focus for me as a scientist shifts to actual scientific questions. Once everything is connected, running an experiment looks like you ask Claude to start [music] it. Here we're looking at images of neurons in the brain. With MHS, I can instruct [music] the microscope to move around, go deeper, or even take side views, all in real time.

If you can test [music] hypotheses faster, you could create general technologies faster, [music] like finding new materials, also solving important scientific problems. This is how a century of progress can compress into a decade.