i dont really understand the revenue model here. i also dont understand how there’s going to be enough computational power to do LLM shit for all windows users all the time? this sounds bad for the environment.
Running a pre trained model is much cheaper than training one. But I’d imagine in this case you’ll be sending it over to Microsoft Servers, so they can keep track of everything you ever search so they can better advertise to you.
i dont really understand the revenue model here. i also dont understand how there’s going to be enough computational power to do LLM shit for all windows users all the time? this sounds bad for the environment.
Running a pre trained model is much cheaper than training one. But I’d imagine in this case you’ll be sending it over to Microsoft Servers, so they can keep track of everything you ever search so they can better advertise to you.