8 weeks is impressive for something like that, and it goes to show just how powerful our off-the-shelf tools have become.
I think it's also a bit scary, because 8 weeks is very little time for testing, tuning, and validation of something as opaque as a machine learning model. If it worked right the first time, that's great. But there is still a lot of inherent uncertainty in ML projects. Decision makers need to take that uncertainty into account when planning.
That, or, the 8 weeks only covers the final training runs and the implementation/deployment, and doesn't include time spent developing and tuning proof-of-concept prototype models.
Comments
8 weeks is impressive for something like that, and it goes to show just how powerful our off-the-shelf tools have become.
I think it's also a bit scary, because 8 weeks is very little time for testing, tuning, and validation of something as opaque as a machine learning model. If it worked right the first time, that's great. But there is still a lot of inherent uncertainty in ML projects. Decision makers need to take that uncertainty into account when planning.
That, or, the 8 weeks only covers the final training runs and the implementation/deployment, and doesn't include time spent developing and tuning proof-of-concept prototype models.
In 2022 you test live in production lol
It’s not life or death though. It’s just giving some useful boilerplate that you’ll have to touch to make useful. They can learn as they go.