Hi ,
Most of the time, if you're trying to predict the weather, yesterday's conditions provide a good enough estimate.
If it was 30C+ yesterday, chances are you won't need your winter coat when you go outside today.
But if you're a farmer trying to determine the optimal time to harvest crops, a precise weather
forecast could be the difference between a successful year and ruin.
In the first scenario, the value of an accurate estimate is low, while in the second, it's extremely high.
Here's the thing...
Every model costs something to train and deploy. Even if it's just your time. ChatGPT, for example, has an estimated running cost of $100k per day.
And that's OK - provided the value produced is greater than the cost.
But if it's not, then perhaps you shouldn't build the model at all.
Sometimes, it's worthwhile
building an elaborate AI-powered weather model. But often, just looking out the window is good enough.
Talk again soon,
Dr Genevieve Hayes.
