fahad (@maddestfhd) profile picture
earth
Joined October 2022
problems working with ai as models get better and cheaper:
1
54
the most efficient model could be found through benchmarks running of every given model for a negligible difference in cost_per_task but in the real world inference the spread will be always there making it harder to determine which is the best model to use to save cost
1
9
and time, so there always will be room for humans to decide about what to choose for long hours of sittings, and waste lot of money in underestimating or overestimating model's effeciency in completion of the task
5