You would probably want to tag several pods with physical trackers and then watch those whales with the satellite. After training it on that data for long enough you could hypothetically make software that spots whales surfacing and then sees what direction they are likely going and it can then start to guess where they think that same pod will pop up later using the data it gathered from those pods being tracked physically with trackers from before.

Yes, we could hypothetically do the things we are already doing for tracking how tagged whales behave in addition to using spottings as a way to estimate numbers of whales.

What you aren’t understanding is the scope of the ocean and that being able to do these things in small areas or with a limited number of tagged animals is vastly different than doing it for the entire ocean. That is why they use crowdsourced techniques like spotting!