After years of AI meaning building your own models trained on proprietary data, one of the hardest things for firms to grasp is that their own data is much less immediately useful for pre-trained models The value in adding data in a use case will require experimentation to learn

→ View original post on X — @emollick