It is ambiguous to me that article has talked about benefits of observability which can come from Model monitoring. Like for monitoring we can set alert/threshold on metric or drift threshold on data or failure alert on model, etc. I hope explainability will not come under observability since it comes as by product of model and we don't have to observe anything for this. It would be helpful if you can provide real world examples for observability.
Thanks for your comment! I hear what you are saying, I think to be fair for a lot of organisations it is true that almost all these points fall under the umbrella of "Model Monitoring". I would argue that as with many things in AI/ML definitions different throughout different organisations, resulting in different ven diagrams with either overlapping or clearly separate entities of scope. So apply it as you hope or what feels best for you :)
This article is actually based on an open discussion about these not entirely clear lines. I would say it's always up for discussion, but I think when we do try to make a distinction (and also use the newer term of ML Observability) it is the intent that really sets the two apart.
It is ambiguous to me that article has talked about benefits of observability which can come from Model monitoring. Like for monitoring we can set alert/threshold on metric or drift threshold on data or failure alert on model, etc. I hope explainability will not come under observability since it comes as by product of model and we don't have to observe anything for this. It would be helpful if you can provide real world examples for observability.
Thanks for your comment! I hear what you are saying, I think to be fair for a lot of organisations it is true that almost all these points fall under the umbrella of "Model Monitoring". I would argue that as with many things in AI/ML definitions different throughout different organisations, resulting in different ven diagrams with either overlapping or clearly separate entities of scope. So apply it as you hope or what feels best for you :)
This article is actually based on an open discussion about these not entirely clear lines. I would say it's always up for discussion, but I think when we do try to make a distinction (and also use the newer term of ML Observability) it is the intent that really sets the two apart.
Thanks for the reply. I totally agree about the varying definition of each pillars. Thanks for sharing your thoughts through these articles.