Indeed, the definition of a weak entity is related to its identity. Peter Chen, the inventor of ERD has defined it as follows:
If relationship relations are used for identifying entities, we call it a weak entity relation
In other words, if all the own attributes of an entity are sufficient to uniquely identify it, it's a regular entity, but if an attribute of another related entity is needed (e.g. a primary key of another entity), it's weak.
If you give a weak entity its own identity by generating a surrogate key (artificial key) it would no longer need the other entity for the unique identification. So it would be by definition a regular entity. By the way, the weak relationship would also have to be changed to a regular one.
An example
Let's take the typical example of a Purchase Order and a Purchase Order Line : several lines with the same line-number and the same material could exist, and the purchase order id would be required to distinguish them. The primary key would typically be composite (purchase order + line number). So the Purchase Order Line is definitively a weak entity.
If you replace the primary key with a surrogate key, the identity of this order line is ensured by the new primary key, independently of the purchase order and the line number. From a relational perspective this is a valid approach and a different reality. You suddenly could imagine a line to exist without an order (even if it doesn't make much sense from a business perspective).
This would not change the constraints or the business rules. Typically, you'd maintain a total (mandatory) relationship to the order. But could imagine reassigning purchase orders, moving lines in the sequence, as flexibly and naturally as you would change other attributes such the material ordered or the desired quantity.