From here they define the $\ell$-Diffie Hellman inversion problem as:
Given $g^{a},g^{a^2}\ldots,g^{a^{\ell}} \in G$, compute $g^{a^{-1}}$
Would this problem become easy if the generator $g$ is also known?
From here they define the $\ell$-Diffie Hellman inversion problem as:
Given $g^{a},g^{a^2}\ldots,g^{a^{\ell}} \in G$, compute $g^{a^{-1}}$
Would this problem become easy if the generator $g$ is also known?
No.
The stated $l$-DHI problem is believed to remain hard when $g$ is known.
Actually, in the quoted page, it is assumed that $g$ is known throughout. This is obvious in particular in the sections on DLP and CDH.
No. The generator $g$ is a public parameter of the group $G$. You cannot perform a Diffie-Hellman handshake unless both parties agree on the generator (as well as any other parameter that defines $G$), so naturally any variation of the Diffie-Hellman problem must, by definition, assume the same thing.
Do you know about the paper Variations of Diffie-Hellman Problem (PDF)? The problem you stated is the generalization of the inversion problem stated in that paper. You can use their technique to prove the relation.