A few answers (Jill's, Marius's, and Roy's) mention the fact that is necessary to choose the correct Python interpreter to make Pylance function properly. I would like to add the fact that this is still necessary to do when using a Jupyter Notebook with the correct Python kernel already chosen.
It is counterintuitive to choose both Python interpreter and notebook's Python kernel to make things work. It is even more counterintuitive considering the fact that Python interpreter's button (on the left bottom of the screen, on status bar) does not necessarily appear when a Jupyter Notebook is open, but when a Python script is open. For instance, in this screenshot, we see the little line under Scikit-learn's import, indicating a problem with the import (even though the import is successful). However, the correct Python kernel, with Scikit-learn installed, is already chosen. Only opening a Python script we notice that the Python interpreter is the reason of this behavior, because a wrong one is chosen, without Scikit-learn. In some sense, one could think that the reason behind this was a problem with the Python kernel or the Conda environment (it is common to experience this kind of problem when experimenting with Jupyter Notebook and Jupyter Lab). I hope this answer may help those who are searching for solving this problem in the specific context of Jupyter Notebooks inside VS Code. They could ignore the other answers because they could think it is not the case for them.