The algorithm he's developed is, we're led to believe, potentially the seed for world-changing, billion dollar business possibilities. The reason he's been offered $10M for it is because the potential buyer believes they can realize that potential, and that without them its inventor can't.
The question he's faced with is not "Take ten mill or start a company", it's weighing up, given that this is a multibillion dollar idea, can you, with a little seed funding, realize at least $10M of value out of it by going it alone?
Saying "take the $10M and then start a company anyway" is ignoring the rather crucial step "*come up with an equally valuable idea which you haven't sold the rights to"...