The document discusses three ideas for candidate selection: using academic performance for initial shortlisting, using artificial intelligence or robots to test candidates, and using peer assessment for initial screening. The author believes academic performance is the most practical idea because test scores provide insight into a candidate's aptitude and knowledge. Using AI or robots to select candidates is seen as the most disruptive idea as it could remove human factors from the process. Peer assessment is the author's favorite idea as it allows for a large initial pool, tests candidates' knowledge and group skills, and saves time.