I agree, but I don't think there is a substantial number of software shops that use relevant degrees as a hard filter, nor is there a significant number of non-degree holders bursting with realized CS talent that the field is somehow blind to.
I took the intended sentiment to be that of finding talent among non-degree holders that won't necessarily present itself in a typical CS interview scenario. Thus we need a way to identify future CS talent that is merely in need of training and mentoring. And so the problem reduces to, in my opinion, to finding people who are smart and unusually self-motivated. But there is just no good way of doing this.
My opinion: applying one-size-fits-all formulaic interviewing to your candidates excludes unconventional talent. Asking personal questions and following through on them unearths the smart and self-motivated talent. As an unconventional CS talent myself, I've had the most success with this approach on both sides of the interview process.
What do I mean by personal questions? My favorite interview so far left all the academic quizzing aside and asked me to talk in-depth about a particular real-world project I worked on. What patterns did I use? How did it interact with the database? How did I model the problem? This was not something that would be easily googled and it let me convey how I think about software on my own terms. Then it was up to the interviewer if my approach to software would be useful to them.
I took the intended sentiment to be that of finding talent among non-degree holders that won't necessarily present itself in a typical CS interview scenario. Thus we need a way to identify future CS talent that is merely in need of training and mentoring. And so the problem reduces to, in my opinion, to finding people who are smart and unusually self-motivated. But there is just no good way of doing this.