I do not think you can do a computer science Ph.D. without a strong quantitative undergraduate degree, ie computer science, math, physics. The Ph.D. application

I do not think you can do a computer science Ph.D. without a strong quantitative undergraduate degree, ie computer science, math, physics. The Ph.D. application is a very clubby process, and it is more or less who you know, who you have worked for, where you interned (that is Google or Facebook). It is not really a five year school so much as a five year employment with minimal rights to the employee, and a lot of risk to the employer since your boss (advisor) has to commit to someone for 5 - 6 years without a prior internship relationship. Framed in this way, you see how difficult it is for you to get in a program now, it is basically impossible. Your best bet is to go through a post-bac computer science program (UPenn and Columbia both offer them), the program themselves are cash grabs for the university, but while in the program you can work for a professor and develop a relationship there, and then he/she will give you pointers. If you're in the program sometimes you can transfer to a related program that offers more math/theoretical comp-sci curriculum as well. You can become a "machine learning" scientist after this program, Ph.D. is not required. Honestly, I would say you're at least three years away from asking if you're qualified for a Ph.D.