Teaching a Computer How to Read: A Summer of Experiential Learning

This past summer I interned at BenchSci, a Toronto-based tech startup, for four months. My job was to teach a computer how to recognize letters in an image. It was a great experience, and I learned a lot about machine learning, computer vision, and what it is like to be a part of a startup.

Sliding Window
This is an example of an image I ran through (part of) my program. It identifies where the text is in an image.

This internship fulfills my experiential learning requirement, which is a requirement every Quest student has. Essentially, one must do something in the “real world” that is specifically not academic, and that relates to one’s Question. My Question, as my constant readers will know, if “How should we create artificial general intelligence?”. Understanding the current state of the art in the machine learning field helps me to understand my Question in more depth, and thus was a particularly good fit for my experiential learning.

This is a (rough) schematic of a convolution neural network, which was was of the key technologies I worked with.
This is a (rough) schematic of a convolution neural network, which was was of the key technologies I worked with.

A few (non-technical) things I learned while interning:

  1. ALWAYS back up your work. Do it.
  2.  Just because Google has a team of experts that can do something does not mean that you alone can do the same thing.
  3. Often, the hardest part of something is not the part you expected it to be.
  4. Often, the hardest part of something is not the exciting part.
  5. Be ready to backtrack; don’t commit to any one approach.
  6. Machine learning is less advanced (powerful/developed) than I thought.
  7. Machine learning is more advanced (powerful/developed) than most people think.
  8. It can hard to work alone.
  9. It can be hard to work in teams.
  10. Break projects up into manageable sizes.

Needless to say, it was a great experience, and I learned a lot.

 

Keep reading, keep learning, keep thinking.

Sapere aude! 

Daniel

Leave a Reply