Frontiers of SE

Lecture 10-2
Bart Massey 2015-12-02

Heads Up

• Today we'll talk about frontiers of SE

• The more things change...

• SE can be done in weird ways

  

Frontiers Of SE: Machine Learning

• Way to build SW systems without programming

• Supposedly analogous to human / animal learning

• Several successes

  • Backgammon play
  • Spam filtering
  • Speech recognition

  • Big data

    

Frontiers Of SE: Massive Parallelism

• Way to bring much hard-to-use computing power to problems

• Nothing new, but scaled a lot

• Several successes

  • Go play
  • Climate modeling
  • Web services

  • Big data

    

Frontiers of SE: Proofs Of Software Security

• Only sure way to defend a human adversary

• Typically limited to specific properties

• Several successes

  • Automatic software validators e.g. Coverity

  • SEL4 kernel

    

Discussion

• SE is a huge field, but we are

  • Making progress: things are better than 20 years ago
  • Learning how to think about software
  • Carving up the field into more manageable chunks

• Hard to say what the future will hold, but kind of neat