Saturday, November 14, 2015

Google's TensorFlow and "mathematical tricks"

TensorFlow is a new open source software library for machine learning distributed by Google. In some ways, this could be seen as a competitor to BlueMix (though much less user friendly). Erik Mueller, who worked on the original Watson Jeopardy system (and has a vested interest in AI with his new company Symbolic AI), just wrote a brief review of TensorFLow for Wired.

Google’s TensorFlow Alone Will Not Revolutionize AI

Unfortunately, it's not really a review of TensorFlow itself, but rather makes a general point against statistical approaches, which I basically agree with, but the argument requires a much more comprehensive treatment.

Some good quotes from the article:

  • "I think [TensorFlow] will focus our attention on experimenting with mathematical tricks, rather than on understanding human thought processes."
  • "I’d rather see us design AI systems that are understandable and communicative."

No comments:

Putting the Linguistics into Kaggle Competitions

In the spirit of Dr. Emily Bender’s NAACL blog post Putting the Linguistics in Computational Linguistics , I want to apply some of her thou...