Collections of papers and articles that I've spotted this week that seem interesting. Comments on some of them.




  • No excuses for moral realism
    • Most people would accept things like 'He was drunk' or 'She didn't know she was feeding bleach to her baby' as excuses, but not things like 'He didn't know that murder was wrong'.
    • The author uses this as an argument against moral realism
    • Here at Nintil we of course go edgy and say that moral ignorance is exculpatory. We used this argument against Bryan Caplan b4 it was cool and it appeared on journals.


Artificial intelligence

  • We need to go deeper, a practical guide to Tensorflow and Inception
  • Attacking machine learning with adversarial examples
    • The more you study ML, the more fascinating it is
    • But the more removed from fully general AI it seems. Adversarial examples just reveal the fact that neural networks still do not know what they are seeing (in the sense that AlphaGo does not know it is playing Go). Neural nets are a clever form of approximating almost any function, but the way they do it seems to be brittle. A human would never fall for adversarial examples as depicted in the literature about adversarial examples.
    • This will be probably fixed in the coming years