How can organizations ensure that their use of AI aligns with ethical and responsible practices?

Part of what makes modern neural networks so powerful is their ability to recognize and extrapolate complex patterns and tasks from massive input data, but this often means that AI is a blackbox where it is hard to decipher why the neural network made the decision it did. This kind of justification is important to many high-profile decisions. One classic examples where this went wrong, are using AI to recommend whether prisoners are likely to relapse, but we also see it where GTP-3 generated flawed medical recommendations or historical inaccuracies. There is some interesting reseeach into how can try to detect biased training data. I recommend both these articles for further reading on bias in NLP: Detecting and Mitigating Bias in Natural Language Processing https://www.brookings.edu/resear... A higher-level summary. A Survey on Gender-bias in Natural Language Processing https://arxiv.org/abs/2112.14168 - a in-depth academic article comparing detailed methods to detect bias.