**Overview** Part of our fortnightly series of paper reading sessions. We're reading BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (Devlin et al., 2018). It takes the pre-training idea from the GPT paper we read last time and makes it bidirectional: instead of predicting the next word from the words before it, the model hides words and fills them in using the text on both sides. Read it before you come: [https://arxiv.org/abs/1810.04805](https://arxiv.org/abs/1810.04805) Check out the journal club channel on our applicant Discord server: [https://discord.gg/y6AXyzbXUX](https://discord.gg/y6AXyzbXUX) **How the session works** This is not a walkthrough. We'll split into small groups to work through a set of questions, then regroup for answers and an open discussion. **Who is this for** Experienced software engineers. You don't need a research background, though you'll get more out of the evening if you're comfortable with the basics of neural networks. You don't need to have come to the earlier sessions. **What to bring** The paper, read beforehand. You won't need a laptop.

Description supplied by the event organiser.