Uploaded February 2020 | Updated September 2026, 8 hours ago
In this video we work on the first part of making an AI that can answer questions. We use DistilBERT, transformer based NLP model pretrained on the Stanford SQUAD dataset to answer questions. In part 1 we load the model and use it to make answer predictions.
Relevant Links:
GitHub code: github.com/ejmejm/AskMeAnything
Google Colab: colab.research.google.com/drive/1DbOL_kNsMct1_9qCrMfnaiyTDGNfBkq3
Simple Transformers: github.com/ThilinaRajapakse/simpletransformers
DistilBERT Paper: arxiv.org/abs/1910.01108
In this video we work on the first part of making an AI that can answer questions. We use DistilBERT, transformer based NLP model pretrained on the Stanford SQUAD dataset to answer questions. In part 1 we load the model and use it to make answer predictions.
Relevant Links:
GitHub code: github.com/ejmejm/AskMeAnything
Google Colab: colab.research.google.com/drive/1DbOL_kNsMct1_9qCrMfnaiyTDGNfBkq3
Simple Transformers: github.com/ThilinaRajapakse/simpletransformers
DistilBERT Paper: arxiv.org/abs/1910.01108
![Self-Supervised RL - Learning Without Data [Zero to Paper]](https://i.ytimg.com/vi/CDKsa06xU0o/mqdefault.jpg "Self-Supervised RL - Learning Without Data [Zero to Paper]
Inverse Reinforcement Learning with Natural Language Goals (LangGoal IRL) offers a way to do sample-efficient IRL and a way to generalize using self-supervised learning. The paper is novel and is a step forward for general AI and ML algorithms. Though it has its cons, I think it is one of the better papers out there that cover RL, IRL, NLP, and generalization.
Zero to Paper playlist: https://www.youtube.com/playlist?list=PL_49VD9KwQ_ONxENRk11jFEI3_pqAwaug
Inverse Reinforcement Learning video: https://www.youtube.com/watch?v=qo355ALvLRI
Paper covered: https://arxiv.org/pdf/2008.06924.pdf")
