Uploaded December 2021 | Updated September 2026, 7 hours ago
Gopher is DeepMind's new large language model. With 280 billion parameters, it's larger than GPT-3. It gets state-of-the-art (SOTA) results in around 100 tasks. The best part of the Gopher paper is the wide and deep analysis on what scales with model size, performance on many tasks like language modeling, Q&A, logic tasks, and also a look into the ethics of large language models. I think the Gopher paper is a great example of what scaling research in Machine Learning should be about.
Gopher Blog Post: deepmind.com/blog/article/language-modelling-at-scale
Gopher Paper: storage.googleapis.com/deepmind-media/research/language-research/Training%20Gopher.pdf
Resource on Transformers: lilianweng.github.io/lil-log/2018/06/24/attention-attention.html#full-architecture
Gopher is DeepMind's new large language model. With 280 billion parameters, it's larger than GPT-3. It gets state-of-the-art (SOTA) results in around 100 tasks. The best part of the Gopher paper is the wide and deep analysis on what scales with model size, performance on many tasks like language modeling, Q&A, logic tasks, and also a look into the ethics of large language models. I think the Gopher paper is a great example of what scaling research in Machine Learning should be about.
Gopher Blog Post: deepmind.com/blog/article/language-modelling-at-scale
Gopher Paper: storage.googleapis.com/deepmind-media/research/language-research/Training%20Gopher.pdf
Resource on Transformers: lilianweng.github.io/lil-log/2018/06/24/attention-attention.html#full-architecture
![ML Research Idea [Zero to Paper]](https://i.ytimg.com/vi/nSEb6w_BBqE/mqdefault.jpg "ML Research Idea [Zero to Paper]
This episode (part 2) of Zero to Paper covers the idea we will be working on: Text to Goal, or TTG for short. TTG aims to translate natural language into a reward function for a Reinforcement Learning problem. This research project aims to leverage recent advancements in NLP and Computer Vision to make reward functions easier and more natural to craft.
I also briefly touch on how I come up with ideas an what is important to me. I hope this will be helpful to anyone wonder how to do ML research!
Zero to Paper playlist: https://www.youtube.com/watch?v=74fKCvr5n5o&list=PL_49VD9KwQ_ONxENRk11jFEI3_pqAwaug
RL Theory series if you want to get into RL: https://www.youtube.com/watch?v=1OI0uuz9jkI&list=PL_49VD9KwQ_OML1Knh-Yb7FUFkhTLS0jL")
![Learning Language Through Games [Zero to Paper]](https://i.ytimg.com/vi/qY0nCUeQlXI/mqdefault.jpg "Learning Language Through Games [Zero to Paper]
Lets talk about natural language in Reinforcement Learning. Its also a form of language grounding because models are trained to learn connections between language and a non-text environments. I think language-conditioned RL is the way forward for making more efficient, general AI training. Let me know what you think of this combination between NLP and RL in the comments!
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
Papers covered:
https://arxiv.org/pdf/2005.09382.pdf
https://arxiv.org/pdf/1902.07742.pdf
https://arxiv.org/pdf/1806.01946.pdf")
