Felix Hill: #GPT3 from @OpenAI showed an emergent ability in large neural language models for rapidly acquiring and using new words.
We develop an agent that does this in a simulated 3D environment.
5 replies, 395 likes
DeepMind: Recent large neural language models have shown an emergent ability to rapidly acquire and use new words. Our researchers developed an agent that does this in a simulated 3D environment.
2 replies, 284 likes
Connor Leahy: https://arxiv.org/abs/2009.01719
This paper (by @FelixHill84 et al) is really an "It's all coming together" moment for @DeepMind I feel.
Let me try to describe my takeaways from my first readthrough.
5 replies, 249 likes
Shawn Presser: DeepMind trained a Transformer in a 3D environment to recognize and manipulate specific objects it learns about on the fly. It has an external memory, meaning it can remember new objects with no training.
"Grounded Language Learning Fast and Slow"
arxiv: https://arxiv.org/abs/2009.01719 https://t.co/9WQ7sNNSq8
2 replies, 88 likes
Edgar Xi: Deepmind is literally playing 3D chess
0 replies, 4 likes
Andrey Kurenkov 🤖: Very cool to see this!
Interesting to see the inspiration from cognitive science and human experiments in how this AI agent was built, and awesome to see the results of that application being pretty strong.
0 replies, 3 likes
Hip Priest: Nifty - NLP language acquisition model...
0 replies, 1 likes
Found on Sep 04 2020 at https://arxiv.org/pdf/2009.01719.pdf