HackerPapers

arXiv:1803.10122cs.LG

World Models

David Ha, Jürgen Schmidhuber

Abstract

We explore building generative neural network models of popular reinforcement learning environments. Our world model can be trained quickly in an unsupervised manner to learn a compressed spatial and temporal representation of the environment. By using features extracted from the world model as inputs to an agent, we can train a very compact and simple policy that can solve the required task. We can even train our agent entirely inside of its own hallucinated dream generated by its world model, and transfer this policy back into the actual environment. An interactive version of this paper is available at https://worldmodels.github.io/

arXiv comments
Journal reference
DOI
doi:10.5281/zenodo.1207631
Subjects
cs.LGMachine Learning stat.MLMachine Learning (Statistics)

On Hacker News

Linked by 10 identifiable commenters in 11 comments across 11 threads, from Mar 2018 to Feb 2026. Found in 1 HN submission from Jun 2018 to Jun 2018. Open submission discussions.

Comments linking this paper, per year.

Discussed in comments11 threads

  1. The Waymo World Model 1 comment

    • > Suddenly all this focus on world models by Deep mind starts to make sense Google's been thinking about world models since at least 2018: https://arxiv.org/abs/1803.10122

      xnxView on HNParent context

  2. Project Genie: Experimenting with infinite, interactive worlds 1 comment

    • Didn’t the original world models paper do some training in latent space? (Edit: yes[1]) I think robots imagining the next step (in latent space) will be useful. It’s useful for people. A great way to validate that a robot is properly imagining the future is to make that latent…

      SequoiaHopeView on HNParent context

  3. 'World Models,' an old idea in AI, mount a comeback 1 comment

    • They also don't mention the famous paper by Ha & Schmidhuber (https://arxiv.org/abs/1803.10122). The worst part is that they namedrop many other tangentially related and/or outright fraudulent "ai experts" like Hinton, Bengio, and LeCun.

      esrhView on HNParent context

  4. Getty Images v. Stability AI – Complaint 1 comment

    • > it seems that this merely exposes the normal operation. Insufficient images of that person prevented it from abstracting the person components from the background yes my point was that this total failure to abstract (or slice or average or whatever it is that it usually seems…

      d110af5ccfView on HNParent context

  5. More Protein Folding Progress – What’s It Mean? 1 comment

    • Your description kind of reminds me of the 2018 World Models publication. Are you familiar with that work and/or am I off base? (https://worldmodels.github.io/) (https://arxiv.org/abs/1803.10122)

      d110af5ccfView on HNParent context

  6. DeepMind says reinforcement learning is ‘enough’ to reach general AI 1 comment

    • That may be true, but that's not to simulate. "Dreaming" is used to "simulate", and it's only a hack, in humans probably not even explicitly programmed. The hack is to have the simulator itself ... also be a learned system, and not to simulate the world, because you don't act on…

      candiodariView on HNParent context

  7. The revolution of machine learning has been exaggerated 1 comment

    • That seems a fairly specific meaning of the term that may not be in wider use, see eg: https://arxiv.org/abs/1803.10122

      nmcaView on HNParent context

  8. Ask HN: 2018 Summer Reading List? 1 comment

    • Not a book but if you’re ML curious, I enjoyed reading the World Models paper. It was the first ML paper I read all the way through, and I found it approachable but still technically informative. I later learned the author has been in the AI game for decades, which explains his…

      TaylorAlexanderView on HNParent context

  9. World Models with a Reservoir Computing Twist 1 comment

    • Background information about Reservoir Computing: https://en.wikipedia.org/wiki/Reservoir_computing World Models paper: https://arxiv.org/pdf/1803.10122.pdf https://worldmodels.github.io Previous HN Discussion: https://news.ycombinator.com/item?id=16860247…

      baylearnView on HNParent context

  10. A step-by-step guide to the “World Models” AI paper 1 comment

    • Hi, I'm one of the authors of this paper (https://arxiv.org/abs/1803.10122, https://worldmodels.github.io). Happy to answer any questions you may have.

      hardmaruView on HNParent context

  11. World Models 1 comment

    • Link to the neural net generated Doom level demo: https://worldmodels.github.io/doomrnn/ Original article: https://worldmodels.github.io/ PDF version of the paper: https://arxiv.org/abs/1803.10122

      wei_jokView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.