News for hackers. Stuff that matters.
Re:Mental time travel: Hierarchical memory for reinforcement... by andreyk on Friday May 27, 2022 @02:57AM (#31526762)
Fun coincidence, I read this paper as part of my AI research just recently. It's actually quite neat! Mostly because there is really not much like it in the sense that it focuses on exploring a memory mechanism for RL agents - most of the time people just stack frames or slap on an RNN and call it a day. The set of tasks evaluated on is cool too.
Re:Mental time travel: Hierarchical memory for reinforcement... by totetsu on Friday May 27, 2022 @01:12AM (#31526203)
> . HCAM stores memories by dividing the past into chunks, and recalls by first performing high-level attention over coarse summaries of the chunks, and then performing detailed attention within only the most relevant chunks.

This sounds just like the techniques the human 'memory athletics' gurus teach.

Re:Mental time travel: Hierarchical memory for reinforcement... by Gravyness on Friday May 27, 2022 @02:43AM (#31526688)
The original gurus or the fake ones after everyone realized it was profitable?
Re:Mental time travel: Hierarchical memory for reinforcement... by totetsu on Friday May 27, 2022 @03:57AM (#31527115)
I think I first heard about chunking from one of the real ones trying to spin a profit.
Re:Mental time travel: Hierarchical memory for reinforcement... by MoritzWall on Friday May 27, 2022 @02:40AM (#31526663)
> or reasoning over memory

Where is the "reasoning" in this? sounds like pattern recognition to me.

Re:Mental time travel: Hierarchical memory for reinforcement... by mach1ne on Friday May 27, 2022 @02:01AM (#31526454)
The authors are seemingly falling into the fallacy of replicating the human experience, which simply leads to ultimately inefficient Alt Intelligence. While an interesting idea which could have a couple very rare useful use-cases, this seems like another of the million [Deep Learning + a minor wacky idea] research papers.
Re:Mental time travel: Hierarchical memory for reinforcement... by px43 on Friday May 27, 2022 @10:59AM (#31530395)
I have not read the paper, but for some high level context, it's written by a team of researchers at DeepMind, and was presented at NeurIPS, so I'm going to go out on a limb and give the authors the benefit of the doubt that they probabally have some idea of what they're talking about.
Re:Mental time travel: Hierarchical memory for reinforcement... by The_Amp_Walrus on Friday May 27, 2022 @06:34AM (#31527971)
Can you explain why it's a mistake in this particular case?