4 ms·

Although it is a very interesting paper, and I am by no means qualified to even understand half of it, while reading up representational drift I was much more c

by mateo1 5y ago

Although it is a very interesting paper, and I am by no means qualified to even understand half of it, while reading up representational drift I was much more convinced by another (and a bit more recent) paper portraying the phenomenon as a way for the brain to switch up the physical layout of the firing neurons in a redundant population without altering the output. All that possibly for biological reasons (ie dead neurons or perhaps repairs). Saying information is stored in electrical fields just doesn't sit right with me. I have a hard time imagining them being more robust than spikes, or really usable inside the brain because of the physics involved.

shepardrtc 5y ago
Was this the paper you're speaking of? "Self-healing codes: How stable neural populations can track continually reconfiguring neural representations": https://www.pnas.org/doi/full/10.1073/pnas.2106692119 https://www.pnas.org/doi/full/10.1073/pnas.2106692119 It came out about a month ago, and I've just started reading it. But it talks about neural drift, and they actually were able to create a model for it.
- hhs 5y ago
  If interested, this is the press release for that paper: http://www.eng.cam.ac.uk/news/study-proposes-how-changes-neural-code-unlock-brain-s-inner-learning http://www.eng.cam.ac.uk/news/study-proposes-how-changes-neu... And in that release, it ends like this: "Dr O’Leary, Associate Professor in the Department of Engineering, said the study emphasises the idea that “drift” may arise from continual learning. “There is a huge unanswered challenge in artificial intelligence, namely the problem of building algorithms that can learn continually without corrupting previously learned information,” he said. “The brain manifestly achieves this, and this work is a step in the direction of finding algorithms that can do the same.”"
acrefoot 5y ago
I'm not sure exactly what you believe, but since you're talking about information storage, can I ask if you mean for working memory or for long-term memory? For long-term memory, the storage is durable and mostly static, while recall is dynamic (and may even affect the storage after further consolidation). For working memory, I imagine it's a pretty dynamic process for both storage and recall. The top-level link seems to only be about working memory, so the title of the HN post should probably be changed to avoid confusion.

by mateo1 5y ago

shepardrtc 5y ago

Was this the paper you're speaking of? "Self-healing codes: How stable neural populations can track continually reconfiguring neural representations": https://www.pnas.org/doi/full/10.1073/pnas.2106692119 https://www.pnas.org/doi/full/10.1073/pnas.2106692119 It came out about a month ago, and I've just started reading it. But it talks about neural drift, and they actually were able to create a model for it.

hhs 5y ago
If interested, this is the press release for that paper: http://www.eng.cam.ac.uk/news/study-proposes-how-changes-neural-code-unlock-brain-s-inner-learning http://www.eng.cam.ac.uk/news/study-proposes-how-changes-neu... And in that release, it ends like this: "Dr O’Leary, Associate Professor in the Department of Engineering, said the study emphasises the idea that “drift” may arise from continual learning. “There is a huge unanswered challenge in artificial intelligence, namely the problem of building algorithms that can learn continually without corrupting previously learned information,” he said. “The brain manifestly achieves this, and this work is a step in the direction of finding algorithms that can do the same.”"

acrefoot 5y ago

I'm not sure exactly what you believe, but since you're talking about information storage, can I ask if you mean for working memory or for long-term memory? For long-term memory, the storage is durable and mostly static, while recall is dynamic (and may even affect the storage after further consolidation). For working memory, I imagine it's a pretty dynamic process for both storage and recall. The top-level link seems to only be about working memory, so the title of the HN post should probably be changed to avoid confusion.