Science

13206 readers

8 users here now

Subscribe to see new publications and popular science coverage of current research on your homepage

founded 5 years ago

MODERATORS

MinutePhrase@lemmy.ml

Dreams are synthetic data and likely help the brain deal with overfitting (www.sciencedirect.com)

submitted 2 months ago by yogthos@lemmy.ml to c/science@lemmy.ml

4 comments fedilink hide all child comments

top 4 comments

sorted by: hot top controversial new old

[–] BaroqueInMind@lemmy.ml 12 points 2 months ago* (last edited 2 months ago) (1 children)

Pretty fucking cool that we are postulating that dreams are not similar to machine learning algorithms in that they are not primarily purposed for adapting to what's experienced during sober consciousness, but rather used in preventing subconscious lower-level instinctual brain functions from over controlling the perception of reality.

The actual opposite is happening with machine learning LLMs where it believes their hallucinations which are derived from obfuscated data is truthful regardless of where it pulled the data from.

[–] yogthos@lemmy.ml 9 points 2 months ago

I think the key problem with LLMs is that they have no grounding in physical reality. They're just trained a whole bunch of text data, and the topology of the network ends up being moulded to represent the patterns in that data. I suspect that what's really needed is to train models on interactions with the physical world first, to create an internal representation of how it works, the same way children do. Once it develops an intuition for how the world works, then it could be taught language in that context.

[–] yogthos@lemmy.ml 8 points 2 months ago

TLDR: Hoel proposes a new theory for the function of dreams, inspired by machine learning's concept of overfitting. He suggests dreams act as augmented samples of waking experiences to improve generalization and robustness of neural representations, preventing them from becoming too specific to waking experiences. It explains why dreams don't become more realistic over time and offers insights for designing algorithms that mimic the phenomenology of dreaming to improve artificial neural networks. Hoel's theory may lead to strategies for promoting robust learning through dream substitutions or other consciousness-altering experiences.

[–] treefrog@lemm.ee 1 points 2 months ago

I hope it pans out for machine learning.

I kept a dream journal every night for over a year. Meditated extensively during that time. And rarely looked at a screen. Prison gives you a lot of free time.

I've also explored the closes eye visual space in drug induced lucid states extensively.

This time article aligns with my understanding of Busdhisms Six consciousness model.

https://time.com/5925206/why-do-we-dream/

In Western language, we would call the eye consciousness the visual cortex. When our eyes are closed, and we're in a relaxed state, we can see muscle memory practicing. Dreams. Visions if in an awake lucid state.

The visual cortex stays active when our eyes are closed. Otherwise it only takes an hour for other senses to start recruiting those neurons.

My dreams tend to be emotionally charged. So, dreams are, in my experience, a combination of the visual cortex doing busy work and unresolved emotional stuff if it's there, or fun joyful stuff, if I'm feeling joy.

I don't know that it has a 'purpose' other than busy work and keeping the cortex wired together. It can be put to purpose, such as dream journaling and deeper self awareness.