Thoughts on NARS?

I’ve come across this: - it sounds interesting to me.

What is everyone thoughts on this approach to AGI? If you’re familiar with the system, did you have any thoughts to share on pros or cons?

Here’s an interesting related paper that delves into the high-level details:

1 Like

I only skimmed their work, but …
I think they should spend more time studying neuroscience. They confidently state basic facts, incorrectly. In particular they claim that most animals are driven by instincts and not learned behavior, which is just not true. Most animals use both instincts and learning.
Edit: i think its important to recognize thats dumber animals are at least somewhat intelligent, because if you cant explain how simple animals learn then what hope do you have of understanding intelligence in complex animals? Mouse brains are a good tool for studying. Theyre a lot easier (and more ethical) to work with than humans or chimpanzees. And im not saying everyone should run experiments, but at least pay a bit of attention to other fields of science outside of your specialized domain.

(For what its worth, neuroscientists could also benefit from studying comp-sci as well.)


These people are logicians-first, just look at this ass-backwards meaningless name:
Non Axiomatic Reasoning System… There is no hope for them.

1 Like

Wow. That’s a pretty negative reception to NARS :smile:

The guy I’ve interacted with a lot (unfortunately) who seems to be pretty invested in OpenNARS as a contributor has to be the most insufferable dude I’ve ever met.

His reception to new ideas is crazy - he peddles ancient, outdated ideas with little motivation of understanding new ideas. A very narrow minded person basically.

So, in short, I admit I’m heavily biased against giving this theory any credence at all. I hope someone can change that.

As I understand it, Patrick’s own thesis states that the architecture is barely able to do MNIST.

I know I kinda stress on ideas not being able to do basic toy tasks, but honestly? calling it a “proto-AGI” is a far fetch for what seems to be a system built on really shaky foundations.

In conclusion, I would need quite a bit of evidence before I change my mind…