I see lots of people asking about "generating more data". You can definitely encode some tasks in program form and use these programs to generate thousands of new input/output pairs. You can do whatever you want! However...
Show this thread
Regarding the ARC challenge feasibility. It is meant to be difficult, and I don't expect it will be fully solved in 10 years. But I know for a fact that it is partially approachable today using program synthesis. I will be very disappointed if no one scores better than 1 :)pic.twitter.com/QxbwajReVg
-
-
-
...don't get caught up in the mindset that more data will solve all of your problems. That's the deep learning way of thinking. More data will have limited usefulness on ARC.
Show this thread -
The only real use of "training" here is to learn Core Knowledge priors. I believe learning Core Knowledge systems via a pattern recognition approach (like deep learning) isn't the most effective way to acquire them. But it is likely to be viable.
Show this thread -
So if that's your thing, feel free to train a system on as much external data as you want to acquire Core Knowledge priors. Personally, I would simply recommend hardcoding what they look like in the context of ARC. It's all symbols on a grid -- straightforwardly computable.
Show this thread
End of conversation
New conversation -
-
-
Just wondering, do you currently have a system capable of achieving <1 top-3 average error (however small of an improvement)? If so, did it leverage external data to learn Core Knowledge priors (whether that be through programmatic simulation anything else)? I ask this because...
-
I would be extremely surprised if any system with zero prior intuitions could grasp these Core Knowledge concepts well enough to apply the ideas generally.
End of conversation
New conversation -
-
-
Is the test set from the same “distribution” as train and evaluation? i.e. did you generate all 1000 tasks then randomly split into 400/400/100 or did you intentionally leave some new types in test set? Thanks!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I decided to enter because I know nothing about it. geeky BigO(learn something new) lol!
-
Looks challenging
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.