3 ms·
I would like to see this repeated with my highly innovative HARC-HAGI, which is ARC-AGI but it uses hexagons instead of squares. I suspect humans would only mak
by aithrowawaycomm 2y ago
I would like to see this repeated with my highly innovative HARC-HAGI, which is ARC-AGI but it uses hexagons instead of squares. I suspect humans would only make slightly more brain farts on HARC-HAGI than ARC-AGI, but O3 would fail very badly since it almost certainly has been specifically trained on squares.
I am not really trying to downplay O3. But this would be a simple test as to whether O3 is truly "a system capable of adapting to tasks it has never encountered before" versus novel ARC-AGI tasks it hasn't encountered before.
- falcor84 2y agoHere's my take - even if the o3 as currently implemented is utterly useless on your HARC-HAGI, it is obvious that o3 coupled with its existing training pipeline trained briefly on the hexagons would excel on it, such that passing your benchmark doesn't require any new technology. Taking this a level of abstraction higher, I expect that in the next couple of years we'll see systems like o3 given a runtime budget that they can use for training/fine-tuning smaller models in an ad-hoc manner.