ARC has already become a mediocre benchmark (the lack of attention tells you this is probably the general consensus). Yes there will be ARC 3 but it is now just looking closer to OpenAI Gym so not particularly ground breaking. The creator is also kind of arrogant, greatly Google biased and has gone down the Gary Marcus route of being a critique of current AI systems rather than a real contributor, while it has clearly surpassed expectations since reasoning models became a thing that has not been properly acknowledged.
An example that you should not appeal to authority as authority is not always correct and can even be often wrong unless proven otherwise.
ARC has already become a mediocre benchmark (the lack of attention tells you this is probably the general consensus). Yes there will be ARC 3 but it is now just looking closer to OpenAI Gym so not particularly ground breaking. The creator is also kind of arrogant, greatly Google biased and has gone down the Gary Marcus route of being a critique of current AI systems rather than a real contributor, while it has clearly surpassed expectations since reasoning models became a thing that has not been properly acknowledged.
An example that you should not appeal to authority as authority is not always correct and can even be often wrong unless proven otherwise.