Question

Computer scientists often train AI models to recognize objects, aiming to emulate human visual cortex processing. However, a study of 5,000 vision models found that while 95% could identify objects accurately, only 5% developed internal activation patterns resembling human brain activity. Researchers noted that even when models showed similar activation, it was largely driven by specific, pre-programmed image-processing algorithms rather than independent, biological-like learning.
Which finding, if true, would most directly support the claim in the underlined sentence?
A) The pre-programmed algorithms were adjusted to closely mirror the neural pathways of the human visual cortex.
B) The models that did not exhibit human-like patterns were still able to identify objects.
C) The algorithms used to program the models have no direct counterpart in the neural mechanisms of the human visual cortex.
D) The models were trained on datasets that were significantly larger than the datasets used in previous studies.

Answer 1

A

Explanation: If the pre-programmed algorithms were explicitly adjusted to mirror human neural pathways, that directly explains why model activations resembled brain activity — the similarity would be built into the algorithms rather than emerging from independent, biological-like learning. The other choices are irrelevant or contradictory to that claim.