One of the seminal insights into the nature of artificial intelligence

This is an inference-based question. Based on the information in the passage about Yamakawa’s views on the developments in neuroscience, we are trying to identify a view that Yamakawa would agree with.

The passage states “Yamakawa is attempting to overcome Moravec's paradox by using deep learning to develop an artificial general intelligence through a "whole-brain architecture" in which each part of the brain is modeled in a machine-learning module and then individual modules are integrated to create a cognitive architecture.

The passage also states that Yamakawa sees two benefits to this approach:
• He believes that artificial intelligence developed in this manner could incorporate human behaviour and thus bring about a better interaction between machines and humans.
• He also believes that such a collaborative platform that resembles the structure of the brain would allow different people to contribute to the project and prevent it from becoming exclusive to one person or entity.

The passage describes Yamakawa’s view as sanguine, which means optimistic. An understanding of the fact that Yamakawa believes that the structure of the brain can be simulated and that this simulation would facilitate more contributions to knowledge about the brain and neuroscience helps us eliminate Options C and D. Both these options are negative as they state that neuroscience is not yet developed enough to allow successful modelling of the brain structure. However, it is clear from the passage that Yamakawa is trying to overcome Moravec’s paradox by developing artificial intelligence through a whole-brain structure. So, Options C and D can be ruled out.

Option B takes only part of Yamakawa’s idea. He believes that a collaborative platform will help neuroscience evolve. However, Option B only states that the whole-brain architecture will improve neuroscientists’ understanding of the brain. Yamakawa, on the other hand, emphasizes the idea that neuroscience will evolve because of people’s contributions to it. Since Option B overlooks a vital point, it can be eliminated.

Option A is an extreme option as it talks about “an accurate computer modelling of all brain functions”. Yamakawa believes that a model can be built but his focus is on the idea that the model will improve the understanding of the brain and that various people can contribute to this understanding. So, Option A also can be eliminated.

The very basis of Yamakawa’s views is that the model of the whole-brain structure will further the understanding of the brain and help to evolve neuroscience. So, Option E is correct.

Jayanthi Kumar.
_________________

Can anyone explain question 3, please!

Q3 is based only on the first paragraph, which tells us that AI is good at abstract reasoning, but is not good at "simple, physical jobs" involving "perception and sensorimotor skills", things "even very young children" can sometimes do. The passage gives an example of such a task: 'efficiently negotiating a cluttered room'.

So we want to know, from the passage, a reason an AI is not able to do something like find its way around a messy room, but why a child might be. The passage explains this in terms of evolution: humans have been working on (by evolving) these skills for "hundreds of millions of years". Humans haven't been evolving their abstract reasoning skills for nearly as long. So perhaps from that perspective it makes sense that these sensorimotor skills are harder (or take longer) to develop, and thus would be harder for an AI to "learn".

Looking at the answers to Q3, you can rule out B, C and E immediately, since they aren't even mentioned in the passage (the passage doesn't discuss how quickly AI will, in future, be able to evaluate complex inputs, nor how easily AI is distracted, nor how easy it is to break chess down into discrete steps). Answer A is a bit tempting, because it mentions the evolutionary justification from the passage. But it frames the explanation incorrectly. We have no idea from the passage whether "machine learning is subject to evolutionary pressure" (it certainly could be, in the sense that evolutionary pressures could be simulated to refine an AI). But answer A misses the point anyway, since all human skills developed through evolution, whether we're talking about sensorimotor skills or abstract reasoning skills. If an AI were unable to emulate a human skill that developed because of evolution, an AI wouldn't be able to do anything, including the abstract reasoning we know an AI can do. The point is that sensorimotor skills took longer to develop than abstract reasoning. So D is the right answer here.
_________________