In the article Why We Shouldn't Fear Artificial Intelligence, the author writes:"A common fallacy suggests that, because AI are hosted on computers, they’ll be good at manipulating them. But let me ask you this: By virtue of living in a house, do you know how to build / remodel / manipulate one?" What is the type of the fallacy behind the criticized reasoning?
The fallacy in the title does not quite match the one in the text, and is closest to appeal to (irrelevant) authority.
For the one in the text the closest match is the association fallacy: As have C, Bs have C, therefore Bs are As. In this case, what is good at manipulating programs are computers, AI are hosted on computers, therefore AI are good at manipulating programs. The association fallacy is more commonly used as an ad hominem however, either as "guilt by association" or "honor by association". So AI receive the "honor" of being good at symbolic manipulation by association with computers.
Somewhat more remote is the fallacy of division, assigning property of the whole to the parts (the opposite of the composition fallacy), assuming we think of AI as "part" of the computer. However, some use "AI" to refer to that aspect of computers that in particular makes them good at manipulating programs, perhaps much advanced to produce "intelligence". In that case AI would be good at that by definition, and the author would be making a false analogy.