These are not my arguments, since I haven't thought about the issue enough. However, the anthropologist Scott Atran, in response to the latest Edge annual question, "What Scientific Idea is Ready for Retirement?", answered "IQ". Here's his response:

There is no reason to believe, and much reason not to believe, that the measure of a so-called "Intelligence Quotient" in any way reflects some basic cognitive capacity, or "natural kind" of the human mind. The domain-general measure of IQ is not motivated by any recent discovery of cognitive or developmental psychology. It thoroughly confounds domain-specific abilities—distinct mental capacities for, say, geometrical and spatial reasoning about shapes and positions, mechanical reasoning about mass and motion, taxonomic reasoning about biological kinds, social reasoning about other people's beliefs and desires, and so on—which are the only sorts of cognitive abilities for which an evolutionary account seems plausible in terms of natural selection for task-specific competencies.

Nowhere in the animal or plant kingdoms does there ever appear to have been natural selection for a task-general adaptation. An overall measure of intelligence or mental competence is akin an overall measure for "the body," taking no special account of the various and specific bodily organs and functions, such as hearts, lungs, stomach, circulation, respiration, digestion and so on. A doctor or biologist presented with a single measure for "Body Quotient" (BQ) wouldn't be able to make much of it.

IQ is a general measure of socially acceptable categorization and reasoning skills. IQ tests were designed in behaviorism's heyday, when there was little interest cognitive structure. The scoring system was tooled to generate a normal distribution of scores with a mean of 100 and a standard deviation of 15.

In other societies, a normal distribution of some general measure of social intelligence might look very different, in that some "normal" members of our society could well produce a score that is a standard deviation from "normal" members of another society on that other society's test. For example, in forced-choice tasks East Asian students (China, Korea, Japan) tend to favor field-dependent perception over object-salient perception, thematic reasoning over taxonomic reasoning, and exemplar-based categorization over rule-based categorization.

American students generally prefer the opposite. On tests that measure these various categorization and reasoning skills, East Asians average higher on their preferences and Americans average higher on theirs'. There is nothing particularly revealing about these different distributions other than that they reflect some underlying socio- cultural differences.

There is a long history of acrimonious debate over which, if any, aspects of IQ are heritable. The most compelling studies concern twins raised apart and adoptions. Twin studies rarely have large sample populations. Moreover, they often involve twins separated at birth because a parent dies or cannot afford to support both, and one is given over to be raised by relatives, friends or neighbors. This disallows ruling out the effects of social environment and upbringing in producing convergence among the twins. The chief problem with adoption studies is that the mere fact of adoption reliably increases IQ, regardless of any correlation between the IQs of the children and those of their biological parents. Nobody has the slightest causal account of how or why genes, singly or in combination, might affect IQ. I don't think it's because the problem is too hard, but because IQ is a specious rather natural kind.

IQ, or intelligence as commonly understood, is a poor proxy for rationality. In many cases it simple makes people better at rationalizing beliefs they acquired irrational.

What does arguments against "reality" mean?

Arguing against misconceptions of what people believe about IQ? In general I consider arguments against reality pointless. Asking whether IQ is real is the wrong question. It makes much more sense to ask what IQ scores mean.

IQ can be used to give scientific justification to our internalized biases.

I don't want to limit your rights, because you are X. I want to limit your rights because I belong to Y, and as Y does better than X on IQ tests, it is only prudent that we know better what is good for you. I am also not interested in listening to counter-arguments coming from people whose IQ is below 99.

Also, in extreme cases, it can be used to push further policies such as eugenics (the bad kind that everyone has in mind when they hear the word "eugenics"):

Ah... I forgot to say that X shouldn't have the right to have children. No offense meant, but we want to avoid dims out breeding brights. Also, keep your stupid daughter away from my son, as I really don't want my own children to pollute genetic purity of our kind.