The same region that makes something happen is also responsible for inhibiting that action... The same problem emerges when different regions "light up" in the different conditions. It can't be determined whether that is excitatory or inhibitory activity.
I am just a grad student in cognitive neuroscience. But from what I can tell from reading the literature, this comment is, at best, a half-truth. While there may be common gross anatomy for inhibitory and excitatory processes in a subset of task domains, there's good reason to believe that this is not generally the case, especially for the kinds of high-level cognitive processes (e.g., beliefs and preferences) that have been the main topic of this discussion. Specifically, early imaging work utilizing the Stroop Task provided evidence that there are a particular set of brain regions (viz., anterior cingulate cortex and bilateral dorsolateral prefrontal cortex) reliably involved in inhibition of prepotent cognitive processes. As is often the case with imaging studies, this correlational finding has since been established as causal through the use of behavioral manipulations such as cognitive load, neuropsychological lesion studies, and transcranial magnetic stimulation. (But note that we wouldn't have had a clear idea of where exactly to look for the causal underpinnings of cognitive control in the first place if we hadn't first done the correlational imaging work.)
While it's true that both the spatial and the temporal resolution of bloodflow-dependent functional imaging measures currently has practical limitations, it's not so limited as to provide no insight into how the human mind/brain works, as the parent post seems to suggest. Examples of things we've learned from functional imaging are easy to come by, from the previously mentioned finding concerning the neural basis of cognitive control, to quite striking demonstrations of the spatial organization of visual processing (i.e., "retinotopic mapping"), to the existence of imagistic/non-propositional mental representations, to (more recently) the finding that there is shared neural real estate for remembering the past and imagining the future, which yields the surprising yet quite plausible suggestion that memory is not as much for recalling past events, as it is for planning for future events. (Again, all of these findings, save perhaps for that last very recent finding, have been extensively confirmed by convergent evidence from behavioral manipulations, neuropsychology, and the like.)
So that's just a sampling of what we've discovered about the human mind/brain using the current imaging methods, which are limited with respect to their spatial and temporal resolution. And that's to say nothing of all of the amazing technologies that vastly improve the spatial and temporal resolution of functional imaging, which aren't yet widely available, and thus aren't yet in common use among cognitive neuroscientists. Examples include high-field magnets (current studies commonly utilize either 1.5 Tesla or 3 Tesla magnets, but 7 Tesla magnets do exist), improved head coils (8 and 16 channel head coils are commonly used, but 32, 64, and even 128 channel varieties exist), and multi-modal methods that combine, e.g., the relatively high temporal resolution of EEG (which directly measures electrical potential rather than blood flow) with the relatively high temporal resolution of fMRI (and especially spatially restricted fMRI, which increases spatial resolution substantially by zooming in on particular parts of the brain, instead of imaging the whole thing).
So I don't see why folks like the person who wrote the parent post are so skeptical of functional imaging. Yes, it has currently has some practical limitations, and is sometimes misused, as in the esquire piece. But it's quite clear that we've learned a lot from functional imaging, and will continue to do so well into the future.
I'd even go as far as to say that current trends should encourage us not rule out the possibility of utilizing fMRI as a "lie detector." We're not close to having such a technology, yet, mainly because we don't yet know whether the average brain pattern associated with lying is sufficiently distinct from the average brain pattern associated with honesty to make such a technology practical. But why not let scientists do their work in trying to answer the question of whether those two patterns are sufficiently distinct, rather than ruling out the possibility before we even get started? I hope I've provided sufficiently strong reasons to undercut the shared inhibitory/excitatory functional anatomy argument, the it's only correlational argument, and the limited spatial/temporal resolution argument. But perhaps there are other, stronger arguments against functional imaging that I am not aware of? If so, I'd be happy to consider them.