Yeah, this smells very fishy to me. It's almost trivially easy to gather a small validation dataset in humans for the paper. At my institution, it's about $800 an hour to scan someone. You can probably get enough data to validate the model with a half hour scan. Surely the group has enough grant funding to pop a few healthy controls in the scanner.
I haven't looked in super close detail to the paper, but their methods section says that they fit a video model (V-JEPA2) to the fMRI dataset in a voxelwise ridge regression, meaning that the baked in assumption is that the visual response affects each voxel independently. Voxelwise models are very nice for making statistical inferences, but are less good for prediction and modelling tasks, because our brains certainly do not work as collections of independent regions.
BOLD is intensely messy data, and their design is far too simple IMO to reflect anything of reality.
For me, the biggest AI writing tell (other than the blatantly obvious ones) is an unnatural consistency in style, whatever style that may be. It's most apparent in longer pieces, and I'm not sure I can really pin down exactly what it is. But human writers seem to lack the ability to keep a 100% consistent voice and lapse into different registers at different times. LLMs don't have this natural rhythm, which makes for an exhausting reading experience.
The Cincinnati Zoo covered all their parking lots with solar panels last year. Your car stays cool in the summer, and there's motion activated lighting under the panels after dark. It's awesome.
"Genetically encoded" is the appropriate term here -- it was used in the original journal article. It's a common industry term in neuroscience research. For example, GEVIs and GECIs are "genetically encoded voltage indicator" and "genetically encoded calcium indicator" respectively. "Genetically encoded adenosine sensor" here is a term of art.
> I have over 14 years of education in developed countries, and out of those, maybe 1 year combined meaningfully helped me in my jobs/career in terms of skills.
I think you're underestimating the effect of 14 years of daily training in literacy and numeracy.
I think Buddhism still (arguably rightly) doesn't sit entirely well with non-religious Westerners. I have studied with a Zen Sangha and transmitted teachers on and off for a bit and have found their explanations helpful. However, it's absolutely undeniably that the Buddhist cannon is full of batshit insane stuff, just like any other religion. You can write them off as skillful means, but in some ways I think it's more honest to say that you practice meditation with Buddhist characteristics than to say that you're a real Buddhist if you don't have the time of day for spirits and dieties.
Again, this isn't saying that Buddhist modernism is bad. I'd argue that having clear eyes about what parts of Buddhist practice you're willing to take and leave is good.
There are specialized architectures (the Tolman-Eichenbaum Machine)* that are able to complete this kind of task. Interestingly, once trained, their activations look strikingly similar to place and grid cells in real brains. The team were also able to show (in a separate paper) that the TEM is mathematically equivalent to a transformer.