Action Potentials for November
Note: I haven't done deep dives on all of these links. Corrections, criticisms, castigations, beratement, &c, are welcome.
1: A new zebrafish serial section electron microscopy data allows for automated reconstructions of synaptic connections.
2: A study on ancient Neanderthal DNA from a cave in southern Siberia finds that the maternally inherited mitochondrial genome had much more diversity than the Y-chromosomes. This suggests that Neanderthal women were more likely to move to a different community from the ones they were born in, while Neanderthal men tended to stay put.
3: Romeo Stevens on threefold training (2021). Argues that there are three types of mind training: concentration (eg focusing on one's breath), insight (eg noting sensations), and integration (eg psychotherapy or habit change). All three are necessary for growth, and getting stuck with progress in one area is usually due to a lack of progress in one of the others. I don't know much about this space, but this short article seems pretty persuasive. As a side note, over the years, I've found that Romeo Stevens probably has the highest ratio of insights per text of anyone I've ever read.
4: A list of things experts said would never happen or were a bad idea before their time.
5: A study finds that at the end of 2021, “birth rates among college-educated women were 5.6 percent higher than had been anticipated prior to the pandemic”. Helping people reach their fertility goals alone seems like a major reason for governments to support WFH policies.
6: Interesting article by Jacob Cannell argues that empowerment — that is, helping others to achieve their goals — is a workable human-AI alignment goal.
7: Open Images V7 from Google, an updated image dataset with 9M images and thousands of object labels. Distinguishes between “things” (which are countable, like cars) and “stuff” (which is not, like dirt). Mainly posting this because makes me happy to know that “things” and “stuff” are now technical terms.
8: New study finds that in aging, CD8 T cells are associated with oligodendrocyte loss and increased interferon-responsive microglia. Mice treated with the checkpoint inhibitors anti-PD-1 and CTLA-4 (ICB) for 6 weeks at 18 months also have worse markers of white matter aging.
9: A study of cognitive changes in aging finds that participants with obesity had lower cognitive scores at baseline than participants with normal range weight (difference in intercepts, -0.36 [95% CI, -0.46 to -0.17]; P<0.001). This was partially but not fully accounted for by elevated blood pressure and fasting glucose levels, both of which are known to worsen brain health and cognition.
However, there was no difference in the rate of cognitive decline in people with and without obesity. In fact, adjusting for blood pressure and fasting glucose levels, obesity was associated with a slower rate of cognitive decline relative to baseline. This is consistent with results showing that APOE e4 Alzheimer’s risk alleles are linked with lower weight in people 53-72 years old.
10: “Blinking lights to slow down Alzheimer's?” I remember when this study came out, which was during my Ph.D. studying Alzheimer’s disease. I found it sort of unbelievable then that merely flashing blinking lights at 40Hz could be helpful, but the data still looks good, so maybe that’s just another L for my absurdity heuristic. I agree with the author that this seems more promising than small-molecule drugs. More generally this falls into the camp of neurostimulation, which is one of the most promising areas in clinical neurobiology right now.
11: Our paper using deep learning to segment neurofibrillary tangles in microscopy images has now been published. Hopefully, the Non-Fungible Token fad will continue to fade so that we can once again focus on the actually interesting and important NFTs — neurofibrillary tangles.
12: Using genome by trauma interactions in the UK Biobank data set to explain major depressive disorder (MDD). They found that genome by self-reported trauma interactions could explain up to 20% of the variation in MDD, and that this effect was 5-fold stronger in male than female participants.
13: eLife will no longer reject papers once they are under review. This is highly innovative, makes sense especially in the context of open review, and I support and appreciate their efforts.
14: Maybe people don’t like inflation because it tends to hurt real wages — this fits with my experience and I’m surprised it’s not already a standard view in economics.
15: Systematic review of metformin for anti-aging in animal models. Finds that “metformin was not significantly associated with an overall lifespan-prolonging effect in either mice or nematodes”. The monarch of endpoints strikes again.
16: Article on the use of drones in the Russia-Ukraine war. It seems to me that modern warfare is already heavily dependent on drone capabilities and that this trend is only going to increase as computer vision and robotics improve.
17: Engineered T cells deplete B cells and remove clinical and laboratory evidence of systemic lupus erythematosus (the most common form of lupus) in an n = 5 clinical trial.
18: New machine learning paper from Google on segmenting connectomics data. The 3d view really helps show how dense astrocytes are.
19: We don’t talk enough about sleep hygiene nocebo. Let’s say you’re a person who has anxiety about the quality of your sleep. Your doctor tells you that you should do sleep hygiene activities XYZ in order to improve your sleep. But because of time, executive dysfunction, fatigue, forgetfulness, oppositional defiant disorder, etc, you don’t do most of XYZ most of the time.
This is then another reason that you are anxious about your sleep and think you won’t be able to fall asleep. This gets to a possible harm of cognitive behavioral therapy, which is not always a benign thing. CBT needs to be done in a thoughtful way. For example, the Law of Equal and Opposite Advice still applies. (This was written on my phone while I was lying in bed.)
20: Eric Drexler on multipolar approaches to AI.
21: Scrolls buried in ash by Mount Vesuvius in 79 C.E., discovered in a compacted form in 1752, may now be readable in 2022 by the use of a non-destructive synchrotron that accelerates light through the scroll and captures 2d images on the other side, followed by 3d reconstructions. Another advance in the field of archaeology. One day, if humanity survives and flourishes, we will have neural archaeology.
22: Epigenetic reprogramming of blood stem cells: deleting the gene Phf6 in 15-month-old mice led to a reversal of blood stem cell transcriptional aging phenotypes when measured 8 months later.