With the Reddit thing, it'd be easy to miss what's going on on StackOverflow. Basically:
1. Site owners allowed LLM-generated content.
2. Mods are on strike.
3. Also, back in March it turns out SO turned off the Creative Commons data pipe, which backs up the site to the Internet Archive, in an attempt to confound using SO for training data.
Alzheimer's went into the textbooks with n=2
Many of you might know (or not) that Alzheimer's disease is named after Dr. Alois Alzheimer who first reported that the brains of some patients with age-related dementia had "a peculiar substance" (which today we know as beta-amyloid plaques).
But did you know that he found that in exactly 2 patients that he studied before his boss put it in a 1910 textbook and coined it "Alzheimer's disease"? That substance could have been a complete coincidence! (And in that case, I guess we never would have heard about it again) ...
Whoa. Let’s get back to rationale discussions in brain research:
Debate is vital … The discussion that followed … this is a shame ….
This is real-world XKCD
"Elsevier…set the NeuroImage APC…at $3,450 USD. Compared against this, estimates of direct article costs at relevant journals are generally around $1,000 or lower.…It is wrong for publishers to make such high profits."
See the #OpenAccessDirectory list of these journal #DeclarationsOfIndependence.
Swept away today by reading about Capgras' syndrome: the unshakable delusion that familiar people around you have been replaced by imposters. Not only because of "how could our brains lead us to think that?!" but also because of its insights into tensions between brain/mind-based causal explanations in psychiatry. This is going to be a long post, but deserves the characters, I think.
The syndrome was first described by Dr. Capgras in 1923. Following the death of her twin boys in 1906, a woman declared that her husband had been murdered and replaced by one among a rotating set of 80 imposters. She wanted to divorce the double and requested a separation from the courts. Capgras speculated that his patient had a disturbance in her ability to recognize familiar people. He also noted that she did not have a distortion in her ability to see or remember who is familiar to her, but that normal sense of familiarity is accompanied by strangeness. You can read a translation of the original paper here:
Capgras is rare and when it happens, is most often associated with Schizophrenia, bipolar disorder and dementia. Over the past century, many theories about what causes Capgras syndrome have emerged. They reflect the tension between "organic" (biological) causes and "functional" causes that are not biological. As recently as 1983, Robert Berson reasoned that Capgras could not be organic because it was such a specific delusion. Instead, he reasoned:
In Capgras’ syndrome, I would suggest, pathological splitting has occurred: Internalized object representations are split into "good" consciously acknowledged and "bad" unconscious images. The images persist until an even in interpersonal relationships triggers and important affective change. Previously repressed feelings enter awareness, the "bad" images surface, and the patient declares that a significant other is not who he seems to be but a double.
You probably won't be surprised to hear that a pile on ensued. They were all very polite about it, though.
Sir: I would like to heartily congratulate Robert J. Berson on his excellent article ... However, I must vehemently disagree ....
In 2017, Michael Fox's group at Harvard used a network mapping approach to identify where the lesions in 17 patients were located. They found that they overlap with brain regions involved in 1) familiarity perception and 2) belief evaluation. In other words, yeah - organic.
Video of one Capgras patient: https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/brain/140/2/10.1093_brain_aww288/6/brain_140_2_497_s1.zip?Expires=1683368818&Signature=bdmW-6ZuKdN~XCCxG6pBZOIjncyX9tCkcjn88tCMv0ifrkz6CsN-uS~iP18hAnUjLRx5VbSb89~d3yUsYkggaBco48IBaTJ1LNdJYPmqAmbdvMP2Gb~4PeNNbxqBYGqRZszD2wtIDwm69QDnFApE9LDcPhsyc45CDNVRiJ7dRU-c~9BPJdmY~V5IhwA4HU9uE5dEdQITRVzDq6MBzGfp1T7jCE2J~iG~U4mp~WDaIj2dJoJsWbctS1AqipTBJM1Uj-U9XT2JRdBJzFzPVSmPUNjqu66uHMCSp7BHeL4skabV8Jin1H2JkdHPdWswudIe5LbTDQQSsMdhX--LU-wJvQ__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA
Still an open question: what exactly is happening in those brain areas that leads to imposter illusions?! Here's a fascinating news and views that puts it in context: https://doi.org/10.1093/brain/aww323
Notably, there's a lot of great work going on both to investigate the nuts-and-bolts of how face familiarity works:
As well as belief evaluation, e.g.:
While I always love the nuts-and-bolts stuff (and have a soft spot in my heart for familiarity in particular), linking this work to Capgras makes it even more fascinating to me.
"Marmosets do not have opposable thumbs. [...] Marmosets were rarely taken in by magic (just 6% of the time). They simply chose the hand in which the marshmallow was initially placed, and stuck with it."
The audacity of for-profit publishing is amazing. If an article is retracted, it is still viewable, and still for $$$. The paper that initiated the retraction is ALSO for $$$.
So they MAKE MONEY WHEN A PAPER IS RETRACTED. Let that sink in when you think about incentives.
Machine learning can easily produce false positives when the test set is wrongly used. Just et al in
@NatureHumBehav suggested that ML can identify suicidal ideation extremely well from fMRI and we were skeptical. Today retraction and our analysis of what went wrong came out.
Here is the retracted paper: https://nature.com/articles/s41562-017-0234-y and here is our refutation https://nature.com/articles/s41562-023-01560-6. If true, the paper's approach could revolutionize psychiatric approaches to suicide.
So what went wrong? The authors apparently used the test data to select features. Obvious mistake. A reminder for everyone into ML: never use the test set for *anything* but testing. Only practical way to do so in medicine? Lock away the test set till algorithm is registered.
Side note: it took 3 years to go through the process of demonstrating that the paper went wrong. Journals need procedures to accelerate this. Also, all the good things of this were by
TikZ library for the aperiodical polykites (and related tiles)
🎉🎉🎉 Student and Teaching Assistant applications are now open for Neuromatch & Climatematch Academy 2023! 🎉🎉 🎉
⚠️ Submit your application through the portal on our website for Computational Neuroscience, Deep Learning, or Computational Tools for Climate Science @climatematch courses by April 24, 2023.
Brain age predictions in longitudinal data reveal the importance of scan quality and field strength | #bioRxiv https://www.biorxiv.org/content/10.1101/2023.03.31.535038v1.abstract?%3Fcollection=
"... flashy presentations and sheer ambition are poor indicators of success when it comes to understanding the complex biological mechanisms of brains. Today, as we bear witness to a game of Pong being mind-controlled by a monkey as part of a typically bombastic demonstration by Elon Musk’s start-up Neuralink, there is more of a need than ever to unwind the cycles of hype in order to grapple with what the future of brain technology and neuroscience have in store for humanity."
On explanations in brain research:
A thread of the same idea comes up again and again in brain research. It's the notion that identifying the biological details (such as the brain areas/circuits or neurotransmitters) associated with some brain function (like seeing or fear or memory) is not a complete explanation of how the brain gives rise to that function (even if you can demonstrate the links are causal). To paraphrase:
Mountcastle: Where is not how https://www.hup.harvard.edu/catalog.php?isbn=9780674661882
Marr: How is not what or why http://mechanism.ucsd.edu/teaching/f18/David_Marr_Vision_A_Computational_Investigation_into_the_Human_Representation_and_Processing_of_Visual_Information.chapter1.pdf
@MatteoCarandini: Links from circuits to behavior are a "bridge too far" https://www.nature.com/articles/nn.3043
Krakauer et al: Describing that is not understanding how https://www.cell.com/neuron/pdf/S0896-6273(16)31040-6.pdf
Poppel: Understanding brain maps does not formulate "what about" the brain gives rise to "what about" behavior https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3498052/
Also, I imagine that some form of the opposite idea must also be percolating: the notion that 'algorithmic' descriptions of the type used to build AI will be insufficient to do things like treat brain dysfunction (where we arguably need to know more about the biology to, e.g., create drugs). Any explicit references of that idea? @albertcardona @schoppik, @cyrilpedia, Anyone?
Had some fun and wrote a post about the different types of neuroscientists I've seen.
Which one are you?
Negativity drives online news consumption
We're a neuroscience blog trying to make neuroscience accessible for everyone! Check it out here: https://neurofrontiers.blog
QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.