Editor’s note

To start, a quick thought.

What if “crazy people” help us stay smarter than AI? In this day and age, we fear losing the sovereignty of human intelligence to machines. Rightly so. Maybe we are making a mistake. We attribute value to people who have gone to i.e Harvard and less to people who are, for instance, schizophrenic. The former, however, are merely better at applying existing knowledge to defined criteria. Machines already do that better than us. The latter have more creative and original ideas, even though the system does not know how to exploit them. Further, LLMs do not generate “genius” ideas because of well-known fundamental or designed restrictions. The same cannot be said about our neurodivergent friends.

Please enjoy!

Unrelated-but-interesting

There is a return to religiosity among previously secular people. The people who read Dawkins are coming back to metaphysical beliefs. According to the author, this is because humans need social practices and rituals but they could and should do without the beliefs. This is leading us back to the Dark Ages he says. This maximally rational critique misses the point. He believes we can separate both. But singing together for the sake of singing together simply is not convincing for most people as that leaves us as “the sum of the parts” whereas it is about creating a whole that is more than that. Where 1+1 = 3 — jimrutt.substack.com

Is terminal lucidity real? This is when patients suffering from dementia are on their deathbed and suddenly come back to making sense just before dying. The author suggests we cannot know for sure even though I am certain my dear readers have experienced or heard of that. The author suggests that the families of such patients want terminal lucidity to be real, for emotional reasons (which does not mean it is a fabrication, just that it is difficult to measure, as very little has been invested in the study of such a phenomenon). The author suggests two reasons why it could be very real: the reduction of brain swelling before death because of food/water refusal, and some kind of strange rearrangement of matter making hormonal release functional again. Interesting throughout — preservinghope.substack.com

Anxiety has been linked to low levels of choline in the brain. This is the first meta-analysis linking a chemical deficiency for anxious disorders, a study conducted at UC Davis. This allows me a little digression. The idea that depression is due to a chemical imbalance (notably serotonin) was used by drug manufactures to promote anti-depressants (notably SSRIs). A 2022 review by researchers at UCL showed that this idea was not substantiated. Indeed, would that not be great? If only you took our miracle drug, then your mind's equilibrium would be restored. Anyway. Foods high in choline are beef liver, eggs (the yolk), beef, chicken, fish, soybeans and milk — medicalxpress.com

We are addicted to our phones because of their simulated physicality. We personify them. Not only because of the psychologically-manipulative apps on them. Say that again? When swiping your finger to the edge of an iPhone, the content behaves like a rubber band, mimicking the suppleness of skin. When we see the blue dot on a map app, we think “I am here” rather than “my phone is here”. When we use facial recognition to unlock the device, we think “it is recognizing me, as a friend would”. Beyond social media, the author argues that we are addicted to the device itself — theconversation.com

US-style polarization is spreading in the West because of the design of cable TV and social media platforms. Extreme, less nuanced content is rewarded with more user engagement. Recommendation algorithms pick up this signal and further fan the fire. Culture war content attracts more attention than say economic topics. This reinforces media players to create more of that. This is great news for the people in power. As long as we are focused on fighting over questions of identity and such, the questions of wealth and redistribution are not addressed — ft.com

AI & software

It is excruciatingly difficult to evaluate LLMs on qualitative tasks. Unlike mathematics or software engineering where improvement is easily measured, tasks like writing resist simple scoring. While analyzing how Kimi found a way to make their K2 model better at creative writing, Drew Breunig realized that an imperfect, fuzzy categorization for what constitutes good writing is still better than no categorization and that AI researchers should not abandon this goal. Kimi K2 is at the top of the Emotional Intelligence benchmark for LLMs, beyond the likes of OpenAI, Anthropic or Google Deepmind — dbreunig.com

AI has flipped product development on its head. It used to go like this: designers create mockups and send to engineering. Because of varied constraints, the designs got into production over months or quarters. So designers were ahead of engineering, which was reflected in the ratio of designers (few) to engineer (numerous). Today, things are wildly different. The time it takes to create a working prototype has been reduced by a factor of 10 or more. Designers now have to “clean up” the work of engineers ex-post. But they are still very much so needed, as these prototypes lack the refinement of stellar products — lukew.com

Loose ends

Czechoslovakia tearoom ambient music, courtesy of NTS. Serene and airy, I listened while writing this edition. Recommended! — nts.live

It is time to say that China is no longer “catching up” to “exceptional” America. It is rather defining modernity with the same weight as their Western adversary — sinicapodcast.com

A complete introduction to Game Design. Long, rewarding read — raphkoster.com

A new gel-like material has been created by scientists to restore tooth enamel. This was previously impossible — nottingham.ac.uk

A collection of words that don't translate into other languages. An example is the German word Packesel, or someone who's stuck carrying everyone else's bags on a trip — eunoia.world