July 2022 Nuggets
A selection of science, technology, and other news from June 2022
And there goes July. Time for some nuggets. Some science and a few bytes of technology that caught my roving eyes in the past month. Life’s complexity, baby AI, happy phones, and more.
Science 🔬
Back in the day. Most of us have heard about the Cambrian explosion, that moment in the distant past where, seemingly all of a sudden, life blossomed into a (mad) collection of complex body shapes. More and more, though, we are realizing that that ‘seemingly’ in the last sentence is warranted. So-called Ediacaran fauna (pre-’explosion’) turns out to have been much more complex than we tough. This study is a great example of this, suggesting extensive ecological specialization in Ediacaran ecosystems
Peas in a pod. On July 20th, 2022, it was Gregor Mendel’s birthday. Mendel was the Austrian friar who provided us with some of the foundations from which modern genetics would be built. You’ve probably heard about his crosses of yellow and green and smooth and wrinkly peas. This article goes into the influences that inspired Mendel, as well as the influence he has had. In celebration:
Imaginative evolution. There are not that many things that set us, humans, apart from other animals, but surely our imagination is one? Think again. In this paper, the authors look at both behavior and brain structure to argue that a core capacity for imagination - episodic-like memory - evolved several times in animals, and:
…that the evolution of prospective, planning-enabling imagination is a major transition in cognition and consciousness.
Depression (not) in the brain. You’ve probably noticed the recent kerfuffle around a review that showed very little (not to say no) links between depression and serotonin levels. As I wrote here, depression is a multifactorial condition that is unlikely to be reducible to a single neurotransmitter/molecular imbalance. Another recent study shows how even brain scans can’t help us (despite what some people on social media seem to think…). The brain scan overlap between healthy and depressed brains is so substantial that it’s not (yet?) possible to diagnose anyone with a brain scan.
Language in the brain. What do brains look like when speaking different languages? New research, comprising 45 languages from 12 language families, suggests that answer is ‘surprisingly similar’.
Technology 💻
Baby AI. A little over a month ago, I argued that AI might benefit from having a body. In that same newsletter, I suggested that it could be a good idea to teach AI as a child. And look here: a study that teaches intuitive physics to a deep-learning system by borrowing ideas from developmental psychology.
AI democracy? One of the major challenges in today’s AI (beyond developing actual general AI) is what’s known as value alignment. Here, researchers present a human-in-the-loop machine learning system that designs a social mechanism that humans prefer by a majority. Or:
The AI discovered a mechanism that redressed initial wealth imbalance, sanctioned free riders and successfully won the majority vote.
I don’t know about you, but that doesn’t sound all too comforting. (Not to mention that ‘majority’ tends to give me an unpleasant feeling in the back of my neck. Tyranny of the majority, anyone?)
Happy phones. We are inundated by news about how our smartphones are making us depressed and lonely. Here is a good article that pushes back a little bit against that narrative and makes the case that those some phones can be helpful too. By using the power of affective computing and digital therapeutics, several researchers are trying to turn those pesky phones into tools that can actually support our mental health.
Philosophy 🤔
What we (don’t) say. You’ll certainly remember the recent media storm around the Google AI, LaMDA, that was claimed to be sentient by one of the company’s (now former) engineers. This long-read relates the whole discussion to the philosophy of language and technology. Not only in terms of the language models that underlie a lot of these machine learning systems but also in how we use language to describe and understand them (in which we might be coming up short, according to the authors).
At what point does the performance of reason become a kind of reason?
Bring back the rainbow. Here is a graph that made me go ‘huh’, so I had to share it. Are we making the world less colorful?
