Unsupervised Learning Podcast reviews

Just a note: I want to drum up more reviews for the Unsupervised Learning podcast. Can you guys please post a review if you haven’t? What matters most is Apple Podcasts, though I’m grateful for any Stitcher reviews.

I’ve got thirteen episodes up now, that should be enough to get a sense of the podcast.

Finally, if you are a paid subscriber, I’ll be talking to Lee Jussim this week. I think the core readers of this weblog will enjoy that.


A brain warped by reading

Stanislas Dehaene’s Reading in the Brain: The New Science of How We Read is an excellent book because it shows how we reuse preexistent cognitive architecture to extend our capacities through cultural creativity. There is, for example, a part of the brain that is localized toward recognizing the shapes of letters to allow immediate “sight reading” of words (higher mathematics is a similar cognitive extension repurposing).

But nothing is without cost. In Reading in the Brain the author recounts evidence that adaptation to reading may have resulted in a diminishment of human ability to localize and situate ourselves on a landscape with few features (and obviously no road signs!). As we live in a world of clerks and not trackers, this is a reasonable trade-off for most. As someone who reads quite a bit, and has read quite a bit since I could read, I’m sure some of my mental peculiarities are the consequence of the warping effect of constantly reading text.

Aside from reading, over the last 15 years, I have written quite a bit. Last I checked >5 million words. That comes to ~10,000 pages. My writing style has evolved and changed. Just as with reading, I’ve reshaped my brain in various ways. Ways I have not reflected on. And perhaps will never be aware of.

The ancients understood the impact of literacy intuitively. The first great transition likely occurred with the utilization of text to record stories and ideas and freeze in place discourses that were previously free-flowing in the ancient agora (as opposed to the accounting function of Linear B and much of Bronze Age writing).

The rise of text also heralded the long and slow decline of the art of memory. The text itself changed qualitatively and quantitatively. Clay tablets and papyrus gave way to parchment, and parchment gave way to paper. The physical form of text also evolved, from scrolls to a codex. The Bible of Christians was famously one of the first major works distributed primarily as a codex. A book as we understand it (though the separate “books” of the Christian Bible hint at its past as a collection of scrolls). Each of these transitions reduced the price and increased the convenience and accessibility of text, but the printing press transformed the game fundamentally. Due to the crash of the cost of books the art of memory what persisted down into the Renaissance finally expired with early modernity.

These reflections are due to the fact that I have now been heavily involved in two major podcasts for some time. One on science and another on broader topics relevant to South Asians. Additionally, I have participated in a few YouTube live streams as well. The first thing to note is that the density of information per unit of time is lower in podcasts than writing. Part of this may just be that I read fast, and I listen with lower than typical comprehension, but part of it is also certainly objective data density because others admit the same. To “fix” this issue most people simply speed up the podcast, to 1.25 or 1.5 times the regular speed.

But there is a second issue: the very form of writing is structured in a way that is different from the necessarily more extemporaneous form of podcasting. Obviously the latter is on a spectrum. Patrick Wyman’s Tides of History podcasts feel like dramatic readings of essays. In contrast, Joe Rogan’s two to three-hour ramble-fests are winding, digressive, and chaotic. I find Rogan quite entertaining, but I suspect the “learning” portion could be condensed into 15 to 20 minutes out of the 2-3 hours.

When it comes to the science podcasts that I run with Spencer Wells I think they are often dense and tight because the topicality is one where both of us are on solid ground, and science itself is a contingent and structured set of ideas and concepts. In contrast, a podcast where several people try to tackle the definition of Hindu nationalism is naturally going to sprawl in unexpected and sometimes muddled directions.

If podcasting is the new blogging, we are in new territory here. Or are we? Perhaps the more extemporaneous and unstructured manner of dialogue that you see in this medium is a throwback to the ancient agora, and the oral cultures which were dominant even among elites more than 2,000 years ago.


The Insight, Episode 19: Roberta Estes, the Golden State Killer, and forensic genetics

Last week Spencer & I took a break from The Insight. We’re at 71 iTunes ratings. I would appreciate it if readers of this weblog could help us make it to 100 (then I’ll stop pestering you). Also, we only have 5 reviews on Stitcher.

This week we’re talking to Roberta Estes about the arrest of the suspect in the “Golden State Killings”. We kind put this together really quickly since it seemed relevant, and Roberta, Spencer and I have some competency in this area (we’ve all been talking to science journalists). The biggest takeaway from our conversation is that we were a little surprised that it took this long to apply 21st century genomics to forensics.

When I first heard about the arrest I told my wife that it probably was due to a relative match on something like GEDMatch. After the media reported that it was a “new method” I dismissed my supposition because relative matches aren’t a new or novel thing. Well, it turned out that’s exactly what they were talking about!

A lot of the story here is how law enforcement snapped a bunch of pieces together that were out there. The horse has left the barn, and everyone is trying to figure out how to deal with it.


The Insight, episode 17: Stuart Ritchie, intelligence and genes

On this week’s episode of The Insight (Stitcher and Google Play) we talk to Stuart Ritchie, a postdoc in Ian Deary’s lab, about recent developments in cognition and genomics. There’s a reason that Deary gets some time in She Has Her Mother’s Laugh; his group is publishing some really interesting work.

Before we get to the good stuff, Stuart gives us a quick review of general intelligence and why it matters. If you want a book-length treatment then his own book should suffice, Intelligence: All That Matters. Richard Haier’s The Neuroscience of Intelligence goes a little more into the “wet biology” aspect of the brain if that is more your style.

There are two reasons I wanted us to have Stuart on the podcast.

First, psychometrics is not a field which was hit by the replication crisis. It’s a pretty robust and reliable discipline. Companies such as the Educational Testing Service (ETS) rely on the predictive power of the constructs in the field to sell their products. And yet most well-educated people don’t really know much about intelligence testing except that it has been “debunked” by the Mismeasure of Man.

Because people don’t understand the history of intelligence testing (i.e., it enabled the meritocracy by removing the importance of “polish” and “good breeding”) it’s easy for American graduate schools to do things like removing the GRE as a criterion on admissions. Privately some academics have told me that this will mostly result in increasing the importance of undergraduate education and pedigree (because anti-GRE sentiment has become connected to “social justice” I think it’s removal is a fait accompli).

Second, the field of cognitive genomics is moving through a major turning point. A publication like this in January, A combined analysis of genetically correlated traits identifies 187 loci and a role for neurogenesis and myelination in intelligence, is going to be superseded in months. I’m not speculating. I know this as a fact, and so do many others. Where will we be in two years?

Ray Kurzweil has many ideas, some of them interesting, some kooky, and some of them wrong. But one idea he’s promoted which I think is correct is humans are not good at modeling exponential rates of growth. The field of psychometric genomics is now moving into the steep phase of ascent, as sample sizes go well above 1 million, and some researchers shift from proxy characteristics such as education and delve into raw intelligence test scores. Most people “outside of the know” are about to smash into the concrete before they even know it’s coming up at them….


Reflecting on Journey of Man 15 years later

Journey of Man, Spencer Wells’ book and documentary, came out 15 years ago. To a great extent the impact of TV is such that one can argue it introduced genetic anthropology to a whole generation.

A lot has happened since then. On this week’s The Insight we review what’s happened since then, and how Spencer, who started out a conventional academic scientist, became a documentarian.

If you subscribe on iTunes, Sticher or Google Play, make sure to post a review.


My new podcast with Spencer Wells

Spencer Wells and I have a new podcast, The Insight. On the first episode, we’ll be talking about the Neolithic revolution.

We’ve already got several more in the pipeline that will come out in the next few weeks (being edited), including one with John Hawks. This will be a regular thing, so please subscribe!

Update: Also see: http://insitome.libsyn.com/website.

Update: Here is the podcast embedded: