Razib Khan's Unsupervised Learning

In the 1600s, Taiwan, then known as Formosa and under Dutch control, was regarded as a barbarian frontier province, nigh unto empty by Han standards, and used as an escape valve for mounting mainland Chinese population pressure. By the 1700s, this unchecked inflow of mostly Fujianese immigrants had tipped the island’s demographic majority over to the Han. This Han ascendance would have unwelcome sequelae both for the short–sighted Dutch colonists who once welcomed the unchecked flow of peasant immigration and for the ethnically Manchu Qing dynasty, wary of yet another Han-dominated province of China. But for our ancient Austronesian story, this recent chapter of Sinicization and intensive settlement is merely a layer to be peeled back. Who were the native Taiwanese people being progressively demographically overwhelmed by that Han tsunami?

Throughout those centuries of Dutch, Tungning and Qing rule, native Taiwanese had continued ranging across most of the island, hunting and farming as they had for millennia. The Chinese economic relationship with the native Taiwanese resembled that then developing in North America between Europeans and Native Americans, with the indigenous inhabitants trading the newly arrived settlers products like deer skins for imported finished goods.

Today 600,000 people of indigenous Taiwanese ancestry remain on the island, maintaining their cultural traditions in eastern-highland strongholds. Sixteen indigenous Taiwanese tribes are officially recognized, including both the Ami and Atayal, who appear among the Human Genome Diversity Project (HGDP) sample populations. They are divided between those of the highlands in the east that occupy two-thirds of the island, and those of the western plains, where most of the population resides, and who are more Sinicized. While the Han speak a Sino-Tibetan language, the native languages of the indigenous Taiwanese are all Austronesian. In fact, of the 14 subgroups of Austronesian languages around the world, 13 appear on Taiwan alone. Though a minority faction of scholars challenge this taxonomy, linguistic analyses of Austronesian languages using best-of-breed phylogenetic methods do seem to concur on Taiwan being basal.

The standard explanation for these sorts of patterns in human evolutionary genetics, where a lone narrow geographic region harbors the lion’s share of existing diversity, is that this busiest, most diverse node usually marks the origination point of a broader population movement. Africa is the most genetically diverse continent, and it is also likely to have harbored the modern human lineage the longest. Likewise West Asia alone boasts more genetic diversity than the rest of the vast Eurasian super continent. And East Asia is genetically more diverse than the New World. As populations branch out, those fateful pioneers whose genes are an arbitrary subsampling of the initial population, inevitably grow more and more homogenized, constrained to only ever draw on a subset of the original population’s heritable variety, progressively cutting into their patrimony of initial diversity over time. The same dynamic applies to cultural traits like language. Given Taiwan’s motherlode of linguistic diversity, we can confidently conclude that this must be where the Austronesian languages began their expansion millennia ago, with a fateful subset of that day’s population spreading southward, before branching west and east into Malayan and Polynesian subgroups.

Though those two Taiwanese aboriginal groups, the Atayal from the north and the Ami from the south, have long figured in human population-genetic studies because they contributed to the HGDP dataset begun in the 1990s, a 2023 paper examined a wider range of Taiwanese tribes to characterize the dynamics of a possible “out of Taiwan” expansion. A simple admixture-based analysis pooling world populations, both modern and ancient, and allowing ancestral clusters to fall out of the data, shows that the dark-brown element predominates in Austronesian speakers, is maximized in the Kankanaey ethnic group from Luzon in the Philippines, and appears ubiquitous in indigenous Taiwanese. Samples from Borneo and southern Sumatra (Semende) in Indonesia show this component as well, but always paired with the dark-blue element that predominates in the Htin ethnicity of Thailand and Laos. The Htin speak an Austro-Asiatic language, and descend from the first wave of rice farmers to arrive in Southeast Asia. This Austro-Asiatic component also appears in a high fraction in neolithic Man Bac samples from northern Vietnam dating to 1800 BC, which correlates perfectly with agriculture’s arrival in the region from the north. In maritime Southeast Asia, it seems clear then that intrusive Austronesian populations mixed with earlier Austro-Asiatic populations.

In contrast, the Tongans, selected here for the purpose of representing modern Polynesians, show a high fraction of the Austronesian ancestral element, but without a substrate of Austro-Asiatic ancestry. Instead, what is evident in these populations is a minority contribution closely related to the Melanesian ancestry of New Guinea highlanders. While the branch of Austronesian voyagers that moved west encountered Austro-Asiatic speakers farming rice in Borneo, Java and Sumatra, those that tacked east encountered Melanesian populations. These contacts are faithfully reflected in their contemporary genetics.

The earliest ancestors of the Polynesians are almost certainly what is known in the archaeological record as the Lapita culture, which spread across Melanesia and western Polynesia between 1500 and 1000 BC, originating from a forward Austronesian hub in the northern Philippines (the first stop out of Taiwan). The ancient DNA from the earliest sites (in Tonga) is notable in showing almost no Melanesian ancestry, suggesting a later admixture event swept across Polynesia. Genetically, the closest modern populations to the Lapita seem to be ethnicities in northern Luzon and indigenous Taiwanese, confirming a north-to-south migration event after 2000 BC (also traceable in the archaeological record).

The Lapita’s Polynesian descendants became famous as the world’s most audacious open-ocean mariners, but their overall cultural package shows evidence perhaps more than anything of their ancestral Austronesian adaptability. Rice agriculture arrived in Taiwan 5,000 years ago, likely as part of the cultural package of the proto-Austronesians, who migrated from the Chinese mainland. Rice and Austronesian culture are closely linked on their southwestern migration, one of their legacies being to have brought a more intensive wet-paddy rice agricultural system to regions that had been previously opened to agriculture by Austro-Asiatic farmers (eventually carrying this package even to Africa, to the Madagascan highlands). But as another contingent of Austronesian mariners sailed southeast from the Philippines, this subgroup seems to have abandoned rice, in favor of crops like taro, yams and banana. The word for rice even disappears in the proto-Polynesian language, around 1000 BC. Why? Rice seeds do not survive well in salty conditions, a common characteristic across islands where much of the land is very near to the ocean. Rice is also ill-suited to the thin soils of many Pacific islands and atolls, providing a poor return on intensive cultivation. What had been essential, western Austronesians’ very staff of life, was quickly jettisoned by pragmatic eastern Austronesians.

Paleogeneticist David Reich has said that one of the major insights of ancient DNA is that markers tracing lines of biological descent can just as well be tracer dye for cultural movements. The Austronesian languages, from Taiwan to Madagascar to Easter Island, are all clearly interrelated. But with ancient and modern DNA, it also grows abundantly clear that their speakers share a common genetic heritage, overlaid at various depths upon the local substrates with which they synthesized. Using ancient DNA samples from northern and southern China dating to the period between the Last Glacial Maximum 20,000 years ago and the end of the last Ice Age 11,500 years ago, researchers have shown that the genetic elements that coalesced in Taiwan by 5,000 years ago seem to have antecedents in local foragers who ranged across the coastlands of southeast China when sea levels were lower and East Asian populations were separated into very distinct northern and southern clusters.

Today on Unsupervised Learning, Razib talks to Kat Rosenfield. She is an American novelist, journalist, and culture critic known for both her fiction and commentary on contemporary political debates. She began her career in publishing and as a reporter for MTV News before branching out into broader cultural criticism, contributing to outlets such as The New York Times, Wired, Vulture, Reason, and UnHerd. As a novelist, she has written You Must Remember This (2023), No One Will Miss Her (2021), Inland (2014) and Amelia Anne Is Dead and Gone (2012). Rosenfield also co-authored the New York Times bestselling A Trick of Light (2019) with Stan Lee. She is currently a contributor to The Free Press and a co-host of the Feminine Chaos podcast.

In the four years since Rosenfield was last on Unsupervised Learning the “culture wars” have seen a changing of the front lines; the woke ascendancy is no more, Elon Musk purchased Twitter, and Donald J. Trump is back in the Whitehouse. After the recording of this podcast Rosenfield was the target of a “cancellation” campaign due to her making light of white liberal gushing over the prose stylings of Ta-Nehisi Coates. But this being 2025, Rosenfield seemed more amused than afraid of the concerted attempt to “drag” her online and notify her employers. The first portion of this podcast discusses where we are now in the culture wars, how things have broadly changed, and which institutional pockets of the old woke ascendancy remain. Rosenfield and Razib also discuss the rise of gender polarization in online culture, and in particular, among Gen-Z.

Subscribe now

Share

Give a gift subscription

Today Razib talks to Eric Kaufmann, a Canadian professor of politics at the University of Buckingham, where he directs the Centre for Heterodox Social Science. He earned his BA from the University of Western Ontario and his MA and PhD from the London School of Economics. Prior to his current role, he held positions at the University of Southampton and Birkbeck, University of London, which he left in October 2023. He is the author of several books, including Whiteshift: Immigration, Populism and the Future of White Majorities, Shall the Religious Inherit the Earth?, and The Third Awokening. His research interests include nationalism, political and religious demography, and national identity. Kaufmann is a previous guest on the podcast.

Razib and Kaufmann begin their conversation by exploring the thesis of one of his earlier works, 2004’s Rise and Fall of Anglo-America. They discuss the definition of “WASP,” White Anglo-Saxon Protestant, and cultural changes in the white American majority because of the massive immigration waves of the 19th and early 20th century. Kaufman argues that a coalition of liberal WASPs and “white ethnics” was instrumental in the eventual overthrow of the cultural hegemony of elite Protestant whites in the second half of the 20th century. Razib and Kaufman then relate the history of the WASPs to his latest book, The Third Awokening, which chronicles the rise of “cultural socialism” centered around race. Kaufman documents the potency of the ideas of the latest variant of wokeness, their traction among the youth, and argues for its historical roots in earlier forms of Anglo liberalism.

Subscribe now

Give a gift subscription

Share

Today Razib talks to Ryan P. Williams. He is president of The Claremont Institute, a position he has held since 2017. He is also a contributor to The Claremont Review of Books and started The American Mind. Williams earned a B.A. in political science and Economics from Hillsdale College and an M.A. in politics from Claremont Graduate University. He has taught American politics and political philosophy as an adjunct professor at California State University, San Bernardino and Cal Poly Pomona.

Razib and Williams first discuss the origins of The Claremont Institute and the influence of Harry Jaffa on the think-tank’s founding and current thought. They explore the influence of Jaffa’s mentor, political philosopher Leo Strauss, upon his worldview, and the differences that define the “west coast Straussianism” associated with the Institute and “east coast Straussianism.” Williams also articulates how the conservative thought of Claremont affiliated scholars and pundits differs from other movements on the right, and in particular, how it is differentiated from both neoconservatism and paleoconservatism. Razib and Williams then go over The American Mind’s decision to publish Michael Anton’s “flight 93 election” piece, and the connection of many Claremont scholars to the Trump administration and the MAGA movement.

Subscribe now

Give a gift subscription

Share

Chad Orzel is a physicist and science writer who has been blogging for nearly twenty-five years. He’s the author of four books, Breakfast with Einstein: The Exotic Physics of Everyday Objects, How to Teach Quantum Physics to Your Dog, How to Teach Relativity to Your Dog, Eureka: Discovering Your Inner Scientist and A Brief History of Timekeeping. The last is a mix of cultural and engineering history, archeology and physics, and reflects Orzel’s wide interests as reflected in his Substack, Counting Atoms.

In this episode of Unsupervised Learning Razib surveys the state of physics communication and science, as well as our broader culture’s relationship with academia. Orzel and Razib first discuss the massive success of physicist-turned-YouTuber Sabine Hossenfelder. Emerging from academic physics and associated with Lee Smolin and the Perimeter Institute, Hossenfelder has shifted from skepticism of mainstream theories like string theory to arguing that academic science as a whole must be restructured. Orzel also notes that contrarian or heterodox views in popular areas such as astrophysics and particle physics receive much more attention than applied fields like solid-state physics. Razib and Orzel reflect on how science communication has changed over the past two decades, moving from the text-driven blog era before 2010 to the rise of podcasts and video. They also discuss the many technological applications of physics in the 21st century, particularly in battery technology, an area that is transforming daily life but rarely serves as fodder for glossy popular-press treatments.

In the second half of the podcast, Orzel considers how science, and academia more broadly, have navigated the adversarial stance of the Trump administration. Razib asks whether institutional science, shaped in the post–World War II era, may be due for a major transformation, or whether it is even approaching the end of its line. Finally, Orzel addresses whether academics can regain broad public trust in the wake of the extreme politicization of the 2010s.

Subscribe now

Give a gift subscription

Share

Recently, the new embryo-selection start-up Herasight has been in the news, finally coming out of stealth. Part of the buzz is because of the public involvement of well-known geneticists and academics like Alex Young and Joe Pickrell in Herasight’s algorithm development. Additionally, Noor Siddiqi, the CEO of Orchid, a competitor to Herasight (and onetime advertiser on this podcast), was a guest on Ross Douthat’s show Interesting Times, triggering another round of conversations around embryo-selection, including in The Wall Street Journal and Breaking Points.

To hash out some opposing viewpoints, Unsupervised Learning decided to bring on two guests that stake out very different positions, Dr. James Lee, a psychometrician and behavior geneticist at the University of Minnesota, and Dr. Jonathan Anomaly, a philosopher and Herasight’s sales lead. Lee has been on the record with his skepticism of reproductive technology, writing an op-ed in The Wall Street Journal four years ago warning against the consequences of polygenic embryo selection. Meanwhile, Anomaly’s last book was Creating Future People: The Science and Ethics of Genetic Enhancement, where he advances the idea that such technologies will unlock human potential.

Subscribe now

Give a gift subscription

Share

Transcript

Transcript (PDF)

On last week’s episode of Unsupervised Learning, Razib spoke with Alex Nowrestah, a vice president at the Cato Institute and a strong advocate for expanding legal immigration. This week, he turned to the other side of the debate with Jason Richwhine, a resident scholar at the Center for Immigration Studies and a vocal supporter of sharply reducing immigration.

Richwine earned undergraduate degrees in mathematics and political science from American University, and later a Ph.D. in public policy from Harvard. Before joining CIS, he served as deputy director of the National Institute of Standards and Technology and worked as a senior policy analyst at the Heritage Foundation.

The conversation begins with an overview of the dramatic swings in U.S. immigration policy under Biden and Trump. Both note the surge of the foreign-born population in the early 2020s, with the unauthorized share now estimated at 15-16 million. Richwine faults Biden for lax border enforcement and the abuse of parole programs, and points to the comparative effectiveness of Trump’s Remain in Mexico policy. He also presses the case for a moratorium, arguing that even legal immigration must be scaled back to sustainable levels. Razib and Richwine weigh the economic and cultural consequences of high-skilled immigration and close by considering whether meaningful reform is politically possible in the years ahead.

Subscribe now

Share

Give a gift subscription

Three years ago, Razib recorded two podcasts with two immigration experts on different sides of the issue, Alex Nowrestah and Jason Richwhine. While Nowrasteh, who works for the libertarian Cato Institute as Vice President for Economic and Social Policy Studies, supports higher levels of legal immigration, Richwine, a Resident Scholar at the Center for Immigration Studies (CIS), favors lower levels of inflows into the US. The initial pair of podcasts was recorded in the midst of the massive increase in immigration that occurred after the loosening of the pandemic-era controls, resulting in the highest proportion of the foreign-born since the turn of the 20th century. Though the Biden administration tightened controls in its last year, the swell of illegal immigration resulted in a backlash that fueled the re-election of Donald Trump to the presidency and a turn in policy toward restriction. Recently, Razib recorded two conversations with Nowrestah and Richwine, revisiting the topic in 2025, and after three years of policy shifts.

Today, Razib talks to Nowrasteh about the record of the Biden administration, the pivot occurring in the first year of the Trump administration, and where he sees the Republicans going in the future. Nowrasteh addresses the reality that the Democratic administration's lack of interest in controlling illegal flows resulted in anger and frustration at migration in general, and emphasizes the importance of borders and rules in allowing for legal immigration. Razib and Nowrasteh also discuss the controversy over H1-Bs, the role that skilled immigration plays in buttressing American power, and the conflicts on the Right regarding how immigration policy relates to geopolitics. They also explore the relationship between immigration and population, and how both connect to urban policy and economic growth.

Next week, Razib will post his episode with Jason Richwine.

Subscribe now

Give a gift subscription

Share

Today on Unsupervised Learning, Razib talks to John Hawks, a paleoanthropologist who has been a researcher and commentator in human evolutionary biology and paleoanthropology for over two decades. With a widely read weblog (now on Substack), a book on Homo naledi, and highly cited scientific papers, Hawks is an essential voice in understanding the origins of our species. He graduated from Kansas State University in 1994 with degrees in French, English, and Anthropology, and received both his M.A. and Ph.D. in Anthropology from the University of Michigan, where he studied under Milford Wolpoff. He is currently working on a textbook on the origins of modern humans in their evolutionary context. Hawks has already been a guest on Unsupervised Learning three times.

In this episode, Razib and Hawks focus on a very specific question: What were the different contributions to the heritage of modern humans in a world more than 200,000 years ago that was inhabited by at least half a dozen hominin species? First, Hawks takes us back to the year 2000 and his early work extending a more multiregional framework of human evolution, exploring what could be gleaned from the archaeological and paleontological record. Then Razib and Hawks discuss the ancient DNA revolution and the discovery that modern humans had ancestry from Neanderthals, as well as from an entirely new species, the Denisovans. They also examine the fact that, unlike Neanderthals, Denisovans appear to have been separated into very different regional populations that made distinct contributions to various modern populations. Razib also asks Hawks about the discovery of new pygmy human species in Luzon, as well as the current state of research on Homo naledi in South Africa and the Hobbits of Flores. Hawks contends that DNA will likely be extracted from all these lineages at some point and, if not, protein sequence data may be obtained. This would finally give researchers the statistical power to evaluate the possibility of extremely archaic admixture events. Hawks and Razib also address the potential role of natural selection driven by introgressed genes from sister lineages of humans and how this shaped modern variation.

Subscribe now

Give a gift subscription

Share

Today Razib talks to Noah Millman. Millman is an American screenwriter and filmmaker, as well as a political columnist and cultural critic based in Brooklyn, New York. He is the film and theater critic for Modern Age; previously he was a columnist for The Week (2015–2022) and a senior editor at The American Conservative (2012–2017). Millman writes the newsletter Gideon’s Substack, and his work has also appeared in outlets such as The New York Times and Politico. He graduated from Yale University and initially worked on Wall Street for 16 years, starting in a hedge fund's mail room, before leaving after the financial crisis to pursue creative endeavors full-time. Millman has been a producer on seven films, and written three and directed three. His most recent film is Resentment, and he is working on a novel, Fables of a Jewish Century.

Razib and Millman begin their conversation discussing their history as bloggers who began writing early in the first decade of the century, in the wake of George W. Bush’s invasion of Iraq. Millman discusses his disillusionment with neoconservatism, and his evolution into a moderate, if heterodox, Democrat. They also discuss their positionality in a political commentary landscape that has radically shifted over the last twenty years, and what it’s like to be strongly partisan. They discuss how their views of religion have changed, especially in the wake of the New Atheist movement after 9/11 and the emergence of psychedelic spirituality in the 2020s. Millman articulates his views as a Jew whose own theological commitments are minimal, stating that he believes that the “Hindus are right about God” but John Calvin was probably right about humans.

In the second half of the discussion, they pivot to the arts, beginning with how film as a medium has developed over the last generation, from the high tide of independent films in 1999 and through the “comic book” movie heyday of the 2010s, and on finally to the reemergence of more classic movies like Tom Cruise’s Top Gun: Maverick 2 and Brad Pitt’s F1. Razib argues that the Marvel universe exhausted its creative possibilities, and the same content no longer compels the younger generations, especially in a 90-minute format. Millman addresses whether film as a medium has reached the end of the line as a mass medium, and how fan-culture and “stan” culture has transformed the experience of the arts. He also asserts that cultural fragmentation is driven by technology, as consumers have a much greater range of options in their choices than in the past. Millman observes that as top-down cultural dynamics have collapsed, shifts are now driven by bottom-up drives. He also argues that movies will ‌continue to be a major art form because filmmaking is now far cheaper than it was in the past, but he is not optimistic about the future of mass-market tent-pole films that can transcend myriad fan subcultures. Movie studios still do not know which films will become hits and which will flop, even the magic of Pixar and Marvel Studios are no longer a sure thing. In fact, Millman argues that fragmentation has masked the revival of art forms like the novel. As the gatekeepers are gone, many consume low art, with middle-aged people reading copious amounts of YA fiction. Millman argues that any aspiring artist needs to grapple with the competitive realities of the new attention economy. Technology has made it easier for anyone to create art because new tools are cheaper and self-publishing is now a real option for writers. However, all of this unleashed creativity is competing for the same amount of funding, support and a relatively fixed audience.

Subscribe now

Give a gift subscription

Share