Razib Khan's Unsupervised Learning

Today Razib talks to Aria Babu, a British think-thank professional who is part of the growing number of young men and women who are taking an interest in population decline and promoting pro-natalism. Babu has a degree in chemistry from University College London, and has long worked in areas related to the study of economic growth and entrepreneurship. Prior to her interest in pro-natalism Babu held conventional views about population growth and its ties to environmental alarmism. But she quickly saw that actually fertility is crashing worldwide, and with that there might be dire economic and social consequences. If that trend is left unchecked, she foresees a worst case scenario of massive economic decline and the replacement of our riotously varied modern civilization by a select few narrow subcultures, like the Amish or Somalis, who continue to favor reproduction as a social value. On the state level, declining populations will likely lead to the rise of culturally stagnant and politically authoritarian societies reminiscent of The Children of Men.

Babu and Razib also discuss what it is like living as an urban professional in Britain in 2024. While the fact that the UK has one megacity can lead to disproportionate focus on London, Babu points out that it allows the entire nation’s intellectual and cultural class to be in close proximity, resulting in powerful synergies. She also argues that the problem in the UK is not immigration, but insufficient housing for larger populations and the lack of a system to allow in very skilled and value-add migrants. Rather than integration into the EU or an American-system, Babu favors an approach closer to Singapore, where the UK goes its own way and crafts its policies to take advantage of specific opportunities offered by blindspots in EU or American politics.

If you have a sibling with autism, your future child’s risk for an autism diagnosis is increased by a factor of 2 to 3.5×. Orchid’s whole genome embryo reports can help mitigate your child’s risk by screening for over 200 genetic variants definitively linked to autism and other neurodevelopmental disorders. Discuss your situation with a genetics expert.

On this episode of Unsupervised Learning Razib talks to Louise Perry. A British journalist known for her commentary on feminism and gender issues, Perry is the author of the book The Case Against the Sexual Revolution. She also contributes to The New Statesman, UnHerd, and The Daily Mail, and has a Substack at Maiden Mother Matriarch. Perry is a graduate of University of London’s School of Oriental and African Studies, with a degree in anthropology.

Perry and Razib first discuss Britain’s current housing crisis, the reasons and possible solutions. Though the Office for National Statistics estimates the UK’s population at 67.1 million, Perry believes that the true number is likely higher because individuals who are present illegally or have a “gray” status are unlikely to respond. But even this population would make the UK over eight times more densely populated than the US, with England being 13 times denser. In fact, England’s population density is similar to India’s. Perry also brings up the reality of massive immigration flows over the last few years; where before 2020 net migration was around ~200,000 per year, since 2021 the figure has been closer to ~500,000. Additionally, many of these immigrants are placed in “social housing,” subsidized or owned by the government. Perry also points out that the legal regulations in Britain stipulate that about 30% of new developments be allocated for social housing, which incentivizes incumbent homeowners to block construction. Additionally, the rate of population growth is much higher than the British construction industry’s capacity to keep up with the theoretical demand. The UK does not produce enough bricks, nor does it have the labor pool of homebuilders.

The conversation continues to a broader discussion of the ennui in modern British society. Perry asserts that a major problem driving the housing crisis is that the UK has only one major city, London, and any professional who wants to settle in a more affordable region must also take a major salary cut. Setting aside London, and its economic engines of finance and commerce, Perry characterizes the rest of the UK as more akin to a developing Eastern Europe nation. She also believes that the next decade will see a mass flight of the upper-middle-class, the primary tax base of the state. Perry herself has Australian citizenship through her parents (who immigrated from Australia to the UK), while her husband has an American mother. Her situation is common to many upper-middle-class Britons, who have connections to Canada, the US, New Zealand and Australia. Perry believes this is one reason the British political culture is not reforming itself: so many have in the back of their head that they can jump ship if it starts sinking. Ultimately, her thesis is that British openness and intellectual curiosity make the national character a poor seedbed for nationalism, and it may be inevitable that the UK is caught up and tossed about in a vortex of globalization.

If you have a sibling with autism, your future child’s risk for an autism diagnosis is increased by a factor of 2 to 3.5×. Orchid’s whole genome embryo reports can help mitigate your child’s risk by screening for over 200 genetic variants definitively linked to autism and other neurodevelopmental disorders. Discuss your situation with a genetics expert.

On this episode of Unsupervised Learning Razib talks to archeologist and historian Bryan Ward-Perkins about his 2005 book The Fall of Rome: And the End of Civilization. Ward-Perkins was born and grew up in Rome, a son of architectural historian and archaeologist, John Bryan Ward-Perkins. Educated at Oxford University, Ward-Perkins eventually became a fellow of Trinity College at the same university, from which he has since retired. An archaeologist with a deep interest in economic history, Ward-Perkins’ standout book The Fall of Rome was to a great extent a restatement of traditional understandings of the Roman fall in the wake of academic revisions stimulated by Peter Brown’s 1989 World of Late Antiquity: AD 150-750. Ward-Perkins scholarship focuses on the outputs of economic production: fine pottery, grand public buildings and copious coinage. In contrast, Brown and his fellow travelers tended to focus on religious innovation and creativity in the centuries coincident with Rome's fall. The Fall of Rome documents in crisp, dense prose the material collapse attendant with the dissolution of the Western Empire in the late 5th and 6th centuries, from the vanishing of pottery in Britain to the cessation of the construction of massive buildings across  the Italian peninsula.

Razib also asks Ward-Perkins his opinions on his colleague Pete Heather’s book The Fall of the Roman Empire: A New History of Rome and the Barbarians.  Ward-Perkins sees Heather’s work as complementary; while Ward-Perkins is interested in the material aspects of everyday Roman life, Heather documents and narrates the diplomatic and military affairs of the Roman elite. Ward-Perkins also comments on Chris Wickham’s work in books like The Inheritance of Rome: Illuminating the Dark Ages 400-1000, which outlines how the Roman and post-Roman states differed, in particular, the disappearance in Europe of professional soldiers paid in currency, rather than feudal levies. They also discuss  Walter Scheidel’s Escape from Rome: The Failure of Empire and the Road to Prosperity, and whether Roman citizens were actually materially better  off than their medieval successors. Ward-Perkins also gives his estimation of the time measured in centuries until Western Europe reattained Roman levels of social, technological and political complexity.

If you have a sibling with autism, your future child’s risk for an autism diagnosis is increased by a factor of 2 to 3.5×. Orchid’s whole genome embryo reports can help mitigate your child’s risk by screening for over 200 genetic variants definitively linked to autism and other neurodevelopmental disorders. Discuss your situation with a genetics expert.

On this episode of Unsupervised Learning Razib welcomes back a returning guest, J. P. Mallory, to discuss his reaction to the recent preprint The Genetic Origin of the Indo-Europeans. Mallory is the author of In Search of the Indo-Europeans: Language, Archaeology, and Myth, The Oxford Introduction to Proto-Indo-European and the Proto-Indo-European World and The Tarim Mummies: Ancient China and the Mystery of the Earliest Peoples from the West. He is also a retired professor from Queen’s University Belfast in Northern Ireland. An archaeologist who trained under Marija Gimbutas, Mallory has long used linguistics to complement his disciplinary training in archaeology to understand the origin and location of Indo-European languages.

Though Mallory admires The Genetic Origin of the Indo-Europeans, he still thinks more work needs to be done to pinpoint the original homeland of the Yamnaya or their ancestors. The fact is that the preprint remains somewhat vague in its final conclusion, and more work is needed to make sure the populace acquires the same level of community. Mallory also discusses the challenges inherent in interdisciplinary work, synthesizing archaeology, linguistics and now genetics. He believes that a key to grasping the emergence of pre/proto-Indo-European is tracing lineage groups through their Y chromosomes, as the genetics, mythology and anthropology indicate that pre/proto-Indo-Europeans were quite patriarchal and patrilineal. Though Mallory is hopeful that we are making progress on the topic of Indo-Europeans he worries that the fraught situation of disciplinary rivalries will retard synergy, where archaeogenetics engages in excessive imperialism vis-a-vis archaeology and linguistics.

On this episode of Unsupervised Learning Razib talks to professor Sean Anthony about his book Muhammad and the Empires of Faith: The Making of the Prophet of Islam. Anthony is a historian in the Department of Near Eastern Languages and Cultures at The Ohio State University. He earned his Ph.D. with honors in 2009 at the University of Chicago in Near Eastern Languages and Civilizations, and has a mastery of Arabic, Persian, Syriac, French, and German. Anthony’s interests are broadly religion and society in late antiquity and medieval Islam, early canonical literatures of Islam (Koran and Hadith) and statecraft and political thought from the foundational period of Islam down to the Abbasid Caliphate over a century later.

Razib and Anthony discuss the state of the controversial scholarship about the origins of Islam, which often comes to conclusions that challenge the orthodox Muslim narrative. This earlier generation of scholars, like Patricia Crone, challenged the historicity of Muhammad, the centrality of Mecca in early Islam and even the distinctive religious identity of the early 7th century’s Near East's Arab conquerors. This revisionist school serves as the basis for Tom Holland’s 2012 book, In the Shadow of the Sword: The Birth of Islam and the Rise of the Global Arab Empire. While Holland’s work was an accurate summary of research before the 2010’s, Anthony argues that since then new findings have updated and revised the revisionism itself. A Koran dating from the mid-7th century seems to confirm the antiquity of this text and traditions around it, while contemporaneous non-Muslim sources refer to Muhammad as an Arabian prophet. While it is true that coinage did not bear the prophet’s name until the end of the 7th century, it may be that earlier generations of scholars were misled by the lack of access to contemporary oral sources themselves necessarily evanescent. Razib and Anthony also discuss whether the first Muslims actually self-identified as Muslims in a way we would understand, as opposed to being a heterodox monotheistic sect that emerged out of Christianity and Judaism. Though classical Islam qua Islam crystallized under the Abbasids after 750 AD, it now seems quite clear that the earlier Umayyads had a distinct identity from the Christians and Jews whom they ruled.

On this episode of Unsupervised Learning Razib talks to Nikolai Yakovenko, a returning guest to the podcast, about his new AI startup, DeepNewz, and the state of the LLM-driven AI landscape circa the summer of 2024, where we are in relation to earlier expectations and where we might be in the next decade. Yakovenko is an AI researcher who has worked at Google, Twitter and Nvidia, and is now a serial entrepreneur. He is also a competitive poker player. He currently lives in Miami, Florida, though he travels frequently to America’s numerous “ideaopolises,” from San Francisco, Austin, Boston to New York City.

Razib and Yakovenko discuss the reality that in the middle of 2024 here they are again, chatting about the world on a podcast, a scenario not everyone anticipated in the heady days of December 2022/January 2023 when the more overheated visiony tech  imaginations swirled with expectations that the advent of artificial general intelligence, the “machine god,” was imminent. Though OpenAI’s GPT 3/3.5 was a leap ahead of GPT 2, GPT 4/4o has been a less spectacular advance. One of the major unforeseen aspects of the LLM-based framework in AI has been the returns to scale in terms of training data, but the last year and a half have not seen any great quantum jumps. The paradigm-shifting revolution that was promised has not arrived. Though AI has increased productivity on the margins, and certain artistic professions and translators have been decimated, overall, it is still a technology with more promise than realized outcomes. Yakovenko points out that AI-driven music creation produces serviceable outputs, but not great masterpieces. To test this, Razib used Suno to create a song “Nikolai’s Dream” within 5 minutesmid-conversation. Though mildly catchy, Suno’s lyrical styling elicited more amusement than awe.

Yakovenko notes the importance of the LMSYS Chatbot Arena Leaderboard to get a sense of the performance of the various LLM projects. Using the feedback of participants, it produced an updated ranking of chatbot performance. The top ten models are from OpenAI, Anthropic, Google and a Chinese vendor. Yakovenko notes the compression, with a very tight distribution of scores at the top. It turns out that OpenAI is not running away from their competition contra  its brand visibility likely being two orders of magnitude greater than Anthropic’s. This brings to the fore the reality that these AI technologies have been viewed as both scientific research projects and potential business and consumer products.

Finally, Yakovenko and Razib talk about DeepNewz. While most LLM-based chatbots tend to exclude very recent data and events, Yakovenko had the idea of creating DeepNewz to aggregate and assemble the breaking news in various categories like science and sports. Instead of a top-down query of news in various categories, the idea behind DeepNewz is to both cater to your preferences in terms of what you might find interesting, but also to surface stories that you might not know you might be interested in, adding more value.

Related: David McKay: AI and the end of the world as we know it and Nick Cassimatis: fear not AI, this too shall pass.

On this episode of Unsupervised Learning Razib talks to Chad Niederhuth, an erstwhile academic plant geneticist now working in industry. Niederhuth and Razib discuss the reality that in 2024 it is often human genetics that gets the glory, even though experiments on plants go back to the field’s very origins with Gregor Mendel and his peas. Niederhuth’s original training is in molecular genetics, and they discuss the relevance of differences in basic biological machinery between plants and animals, for example the reality that the former have chloroplasts while the latter have mitochondria. They also extensively discuss the flexibility and variation across plants in terms of size and organization of the genome; plants much more often deviate from a diploid two-gene-copy setup than animals, and their range of genome size is enormous. While the smallest plant genome is 61,000,000 bases, the largest is 148,800,000,000 bases (2,400 times larger than the human genome). 

Razib and Niederhuth discuss the flexibility and utility of plants in basic genetic research, but also in the applied agricultural context. Though classic techniques of selection are still relevant, more and more researchers are using genomic methods that look at variation at the DNA level to predict traits in the next generation, and so allow for more robust and productive cultivars. Razib also notes that public queasiness over genetic engineering in animals, let alone humans, does not seem to apply equally to plants, meaning that GMO techniques can be perfected in crops first before transferring to animal or medical contexts.

Finally, Niederhuth talks about his transition from being faculty at a research university to a scientist in the private sector. Overall Niederhuth is happy because his pay is greater, and his responsibilities are narrower and more focused. While as a professor he had to also split his time between teaching and serving extensively on committees, his current position is focused entirely on the research he finds so gratifying. Razib and Niederhuth also discuss the politicization of academic science that has occurred over the last 15 years, and the institution’s future prospects.

On this episode of Unsupervised Learning Razib talks to Jonathan Keeperman, an former lecturer in writing at UC Irvine and proprietor of Passage Press. Keeperman also posts on the internet under what was until recently an anonymous pseudonym, Lomez. Unlike many anonymous accounts on X, “Lomez” developed a decade-long identity, to the point where Keeperman wrote articles under that name for publications like First Things, The Federalist and The American Mind.

Razib and Keeperman talk about what it is like to go from someone with distinct and separate identities, a well-developed online life as well as a fairly conventional offline world, and how to reconcile them when they collide. Keeperman talks about the peculiar and often offensive scripts and modalities of the world of anonymous commentators, whose goals seem to be to have hidden discussions in plain sight, hiding their discourse through shock and obfuscation, and how difficult it can be to communicate this reality to people with more conventional outlooks. 

Keeperman admits that he understood that at some point his anonymity would be stripped away from him, but admits that it is still a difficult path to negotiate. The Lomez identity was unabashedly on the political Right, but as an academic and writing lecturer he was much more discreet about his views, and many of his friends and acquaintances were shocked as to his true politics. Keeperman’s father was a liberal and a Jewish American, so many of his relatives would no doubt have been surprised by his political commitments. 

Razib also asks Keeperman what exactly an MFA means as a credential, and what it teaches you. Though he does not think much of the credential itself, Keeperman explains that the MFA is a terminal degree for many interested in writing and literature, two loves that pulled him away from a life in the corporate world. He explains that one of his goals in entering the writing profession was to bring a masculine sensibility that he feels has been marginalized in the world of creative writing, which is today dominated by women. Razib and Keeperman talk about the marginalization of certain masculine values of vigorous competition and biting debate in many parts of the culture-producing industries, and how Passage Press is an attempt to cultivate voices that otherwise might not find  a platform. In this vein, Keeperman ends by asserting the importance of free speech for all views, from the most offensive to the most anodyne, as an essential part of American culture and the life of the mind. If you have a sibling with autism, your future child’s risk for an autism diagnosis is increased by a factor of 2 to 3.5×. Orchid’s whole genome embryo reports can help mitigate your child’s risk by screening for over 200 genetic variants definitively linked to autism and other neurodevelopmental disorders. Discuss your situation with a genetics expert.

On this episode of Unsupervised Learning Razib talks about religion with Ryan Burge, professor of political science at Eastern Illinois University, and author of The Nones: Where They Came From, Who They Are, and Where They Are Going and 20 Myths About Religion and Politics in America. Burge also has a Substack, Graphs about Religion, where he posts the latest data on trends in American society.

First, Razib asks Burge to outline the wave of secularization that has impacted American society over the last 25 years, from its causes to its potential end. Burge points out that mainline Protestantism looks to be on the verge of extinction in the 21st century, while evangelical Christianity saw its high point in the 1980s. Then they talk about the nature of religiosity in America, and Burge asserts that in some ways the rise in “religious nones” is probably just social desirability bias. With the fall of Communism, atheism and irreligiosity lost some of their  negative connotations, and more and more people have “come out of the closet” or just accepted their revealed preferences. Razib also asks if Christianity will become a minority religion in the US, and if it is true that people become more religious as they get older. Finally, they discuss extensively the connection between religion and politics and how that drove the rise in defections from Christianity, and Burge talks about how the 21st century will see a normalization of extreme polarization between Christian conservatives and secular Americans. If you have a sibling with autism, your future child’s risk for an autism diagnosis is increased by a factor of 2 to 3.5×. Orchid’s whole genome embryo reports can help mitigate your child’s risk by screening for over 200 genetic variants definitively linked to autism and other neurodevelopmental disorders. Discuss your situation with a genetics expert.

On this episode of Unsupervised Learning Razib discusses the idea of “lost civilizations,” the possibility that there were complex societies during the Pleistocene Ice Age. This topic recently rose to salience after a dialogue between writer Graham Hancock and archaeologist Flint Dibble on Joe Rogan’s podcast. Hancock is a longtime guest on Rogan’s show and he promotes a theory that an advanced “lost civilization” during the Ice Age left remnants of its culture across the world, for example the pyramids of both Egypt and the New World. In the exchange, Dibble, a vigorous online critic of pseudo-archaeology came back with scientific orthodoxy; civilization emerged after the end of the Ice Age, and there is no evidence for anything during the Pleistocene.

Razib explains why evidence from biology makes it clear that Hancock is almost certainly wrong. No massively advanced global civilization during the Ice Age left its imprint across the world. Though the archaeological evidence is strong, the data from DNA is even more unambiguously robust and informative. But then Razib steps back and asks what “civilization” is, and presents the possibility that stillborn cultures might have existed during the Ice Age. Civilizations that left no descendants and barely any archaeological footprint. He also argues that the modern conception of civilization starting in Sumer and Egypt 5,000 years ago is simplistic, and American ideas about an arrow of history ascending onward and upward tend to be misleading as a guide to the past. Though the orthodox view is mostly right, there are always gaps in our knowledge and surprises around the corner. Graham Hancock’s enthusiasm for what we can know is commendable, his conclusions long ago outpaced his evidence. In the future, understanding the past will be done with even more powerful tools, but we must proceed with humility upon the foundations of all we know while acknowledging what we don’t.

Related: Why Civilization Is Older Than We Thought and Paradise Lost.

If you have a sibling with autism, your future child’s risk for an autism diagnosis is increased by a factor of 2 to 3.5×. Orchid’s whole genome embryo reports can help mitigate your child’s risk by screening for over 200 genetic variants definitively linked to autism and other neurodevelopmental disorders. Discuss your situation with a genetics expert.