Allow us to introduce a fresh perspective in the realm of economic thought here at Econlib, Sam Enright. Sam’s expertise lies in innovation policy at Progress Ireland, a Dublin-based independent policy think tank, and he also oversees a publication known as The Fitzwilliam. Notably, his personal blog features a widely read link roundup that provides concise insights on the most intriguing content he has consumed monthly. His ‘linksposts’ have garnered a playful reputation for their remarkable length; what follows is a condensed edition of his Links for October.
Blogs and Short Links
1. Ava Huang delves into the friendship theory of everything. (I find this theory quite compelling.)
2. The debate isn’t about choosing between environmental sustainability and economic growth; both can coexist.
3. Surprisingly, free market economics is showing promising results. As Noah Smith notes, the Argentine economy’s recent improvements under Milei are likely the result of orthodox macroeconomic stabilization policies. Only time will tell if other reforms succeed. Could we rephrase this as “Are We All Indebted to the IMF?”
4. The U.S. is one of only two countries that tax non-resident citizens on global income, alongside Eritrea. Check out this wiki on the other financial and legal hurdles faced by American expatriates, including restrictions on investing in the UK’s best tax vehicle, the Individual Savings Account. This insight comes courtesy of Bogleheads, a community deeply devoted to the teachings of John Bogle.
5. Sooner or later, we will all embrace congestion pricing.
6. Sebastian Garren offers a rapid exploration of Chilean economic history. Expect to hear more on this topic soon:
“Gratitude to Sam Enright and The Fitzwilliam for inspiring this journey.”
Music and Podcasts
7. Chakravarthi Rangarajan discusses the evolution of Indian monetary policy post-1991 liberalization. I was surprised to learn about the significance of fiscal dominance in India prior to the 1990s (or even to understand what it entails).
8. Dmitri Shostakovich’s Symphony No. 8 accompanied by the Sticky Notes episode. This symphony has a darker complexity compared to the more uplifting Symphony No. 7, which might have been a more suitable introduction. The cautious optimism reflected in the music about the Red Army’s progress is palpable, and I find composers associated with specific historical contexts to be particularly engaging (#8 premiered in 1943, while #7 premiered in 1942).
9. Tabla Beat Science’s Tala Matrix showcases another one of Zakir Hussain’s endeavors. If you haven’t yet read Shruti Rajagopalan’s obituary for Zakir, I highly recommend it as the best piece I’ve encountered on Indian music.
10. Richard Sutton, a pioneer in reinforcement learning, expounds on his belief that LLMs are reaching a dead end. When will I comprehend my own “bitter lesson” that I ought to read transcripts instead of trying to follow these podcasts audibly?
Papers
11. P.W. Anderson’s More is Different: Broken Symmetry and the Nature of the Hierarchical Structure of Science is a paper I’ve often heard referenced but never read until now. Anderson advocates for anti-reductionist pluralism, a stance that seems aligned with Daniel Dennett’s arguments in Real Patterns. My previous sympathies towards the notion that “chemistry is merely applied physics” appear philosophically naive upon reflection. I also came across a 50-year retrospective by Steven Strogatz et al. It’s fascinating sociologically that a non-philosopher could produce such a widely discussed philosophical paper in just four pages.
12. Richard Sutton’s The Bitter Lesson is another must-read. The core insight is as follows:
“The most significant takeaway from 70 years of AI research is that general methods leveraging computation are ultimately the most effective, and by a significant margin… We must accept the bitter lesson that attempting to embed our own cognitive frameworks into AI does not yield lasting results.”
Sutton’s work illustrates that the more general strategies for building AI—which prioritize computational scaling and avoid the symbolic methods of GOFAI—were once dismissed as “weak methods.” It’s no wonder skepticism prevailed at the time.
13. David Silver and Richard Sutton’s Welcome to the Era of Experience. I engaged with this accessible essay through a machine learning reading group at the delightful Mox coworking space. Their 90/30 Club reads through Ilya Sutskever’s list of the 30 essential AI papers that cover 90% of the field’s critical concepts. Although they have since moved on to other papers, I thought I might struggle to keep pace with the brilliantly “cracked” San Francisco engineers; however, listening to Sutton on the Dwarkesh podcast helped me prepare.
To be candid, the overwhelming vibrancy of the Bay Area can be overstimulating, which contributed to my low mood during my visit. In contrast, Dublin feels like a place where you can connect with nearly everyone who shares your interests. Small communities have their own unique charm.
Ultimately, Silver and Sutton argue that AI is reaching its limits in learning from human-generated data. Moving forward, AI will rely heavily on experiential learning, embracing trial and error. Achieving superintelligence will necessitate a significant paradigm shift, heavily depending on reinforcement learning. The pivotal graph from page 6 encapsulates this idea:
Figure 1: A sketch chronology of dominant AI paradigms. The y-axis suggests the proportion of the field’s total effort and computation that is focused on reinforcement learning. From Silver and Sutton, “Welcome to the Era of Experience.”
They propose a more nuanced vision where advanced AI will be guided by human values and feedback, a concept I found somewhat intricate. This paper, released in April, will eventually feature in a book titled Designing an Intelligence, which I plan to pre-order once a release date is set.
This content is quite intellectually stimulating, and as I wrap up this discussion, I will leave you with a recent nugget of wisdom from my friend David:
They should refer to the opposite of an AI doomer as a sloptomist.
- You can read the full version of this post here.
[1] I was reminded of a Marginal Revolution comment from 2023 suggesting that John Bogle deserves the (hypothetical) Nobel Prize for his contributions to economics.
[2] The name David Silver didn’t ring a bell, but I now recognize him from that remarkable documentary on AlphaGo.
