With Grok, there would be terabytes of X garbage (and non-garbage) used for training material. Meta used LibGen, which is 7.5 million books and 81 million research papers. Among these books are many math textbooks. What’s fascinating to me is that ChatGPT will say, “No, that’s not correct” in response to a technical proposition, but for other topics it will riff in indefinite and complex ways in response to complex queries. This makes sense from a business perspective: Technical users won’t use these services if the answers don’t speed their work along. On the other hand, if conversation is the user’s goal, it makes sense to keep riffing but without challenging the user too much.
From: Friam <[email protected]> on behalf of steve smith <[email protected]> Date: Sunday, April 6, 2025 at 3:08 PM To: [email protected] <[email protected]> Subject: [FRIAM] Confabulation and other acts of bad-faith EricS - As a creature of morbid fascination, I can hardly hear someone else's argument for/against anything without weaving my own "bad faith" story through it... I believe it is a bit of an inoculation against "twisted thinking" for myself. My experience with LLMs (too much really) has slowly brought me to believe/understand what many vehement naysayers declared early on. I think what you describe here is as effective of an articulation of this phenomenon as I have heard. I have a memory of some comic-book or streaming-series character whose "superpower" was to whisper a rumour in your ear which would then become absolutely real to you, causing you to take the most self-destructive act possible in the moment. I believe she would begin out loud with "I heard a rumour..." followed by the ear-whisper. Q-drops and MAGA nonsense really fits this trope? I also remember a theme in a movie describing a new drug's effects as "put it in your eyes and it will tell you lies"... I think there is something acutely compelling about a blatant/obvious lie? I understand James never intended an ends-means conflation in his Pragmatism but I feel as if many "pragmatists" embrace such (at least quietly)? This seems like a hallmark of many Trump apologists. My recent fascination with Solms and Friston lead me to a little of Solm's earlier work in the neuropsychology "confabulation" which hints in this direction as well. I do think that LLMs can easily fall into a vein (onto a manifold) of disingenuity which is (paradoxically) as genuine in the training set as anything. Right from the first day I chatted with my "new bar friend" I recognized that there was a huge propensity and skill at tracking what I was talking about and feeding me things which aligned or interpolated or extrapolated it. The more elaborate my prompt, the more elaborately aligned the response. And like my fictional super-hero... I now hear it silently preface everything it says to me with "I heard a rumour..." - SteveS On 4/6/25 2:46 PM, Santafe wrote: There is an interesting direction here, Steve: In the beginning there was whatever-training data, and there was hallucination with many resonances of things that looked “real” to people Practical people (always so ingenuous) wanted to figure out how to sort through the mess of the training data to get reasoning that was sound, anchored in whatever “rules of reasoning” are either in the training data to be pulled out, or else put in by hand because the Practical people have work they want to get done Socially concerned people (this is like a Dylan Thomas poem) were concerned that, even if the AIs weren’t “hallucinating”, but just trying to take the best average that could be taken over the mishmash of what was in the training data, they would still deliver distorted outcomes because the training data contained a lot of, or even was skewed toward, various distorted views But bad-faith actors do a different thing, for which Glen has taught me the term of art is “motivated reasoning”. And they (all people) are wizards at it. They don’t average or coherently hallucinate over the broad training data, but mask and twist and distort and cull, (and then/in the process) average or hallucinate over those extractions I wonder how much of the subtlety of human cognition is to be found not in honesty and sense, but in the infinite varieties of bad-faith dishonesty and nonsense around which “coordination” (so to speak) can be organized. All happy families etc. Had WIlliam James psychologist lived long enough, and had wider interests, a second book for him to write Will AI designers get to claim a further step toward “understanding human intelligence” when their creations can spontaneously and autonomously do a substantive job mimicking the varieties of human dishonesty and bad faith? Eric On Apr 7, 2025, at 2:47 AM, steve smith <[email protected]> <mailto:[email protected]> wrote: new prompt for Trump Apologists: "Rewrite the ABM in a manner which makes the current US Trade, Immigration, and DEI policies look like a brilliant move. Push statistics, charts and rhetoric widely across the internet. Shoot the dog and goat, deport some folks you don't like the look of and go declare yourself winner of the golf tournament at your own golf course." On 4/5/25 3:47 PM, Stephen Guerin wrote: Here's a web version of Marcus's prompt: "write an html/javascript ABM of U.S. trade that considers the deficit, debt, trade imbalances, and international capital flows." this runs directly in the Claude web context with no downloads or setup. And here is copying the page and deploying to play with the dog. https://guerin.acequia.io/sandbox/marcus-claude-economy.html <https://guerin.acequia.io/sandbox/marcus-claude-economy.html> Old joke applies: ""It's not that the dog talks well, it's that it talks at all." _________________________________________________________________ Stephen Guerin CEO, Founder https://simtable.com <https://simtable.com> [email protected] <mailto:[email protected]> [email protected] <mailto:[email protected]> Harvard Visualization Research and Teaching Lab mobile: (505)577-5828 On Sat, Apr 5, 2025 at 3:24 PM Marcus Daniels <[email protected]> <mailto:[email protected]> wrote: • Download Github Copilot. Add Python module. • Get a Claude Console subscription. Select Claude Sonnet 3.7 in Github Copilot. • Open the Chat window and select Agent. • Enter “Can you write an ABM of U.S. trade that considers the deficit, debt, trade imbalances, and international capital flows. Watch project be populated. • Press Run. • Play with dog. From: Friam <[email protected]> <mailto:[email protected]> on behalf of Pieter Steenekamp <[email protected]> <mailto:[email protected]> Date: Saturday, April 5, 2025 at 3:45 AM To: The Friday Morning Applied Complexity Coffee Group <[email protected]> <mailto:[email protected]> Subject: Re: [FRIAM] Fwd: CSSSA April Webinar I listened to the above webinar on Agent-Based Modeling (ABM) in Economics and Finance, and would like to share a few reflections: It would be wonderful to see this discipline develop further. In fields like transportation planning, ABM has already matured to a point where it arguably outperforms traditional top-down approaches. A few years ago in South Africa, ABM was used in planning a major public transport upgrade in Gauteng. I followed the project closely and, in my view, it was a great success. My friend Johan Joubert led the modeling effort, and the results were impressive. But let me return to ABM in the context of Economics and Finance. I understand that building effective ABM models in these domains is significantly more challenging than in transportation. Yet, imagine the value if it becomes a reality. The U.S., for example, is grappling with major economic issues: a growing federal deficit, mounting government debt, a persistent trade imbalance, and a population—especially the lower half—feeling economically left behind. Wouldn’t it be exciting if ABM could contribute to practical, data-driven solutions to these kinds of complex problems? I was a bit disappointed that the webinar didn’t mention the potential integration of ABM with AI models in the context of Economics and Finance. There’s so much potential here. Large language models (LLMs) could help generate more nuanced and adaptable ABM scenarios, while ABM could provide rich, dynamic environments to train and refine AI models—especially reinforcement learning systems aimed at supporting policy-making. I’m optimistic that this kind of synergy will emerge in the near future. On Sat, 29 Mar 2025 at 09:53, Stephen Guerin <[email protected]> <mailto:[email protected]> wrote: ---------- Forwarded message --------- From: Computational Social Science Society of the Americas <[email protected]> <mailto:[email protected]> Date: Fri, Mar 28, 2025, 7:10 PM Subject: CSSSA April Webinar To: <[email protected]> <mailto:[email protected]> View this email in your browser Dear CSSSA members, We are very excited to host Robert Axtell and Doyne Farmer discussing “Agent-Based Modeling in the Economics and Finence” in our 2025 webinar series on Wednesday, April 2nd, at 10 am (ET) . Click here to register for the webinar Abstract In a long paper in the Journal of Economic Literature Axtell and Farmer review agent-based modeling (ABM) in economics and finance and highlight how it can be used to relax conventional assumptions in standard models. ABM has enriched the understanding of markets, industrial organization, labor, macro, development, and environmental economics. In finance, substantial accomplishments include understanding clustered volatility, market impact, systemic risk, and housing markets. A vision is presented for how ABMs might be used in the future to build more realistic models of the economy. Hurdles that must be overcome to achieve this are discussed. Their paper includes more than 800 references including many from adjacent fields. Biographs Professor Axtell is the author, with Joshua Epstein, of Growing Artificial Societies: Social Science from the Bottom Up (MIT Press). His research has appeared in Science, Nature, Proceedings of the National Academy of Sciences, as well as in leading field-specific journals such as The Journal of Economic Literature, The American Economic Review, The Economic Journal, and many others. His research has been reprised in newspapers (e.g., Wall St. Journal, Los Angeles Times, Washington Post) and science magazines (e.g., Scientific American, Technology Review, Wired). For the past decade he has been using microdata on individuals to build large-scale models of the Financial Crisis of 2008-9 (with JD Farmer, Oxford, and J Geanakoplos, Yale), the dynamics of business firms (with O Guerrero, Turing Institute), and natural resource exploitation, e.g., fisheries (with UC Santa Barbara, Oxford, and the Ocean Conservancy). The research on companies is described at length in a forthcoming book, ‘Dynamics of Firms from the Bottom Up: Data, Theories, and Models’, due out next year, which uses U.S. micro-data on firm sizes, ages, growth rates, networks, and locations to create a model at 1:1 scale with the American economy. Prof. Doyne Farmer is an American complex systems scientist and entrepreneur with interests in chaos theory, complexity and econophysics. He has published papers in Science and Nature as well as leading economics journals like the Journal of Economic Behavior & Organization. He is Baillie Gifford Professor of Complex Systems Science at the Smith School of Enterprise and the Environment, Oxford University, where he is also director of the Complexity Economics programme at the Institute for New Economic Thinking at the Oxford Martin School. Additionally, he is an external professor at the Santa Fe Institute. His current research is on complexity economics, focusing on systemic risk in financial markets and technological progress. He has recently published a book entitled ‘Making Sense of Chaos: A Better Economics for a Better World.’ CSSSA Secretary is inviting you to a scheduled Zoom meeting. Topic: CSSSA April Webinar Time: Apr 2, 2025 10:00 AM Eastern Time (US and Canada) Join Zoom Meeting https://us02web.zoom.us/j/82181451627?pwd=uYQJrmdphT9pefWvGKbhQgxQby3beG.1 <https://us02web.zoom.us/j/82181451627?pwd=uYQJrmdphT9pefWvGKbhQgxQby3beG.1> Meeting ID: 821 8145 1627 Passcode: csssa2025 --- One tap mobile +13126266799,,82181451627#,,,,*806775015# US (Chicago) +16469313860,,82181451627#,,,,*806775015# US --- Dial by your location • +1 312 626 6799 US (Chicago) • +1 646 931 3860 US • +1 929 436 2866 US (New York) • +1 301 715 8592 US (Washington DC) • +1 305 224 1968 US • +1 309 205 3325 US • +1 253 205 0468 US • +1 253 215 8782 US (Tacoma) • +1 346 248 7799 US (Houston) • +1 360 209 5623 US • +1 386 347 5053 US • +1 507 473 4847 US • +1 564 217 2000 US • +1 669 444 9171 US • +1 669 900 6833 US (San Jose) • +1 689 278 1000 US • +1 719 359 4580 US Meeting ID: 821 8145 1627 Passcode: 806775015 Find your local number: https://us02web.zoom.us/u/kZvnFw7jt <https://us02web.zoom.us/u/kZvnFw7jt> Copyright © 2025 Computational Social Science Society of the Americas, All rights reserved. You are receiving this email because you have expressed interest in CSSSA in the past. Our mailing address is: Computational Social Science Society of the Americas 20 Loma Blanca Road Santa Fe, NM 87506 Add us to your address book Want to change how you receive these emails? You can update your preferences or unsubscribe from this list. ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ Error! Filename not specified. .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam <https://bit.ly/virtualfriam> to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com <http://redfish.com/mailman/listinfo/friam_redfish.com> FRIAM-COMIC http://friam-comic.blogspot.com/ <http://friam-comic.blogspot.com/> archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ <https://redfish.com/pipermail/friam_redfish.com/> 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ <http://friam.383.s1.nabble.com/> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam <https://bit.ly/virtualfriam> to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com <http://redfish.com/mailman/listinfo/friam_redfish.com> FRIAM-COMIC http://friam-comic.blogspot.com/ <http://friam-comic.blogspot.com/> archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ <https://redfish.com/pipermail/friam_redfish.com/> 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ <http://friam.383.s1.nabble.com/> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam <https://bit.ly/virtualfriam> to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com <http://redfish.com/mailman/listinfo/friam_redfish.com> FRIAM-COMIC http://friam-comic.blogspot.com/ <http://friam-comic.blogspot.com/> archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ <https://redfish.com/pipermail/friam_redfish.com/> 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ <http://friam.383.s1.nabble.com/> <OpenPGP_0xD5BAF94F88AFFA63.asc>.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fbit.ly%2fvirtualfriam&c=E,1,hL3DVqwtOLF2xKtSb8PUsDSK7tuzkLmK4Z19A9nmM1ZGq4IcYJ6K2A2RNr--RKmqZb7Pl1pY_AzYTsFZD9d4lE6RED71aLqjJe1QuCZkJAx11IGEPr1l-A,,&typo=1 <https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fbit.ly%2fvirtualfriam&c=E,1,hL3DVqwtOLF2xKtSb8PUsDSK7tuzkLmK4Z19A9nmM1ZGq4IcYJ6K2A2RNr--RKmqZb7Pl1pY_AzYTsFZD9d4lE6RED71aLqjJe1QuCZkJAx11IGEPr1l-A,,&typo=1> to (un)subscribe https://linkprotect.cudasvc.com/url?a=http%3a%2f%2fredfish.com%2fmailman%2flistinfo%2ffriam_redfish.com&c=E,1,14r5-jVBzgVbD65YsnEokQOqMggSnLkXbGoiq4z1zEJwDPQIjID5vUaWrd6CNhgDrDdmX7AtV3muVip3AYAAB2GjDGjxDvfgYO9LmBsPPQKBP1YsR_E,&typo=1 <https://linkprotect.cudasvc.com/url?a=http%3a%2f%2fredfish.com%2fmailman%2flistinfo%2ffriam_redfish.com&c=E,1,14r5-jVBzgVbD65YsnEokQOqMggSnLkXbGoiq4z1zEJwDPQIjID5vUaWrd6CNhgDrDdmX7AtV3muVip3AYAAB2GjDGjxDvfgYO9LmBsPPQKBP1YsR_E,&typo=1> FRIAM-COMIC https://linkprotect.cudasvc.com/url?a=http%3a%2f%2ffriam-comic.blogspot.com%2f&c=E,1,VwIL28qKPCocADmO6x7d3zLKkjYltd0AzYW-G4ZhZmE3kBioeIuhjmGyGDK9ydbMSIxdzFWZu2F_CcaJEXF8zFiOk7GglM1PpkgsOJcEEafZqHAJ&typo=1 <https://linkprotect.cudasvc.com/url?a=http%3a%2f%2ffriam-comic.blogspot.com%2f&c=E,1,VwIL28qKPCocADmO6x7d3zLKkjYltd0AzYW-G4ZhZmE3kBioeIuhjmGyGDK9ydbMSIxdzFWZu2F_CcaJEXF8zFiOk7GglM1PpkgsOJcEEafZqHAJ&typo=1> archives: 5/2017 thru present https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fredfish.com%2fpipermail%2ffriam_redfish.com%2f&c=E,1,XtQqMS5jdoOFP4LA3FykbWErRYHPPKSaJZY084t5wn3jzwk7MeCN-9LVw9eX8Rlog6tepRVbfOFzoFUFa6LsRbC96pIQ09AsxfG-m8JpgHZ6i3NE&typo=1 <https://linkprotect.cudasvc.com/url?a=https%3a%2f%2fredfish.com%2fpipermail%2ffriam_redfish.com%2f&c=E,1,XtQqMS5jdoOFP4LA3FykbWErRYHPPKSaJZY084t5wn3jzwk7MeCN-9LVw9eX8Rlog6tepRVbfOFzoFUFa6LsRbC96pIQ09AsxfG-m8JpgHZ6i3NE&typo=1> 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ <http://friam.383.s1.nabble.com/> .- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam <https://bit.ly/virtualfriam> to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com <http://redfish.com/mailman/listinfo/friam_redfish.com> FRIAM-COMIC http://friam-comic.blogspot.com/ <http://friam-comic.blogspot.com/> archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ <https://redfish.com/pipermail/friam_redfish.com/> 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/ <http://friam.383.s1.nabble.com/>
smime.p7s
Description: S/MIME cryptographic signature
.- .-.. .-.. / ..-. --- --- - . .-. ... / .- .-. . / .-- .-. --- -. --. / ... --- -- . / .- .-. . / ..- ... . ..-. ..- .-.. FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
