Archive for LLM
Bayesian persuasive privacy at ICML²⁶
Posted in Books, Mountains, pictures, Statistics, Travel, University life with tags #ERCSyG, adversarial privacy, Bayesian inference, Bayesian privacy, ICML 2026, International Conference on Machine Learning, LLM, LLM reviewing, Ocean, Oceanerc, peer review process, persuasive privacy, proceedings, Seoul, South Korea, statistical machine learning, watermarking on May 14, 2026 by xi'anSEINE AI
Posted in pictures, Statistics, University life with tags #ERCSyG, adversarial learning, agent-based models, bandits, Bayesian learning, Bradley–Terry–Luce model, ChatGPT, hallucinations, Information and Communication Technology, INRIA, John Nash, Jouy-en-Josas, LLM, München, multi-agent decision theory, Nash equilibrium, Paris, Queen Mary University of London, reinforcement learning, Seine, slideshare, workshop, ye Norse farce on March 23, 2026 by xi'an
Other speakers included Simon Lucas from Queen Mary London on Simulation-based AI, which I had trouble distinguishing from building a statistical model by goodness of fit (and using bandits used for update), while focussing on competing on some computer game challenges. And Volker Tresp from LMU München on a tensor brain model that he opposes to a Bayesian brain (with a related paper entitled Bayes or Heisenberg: Who(se) rules? which we discussed in general terms over lunch, namely Bayesian learning vs. quantum updating. And Michal Valko from INRIA Paris (and other companies), who went full blast against the Bradley-Terry model!, with a title of Nash and Nemirovski walk into a bar! With a half-time technique approximating Nash equilibria that reminded me of leapfrog. Much entertaining talk that further provided a game-theoretic transition to mine’s.
As an aside, I played yesterday with ChatGPT composing my talk slides out of our arXiv document and it proved a disaster, with hallucinations of results and concepts not in the paper and a complete mess of handling graphs, first creating generic, fake, unrelated pictures, then inserting actual graphs haphazardly throughout the slides. The sorry result I obviously did not use as the workshop did not seem the ideal place for this sort of prank! The actual version only recycles a few of its summarising slides. (With ye Norse farce proper colour choice!)
Nature tidbits [15 Jan 2026]
Posted in Books, pictures, Travel, University life with tags AI scare, black holes, cover, cybersecurity, encyclopedia, England, French Guiana, James Webb telescope, Jimmy Wales, Kourou, little red dots, LLM, Nature, prehistory, tortilla, wikipedia on March 13, 2026 by xi'anThe cover of the 15 Jan issue of Nature is a blurry reconstitution of snapshots from the James Webb Space Telescope, which contributed to uncover these new astronomical objects, deemed to be young supermassive black holes. A notion I only came across recently, during a pre-defence lunch near the Paris Observatory.
Apart from that astronomical advance, Nature puts its focus on the 25th anniversary of Wikipedia. With a tribune bemoaning the insufficient investment of academics in the platform (with mentions of hypocrisy and betrayal). Which sounds rather unfair, since it requires an additional levy on research time, even though I did contribute to a few entries. And recognise the worth of most scientifics pages, as well as the parasitism by LLMs. And an interview of of Jimmy Wales’ about his Wikipedia memoir, Seven Rules of Trust. Kudos to his vision! He sounds rather optimistic about the chances of Wikipedia surviving the tsunAImi, but only if the users keep resorting (and indeed contributing) to the platform rather than accepting the LLM production at fa(r)ce value!
Other entries on
- LLMs suffering (!) from “anxiety, trauma, shame and post-traumatic stress disorder”, although the arXiv reporting the experiment is criticized by others for anthropomorphising the machines. The danger is more in them inducing real trauma in vulnerable (human) users!
- LLMs exhibiting aggressive behaviour (if trained accordingly)
- the oldest evidence on human controlled fires using pyrite (in SE England), 400,000 years ago
- the rise of academics being harassed (and not only in the US) and six recommendations for protecting our digital security (mentioning organisations such as Scholars at Risk Europe, Expert Voices Together, and Faculty First Responders)
- a “Where I work” picture of a food scientist at the National Autonomous University of Mexico flipping a tortilla, with fermented ingredients to improve health and combat malnutrition in poor communities
ChatGPT’ed Monte Carlo exam
Posted in Books, Kids, R, Statistics, University life with tags AI, ChatGPT, final exam, graduate course, LaTeX, LLM, Monte Carlo Statistical Methods, multilevel Monte Carlo, R, unbiased MCMC, Université Paris Dauphine on January 22, 2026 by xi'anThis semester I was teaching a graduate course on Monte Carlo methods at Paris Dauphine and I decided to experiment how helpful ChatGPT would prove in writing the final exam. Given my earlier poor impressions, I did not have great expectations and ended up definitely impressed! In total it took me about as long as if I had written the exam by myself, since I went through many iterations, but the outcome was well-suited for my students (or at least for what I expected from my students). The starting point was providing ChatGPT with the articles of Giles on multi-level Monte Carlo and of Jacob et al on unbiased MCMC, and the instruction to turn them into a two-hour exam. Iterations were necessary to break the questions into enough items and to reach the level of mathematical formalism I wanted. Plus add extra questions with R coding. And given the booklet format of the exam, I had to work on the LaTeX formatting (if not on the solution sheet, which spotted a missing assumption in one of my questions). Still a positive experiment I am likely to repeat for the (few) remaining exams I will have to produce!
Nature tidbits [06 Nov 2025]
Posted in Books, Kids, pictures, Travel with tags 5 year plan, @ScientistTrump, AI, Amazon, brain model, Brazil, CDC, China, Chinese politics, civet coffee, climate change, coal mining, coffee, coral reefs, cover, DNA helix, EPA, evidence, Florida, Francis Crick, India, James Watson, koti luwak, LLM, Margaret Thatcher, minerals, mining, MRC, NASA, Nature, NIH, Nobel Prize, non-fossil energy, non-fossil fuels, NOOA, pangolin, PhD scholarship, proteins, rare earths, renewable energy, Singapore, Trump administration, US politics on January 3, 2026 by xi'anIn this issue of Nature (I read on my way to Warwick), a pre-COP30 tribune, to be opposed to later issues!, with a positive take on the impact of the Trump Administration ignoring the conference, with the advances made by China and India (with a surprising 50% of “installed electricity generation capacity coming from non-fossil sources”, if more critical on Brazil’s efforts than the subsequent tribune by the Brazilian undersecretary for ecological transformation for environment, plus a tribune on the ambiguous terms used by countries to secure access to “critical” minerals, in tune with the on-going muscle-flexing attitudes of China and the US. Although the comment is more focussing on the universal access to minerals than to the protection of the workers extracting it and to the environmental impact of it. Followed though by another comment on the climate impact(s) on mining as (no longer) extreme weather events hinder mining all over the (mining) world.
A reflection on China’s 5y plan for science and its reaching a $500 billion annual investment in R&D, predicting (with a large confidence margin) that it will become the #1 power in sciences and technology in the coming decade. I am actually surprised that China has not yet achieved this goal for semi-conductors. And a tribune on the mixed signal of Takaichi Sanae becoming Japan’s first female prime minister, for science as a whole and for gender equity. (My take being that her having UK’s first female prime minister, Margaret Thatcher, as a role-model is not particularly promising. Just like the projection of Marine Le Pen on the verge of becoming France’s first female president does not carry any optimistic message!)
A light entry on a chemical analysis of the specifics of koti luwak (or civet coffee) that does not tell much, beside civet digestion adding caprylic and capric acids making beans lower in proteïns and higher in fat. Not yet reaching the goal of “leaving the animals out” from producing this luxury coffee ($75 a cup!).
An article on the shrinking number of US PhD admissions (in some colleges) conflicting with another article in a later issue of a stable influx. And a rather shallow article on the creativity or lack thereof of AI, along with the high sycophancy of LLMs, to be opposed to a thoughtful reflection on how AI is radically changing the PhD experience and focus, if almost shelving statistics as a thing from the past! But insisting on graduates keeping their ability to check for the validity of their (AI’s) statistical conclusions!! And another entry on the systematic dismantling of US federal scientific agencies like EPA, CDC, NASA, NOOA, NIH, &tc., by Trump -2.0, which beyond terminating staff contracts in huge proportions is culling the independence of these agencies. With generational impacts on science, training, and evidence-based policies.
A Where I work column featuring a pangolin treated by a Singapore vet, Charlene Yeong. (Unfortunately said pangolin was euthanised after the surgical intervention.) And a book review on the background and motivations of Francis Crick, just prior to his collaborator James Watson passing away. As noted by the author, Cobb, as MRC staff and later non-resident fellow of the Salk Institute for Biological Studies in San Diego, “Crick never had to teach or grapple with university administration: he applied for a grant only once in his life.” And concludes that he was not a saint or a hero but “an extraordinarily clever man with limits to his interests and perception”.
Xi'an's Og