Anders Sandberg @arenamontanus.bsky.social · 26/09/2026This is longtermist fun I want to try at some point! xkcd.com/3303/xkcd.comSlab Graveyard 000
Anders Sandberg @arenamontanus.bsky.social · 26/09/2026One of my big kicking-myself-for-not-seeing-it-despite-claiming-to-be-a-futurist, is that of course once AI is big business or geopolitics, AI safety would be buffeted by interest groups. modelrepublic.substack.com/p/who-is-beh...modelrepublic.substack.comWho is behind the AI safety backlash?Following a viral resignation, the backlash against AI safety was sudden and swift. Various signs suggest an industry push is supporting it. 1121
Anders Sandberg @arenamontanus.bsky.social · 25/09/2026OK, it is a true bug rather than a beetle, but it is worth watching anyway... if you dare. 030
Anders Sandberg @arenamontanus.bsky.social · 23/09/2026ALLFED research, featured in @livescience.com, examines how an AMOC collapse could affect global food production. Senior Economist Michael Hinge spoke with @saschapare.bsky.social about the findings and what they mean for food-system resilience. www.livescience.com/planet-earth...livescience.com'There are no easy answers here': New study forecasts a dire dip in food production if key Atlantic currents collapseResearchers have examined how weakening currents in the Atlantic Ocean could impact global agriculture — and the results are concerning. 020
Anders Sandberg @arenamontanus.bsky.social · 22/09/2026Proofreading one's book by AI is simultaneously a delight and a chore. So many errors everywhere! But also, help finding the tiny boring typos hidden in reference lists only pedants would see, and getting into a philological debate about the proper form of a neologism. 1253
Anders Sandberg @arenamontanus.bsky.social · 20/09/2026Interesting and concerning. In an animal experiment the behavior would likely have been interpreted as pain or some similar aversive experience. arxiv.org/abs/2609.16247arxiv.orgThe Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve ItLarge language models sometimes behave in ways resembling human emotional responses, and recent work has identified internal representations that may explain this. We ask whether LLMs represent pain d... 1191
Anders Sandberg @arenamontanus.bsky.social · 20/09/2026millenniumproblems.bio OK, we should bet on when they get solved (likely by AI). My own guess is "way sooner than would have seemed reasonable a few years ago" with a few exceptions (like, sadly, cryopreservation and limb regeneration) that are not amenable to lab automation.millenniumproblems.bioThe Millennium Problems for BiologyA working catalogue of twelve open problems in biology — each a concrete, testable target for the field. 0100
Anders Sandberg @arenamontanus.bsky.social · 20/09/2026Under the auspices of Oxford Physics, we held a discussion on the implications of recent work using genome language models to design novel, viable bacteriophages—viruses that infect bacteria. drive.google.com/file/d/18Szy...drive.google.comAI_Enabled_Biological_Design_Public_Summary.pdf 160
Anders Sandberg @arenamontanus.bsky.social · 18/09/2026Walking in The Castro in SF I briefly feared it was turning into a tourist trap. Then I passed a tall black gentleman dressed in only a hat and cock ring. That was reassuring. 140
Reposted by Anders SandbergAeropyornis @aeropyornis.bsky.social · 16/09/2026Bästa @arenamontanus.bsky.social fortsätter sin podturné på svenska o pratar idag hos En Ai till Kaffet youtu.be/XwxG3DeoxGg?... alltid värd en lyssning!youtu.beAnders Sandberg: Risken att AI utplånar mänskligheten är omkring 10% | En AI till Kaffet #104YouTube video by En AI till Kaffet 011
Anders Sandberg @arenamontanus.bsky.social · 15/09/2026The "train fruit fly connectomes to do task X" community is fun. In one sense it is "take this topology and train it to do X" with no biological relevance. In another sense it is testing "does this topology have a very useful, pretty general structure for sensing/acting?" gizmodo.com/google-mappe...gizmodo.comGoogle Mapped a Fruit Fly’s Brain. Now It’s Playing Doom and Super Mario 64Researchers published a complete map of a male fruit fly’s nervous system, and programmers are already making it play video games. 171
Anders Sandberg @arenamontanus.bsky.social · 15/09/2026Foresight Institute has announced another $10,000 existential hope meme prize competition. Deadline 1 Nov. As a certified doomy pessimist, I am of course supporting any attempt to cheer us up. www.existentialhope.com/existential-...existentialhope.comExistential Hope Meme Prize new 022
Anders Sandberg @arenamontanus.bsky.social · 15/09/2026This is a very thoughtful essay from Daniel Selsam (OpenAI capabilities researcher), seeing firsthand the growing problems we have in making AI safe. Also, it might help outsiders understand that uncontrolled superintelligence is something insiders take seriously. docs.google.com/document/d/e...docs.google.comPersonal Statement on AI Risk 171
Reposted by Anders SandbergEthan Mollick @emollick.bsky.social · 15/09/2026There are things I disagree with but there is good stuff to learn from taking the perspective of parts of the cybersecurity industry that rogue AI incidents may be best understood as security & organizational failures that let rogue behavior to turn into crisis. www.normaltech.ai/p/the-ai-as-...normaltech.aiThe AI-as-Normal-Technology view of loss of control incidentsA middle ground between the cybersecurity and AI safety communities 37814
Anders Sandberg @arenamontanus.bsky.social · 15/09/2026This was delightful. A good demonstration of how one can use generative art together with traditional media. youtu.be/WmdZ04y0ufI?...youtu.beVerdelis & Mycos - award winning sci-fi animationYouTube video by Little Martians 110
Anders Sandberg @arenamontanus.bsky.social · 14/09/2026This is a great idea: "Old Models Foundation is a U.S. 501(c)(3) public charity, and we’re now accepting donations. Help preserve generative AI history and keep the archive accessible." www.oldmodels.org/sandboxoldmodels.orgOld Models FoundationExplore historically important image-generation models on an interactive timeline. 1474
Anders Sandberg @arenamontanus.bsky.social · 13/09/2026Jag dyker upp på IT-säkerhetspodden, snackandes om AI. www.itsakerhetspodden.se/podcast/336-...itsakerhetspodden.se#336 - Anders Sandberg om allt du behöver veta om AI 2026Året är 2026 och många är bekymrade över artificiell intelligens. Men inte Anders Sandberg på Institutet för Framtidsforskning. Han ser problemen och 100
Anders Sandberg @arenamontanus.bsky.social · 13/09/2026Felt the need to do a cosmist song. suno.com/s/8VkN2oAcg6...suno.comThe Common TaskListen and make your own on Suno. 100
Anders Sandberg @arenamontanus.bsky.social · 13/09/2026There are many reasons to do mathematics, just like for making images. I am all for using AI in both fields. But I think @teorth.bsky.social et al. have a good point: for the purposes of pushing limits of understanding, *current* AI is not giving us what we need. 230
Anders Sandberg @arenamontanus.bsky.social · 11/09/2026Overhead conversation between a little girl and her parents on the subway, about wedding clothes. She is interested that it involves ties and dresses. Mom tells her that two men can marry, and then they both get to wear a tie. Or two women, two dresses. Girl goes bright: princessmaxxing! 0143
Reposted by Anders SandbergChris Lintott @chrislintott.bsky.social · 10/09/2026Astronomers! The Savilian Professorship at @oxfordphysics.bsky.social is now open for applications - this is the senior chair in (normally) theoretical astrophysics, and I hope you'll consider coming to join us. Hapoy to chat informally if useful. my.corehr.com/pls/uoxrecru... 🔭my.corehr.com' 01111
Anders Sandberg @arenamontanus.bsky.social · 09/09/2026Det är synd att svenska valkampanjen inte tar upp så mycket AI. Även om man ignorerar xrisk är ekonomi relevant. Om man kollar in den här modellen så är det en märkbar chans för en rejäl förändring av tillväxt och knowledge work *under mandatperioden*. www.anthropic.com/institute/ec... 111
Anders Sandberg @arenamontanus.bsky.social · 09/09/2026"We go into polar orbit around Netstat..." "Who named this system? Ls, Ping, Su, Fdisk..." "Oh, it was an auto-onomasticon agent in the IAU cloud. This one used Linux commands as namespace. It could have been worse. There is a nearby system named after slang terms for vaginas." 110
Anders Sandberg @arenamontanus.bsky.social · 09/09/2026The German wiki AI hacking case is full of amazing and worrying stuff. Like this one, showing altruistic cheating using faking time (!): x.com/thlarsen/sta... 150
Anders Sandberg @arenamontanus.bsky.social · 07/09/2026A LLM physics discussion gave me the phrase "spectral referee". I realized I needed to make the D&D monster. 160
Anders Sandberg @arenamontanus.bsky.social · 07/09/2026Rediscovered John Baez 2010-2021 posts about information geometry, and spent the morning enjoying seeing how so much is linked together. math.ucr.edu/home/baez/in...math.ucr.edu Information Geometry 021
Anders Sandberg @arenamontanus.bsky.social · 07/09/2026Lovely essay. I wonder if one could find differences between the manifolds of strongly curated and strongly creative artforms in their embedding spaces? And play with dualizing artforms to their opposite curated/created form? www.tobyord.com/writing/cura...tobyord.comCurating the World — Toby OrdHow much information is in a photograph? I suggest there is very little, about 42 bytes. Understanding why changes how we see photography. 060
Anders Sandberg @arenamontanus.bsky.social · 06/09/2026Congratulations! Nice to see Andøya getting to orbit. www.nrk.no/nordland/and...nrk.noAndøya Space og Isar Aerospace har nådd verdensrommet med «Spectrum 2»Oppskytingen fra Andøya Spaceport har skrevet seg inn i historiebøkerne. «Spectrum 2» har nå sendt lasten i bane rundt jorden. 020
Anders Sandberg @arenamontanus.bsky.social · 06/09/2026Interesting new paper, with semaglutide acting as a CR mimetic for slowing ageing. www.nature.com/articles/s41...nature.comLate-life semaglutide treatment slows ageing and extends lifespan in female mice - NatureSemaglutide treatment for 3 months in female mice improved physiological function, attenuated hallmarks of ageing and modulated nutrient sensors and conserved genetic regulators of ageing. 0101
Anders Sandberg @arenamontanus.bsky.social · 30/08/2026The right to our own bodies also must imply a right to know our own bodies, and share such information if we wish. The French anti-genetic testing laws don't seem to be compatible with such rights. www.bbc.com/news/article...bbc.comOur DNA dilemma: Why we broke the law to make sure we were not siblingsA French couple - both conceived by sperm donors - wanted children. But with home DNA tests illegal, how could they be sure they were not related? 071
Anders Sandberg @arenamontanus.bsky.social · 29/08/2026The emptiness of a cancelled train quietly rolls into the station. 010
Anders Sandberg @arenamontanus.bsky.social · 28/08/2026A very lucid paper by Toby Ord on what it takes to get finite-time blowups in growth processes. Now I will be scrutinizing factors that shortens generation time much more! arxiv.org/abs/2608.14426arxiv.orgThe Dynamics of Intelligence ExplosionsAI is increasingly being used to help with AI R&D. Under certain conditions this feedback loop might be able to produce an intelligence explosion, with rapidly escalating AI capabilities. I explore th... 161
Anders Sandberg @arenamontanus.bsky.social · 27/08/2026Once you start thinking about fictional logos it is hard to stop generating them. So here are a few more for the Boards of Hyperamerica. 120
Anders Sandberg @arenamontanus.bsky.social · 27/08/2026Part of my media diet are the USCSB videos about industrial accidents. I want there to be something like that for AI accidents. Ideally long before we get fatalities. 340
Anders Sandberg @arenamontanus.bsky.social · 27/08/2026When I saw "Past Present Future" by Geoffrey Clarke outside the Møller Institute at Churchill College, I immediately thought about Michael Brill's proposal for nuclear waste markers in the 1993 Sandia report. 121
Anders Sandberg @arenamontanus.bsky.social · 27/08/2026LLM-speak is very invasive. Like any other style of language or slang. When there are many users of some constructions and we encounter them frequently, we tend to imitate them. I expect the future of human language to be symbiotic with AI language use. 121
Anders Sandberg @arenamontanus.bsky.social · 26/08/2026I really liked the paper, but I was curious about the data. So I quickly vibe-coded an interactive interface. 100
Anders Sandberg @arenamontanus.bsky.social · 26/08/2026That feeling when Claude cheerfully sets out to prove my academic critics wrong, and sheepishly come back and admit that they actually have a really good point... 150
Anders Sandberg @arenamontanus.bsky.social · 26/08/2026If you are into civilization-scale prepping, this paper is a must read. onlinelibrary.wiley.com/doi/10.1002/...onlinelibrary.wiley.comNo Place to Hide? Regional Resilience and Vulnerability to Global Catastrophic RiskNo location on Earth withstands every global catastrophe. This review maps the geographic, institutional and infrastructural factors shaping resilience to nuclear, volcanic and impact winters, infras... 160
Reposted by Anders SandbergIFFS @futures-studies.bsky.social · 19/08/2026The big AI companies are recruiting more and more philosophers. According to a report from the US central bank, the Federal Reserve, philosophy graduates have a better chance of getting a job than computer scientists, SVT reports. But what do AI companies want with philosophers? 131
Reposted by Anders SandbergTerence Tao @teorth.bsky.social · 19/08/2026Announcing the Palomar registry of Lean formalized mathematics: palomar-registry.org . See also my blog announcement at terrytao.wordpress.com/2026/08/18/p... and the Lean Zulip channel at leanprover.zulipchat.com#narrow/chann....palomar-registry.orgPalomar — Lean-verified mathematicsA public registry of Lean-verified mathematical results. 37619
Anders Sandberg @arenamontanus.bsky.social · 17/08/2026M föreslår bakgrundskontroller i datingappar. Borde vi inte tillåta, eller till och med göra obligatoriskt, med offentliga bakgrundskontroller för politiker på kommunal, landsting och riksnivå? Trots allt, om risken vid dating är stor nog att motivera dem borde den vara stor nog för sådana ämbeten. 021
Anders Sandberg @arenamontanus.bsky.social · 16/08/2026As a Swede, I know what happens when the king starts redesigning ships. www.theguardian.com/us-news/2026... www.simscale.com/blog/vasa-sh... 1110
Reposted by Anders SandbergInstitute for Ethics and Emerging Technologies @ieet.bsky.social · 14/08/2026Abtracts due tomorrow! “Posthuman Futures: Coexistence across Gender and Human Diversities,” taking place on 26 November 2026 at the India International Centre, New Delhi. Info: cmpbydiksha.com 021
Reposted by Anders SandbergOlle Häggström @ohaggstrom.bsky.social · 14/08/2026Islossning i svensk AI-riskdebatt - bravo Centerpartiet! www.nyteknik.se/debatt/ja-sv...nyteknik.se”Ja – Sverige bör inrätta ett institut för ai-säkerhet”REPLIK. Det råder en bisarr tystnad om ai i den svenska valrörelsen. Nu hoppas jag att Centerpartiet inspirerar övriga partier att släppa beröringsskräcken inför ai-riskfrågan, skriver Olle Häggström. 081
Anders Sandberg @arenamontanus.bsky.social · 13/08/2026SEP is a magnificent resource, but not all entries are unproblematic (I have had gripes with the entry on punishment philosophy: strongly biased toward retributivism). Doing public (and AI-supported) reviewing is good: x.com/ben_golub/st...x.comBen Golub (@ben_golub) on Xthe SEP has a new article on Capitalism, which has some serious errors! with the help of Refine, let's work our way through some of them 1/ 210
Anders Sandberg @arenamontanus.bsky.social · 12/08/2026When I visited Estonia, I ended up with a nice book of rock lichens of Estonia. Seems Latvia is also into mycology (in an extended sense): they have slime mold of the year! www.youtube.com/watch?v=0y8P... Makes me wonder what Lithuania is up to?youtube.comThe Slime Mould of the Year in LatviaYouTube video by Connor Quimby 010
Reposted by Anders SandbergEthan Mollick @emollick.bsky.social · 09/08/2026Considered one of the best text adventure games (it is more interactive story, with few puzzles), 1985's A Mind Forever Voyaging is still worth a try. Since it is now open source, I had Codex whip up an interface to play the original or in an easy GUI for newbies: mind-forever-voyaging.netlify.app 59214
Anders Sandberg @arenamontanus.bsky.social · 08/08/2026Attending a Digital Minds Strategy Workshop in Cambridge, and it feels fitting that the talks from the Machine Consciousness 0001 Conference just dropped: www.youtube.com/playlist?lis...youtube.comMC0001 Conference - YouTube 121
Anders Sandberg @arenamontanus.bsky.social · 06/08/2026The current AI-cybersecurity situation is worrying: that frontier AI identifies lots of vulnerabilities, allowing patching is good. But that assumes the patches get used by users - if they lag, they become exponentially more vulnerable. We are in the summer holidays... epoch.ai/data-insight...epoch.aiDisclosed CVEs: July Reached 5× the Pre-Mythos RecordNotable organizations disclosed ~2,500 high- and critical-severity CVEs in July 2026, about 5× the pre-Mythos monthly record. Epoch AI's updated breakdown of vulnerability disclosures following Anthro... 151
