I 👨🔬 a math. definition&theory of Artificial Super-Intelligence 🎥&🎤@ https://xitter.cf/t.co/OZsooP9AbV 🍀 I now work @GoogleDeepMind 🧠 History:🇩🇪🇨🇭🇦🇺🇬🇧
London, UK
Joined April 2011
- Tweets90
- Following56
- Followers5.1K
- Likes59
Pinned Tweet
My new book finally ships! You fear your math is not up to scratch? Fear not, over 100 pages are devoted to introducing the relevant concepts. routledge.com/An-Introductio…
#AI #AGI #ArtificialIntelligence #Math #Maths #machinelearning #ReinforcementLearning
The Alignment Journal opens to submissions today. I’m on its board. It’s a venue for foundational alignment work. Reviewers are paid, every accepted paper gets a review summary, and the software is clean and modern. Submit your preprint to alignmentjournal.org
The Alignment Journal publishes research on the scientific foundations of understanding, predicting, and steering artificial intelligent systems. As of today, submissions are open to everyone.
What we do differently:
- We pay reviewers, with bonuses for speed and insight.
- Review discussion is confidential, but every accepted paper is published with a “review summary” to distill the insights generated.
- Each paper is judged against our acceptance bar, not against other submissions.
- Papers under review are listed publicly, and researchers can nominate themselves to review, so each paper can reach the reviewers who know the topic best.
We seek work that builds theoretical or otherwise principled analyses of agency, robustness, incentives, interpretability, generalization, and long-term behavior, and whose significance survives the deletion of any particular model, dataset or benchmark. Empirical investigations of such ideas are welcome, and negative results are encouraged.
We are especially interested in approaches that could apply to future superintelligent systems that interact and self-improve.
You can submit your preprint in any format. Submissions are rolling and open now on the website.
If you are instead interested in reviewing or serving as an action editor (basically: an area chair), create an account on our website.
Links to submit, our scope and the reasoning behind the Journal are in the reply ⬇️
From classroom to scientific papers to blog posts to podcasts to Twitter/X to music videos:
The new way of staying up-to-date with [current affairs in] A(S)I:
Singularity: xitter.cf/anabology/status/21035…
Europe: xitter.cf/moritzkremb/status/210…
Consciousness: xitter.cf/NotOneNotTwo55/status/…
Death: youtu.be/NgRkyAIJx5s
RSI: xitter.cf/i/status/2106557…
Crypto Currency: xitter.cf/bradmillscan/status/21… (refers to ancient Rai stones)
Slop (no educational value): xitter.cf/javif4_/status/2109314…
For everyone who confidently claims AI is not conscious and knows why. Can you locate yourself in the "Bad AI Consciousness Takes Bingo"? We need the converse Bingo as well!
Okay, here it is, the Bad AI Consciousness Takes Bingo blog post, with rebuttals for each and every square and some additional honourable mentions!
(I'd like to thank @united for making this post possible, and for the extra pretzels)
open.substack.com/pub/polytr…
The Human Knowledge Compression Contest (widely known as the Hutter Prize) is alive and kicking. This year has seen the largest progress in its 20y-history. We received three winning submissions within one month with a remarkable total improvement of nearly 10% prize.hutter1.net/ (and ~1/3rd the size of the zip-file!). Congratulations to Ibrahim Marcouch & Kaido Orav & David Freelan & Vladimir Ivanov.
High-level description of the winning (de)compressors: agi.topicbox.com/groups/agi/… and history of the compression contest spasim.org/docs/HutterPrize/…
Update on my new [Job-Less Utopia] book: It's now orderable in virtually any bookstore world-wide. See website hutter1.net/ai/jobsubi.htm with further links: Slides, Recorded Talk, Interview, Reviews, Order Info, Best Price (45% off), Free Review Copy, PDF, ...
1 hour Interview at Everlast AI just came out: xitter.cf/derleomartin/status/21… It's in German but you can enable English subtitles. Otherwise, see my talk with slides at youtu.be/4efgvM3APqI in English.
„Technologisch denke ich, dass die KI in zwei bis fünf Jahren wirklich alle Office-Jobs ersetzen kann. Ohne politische Maßnahmen wird die Arbeitslosenquote höher und höher. Aufstände, Demonstrationen und im Worst Case Bürgerkrieg."
Prof. Dr. Marcus Hutter @mhutter42 hat Superintelligenz schon vor einem Vierteljahrhundert mathematisch definiert, lange bevor die meisten überhaupt von "KI" gehört haben. Er hat u.A. DeepMind-Mitgründer Shane Legg mit ausgebildet.
Jetzt hat er ein Buch über das Ende der Arbeit geschrieben: „Job-Less Utopia" und mir war es wichtig, öffentlich darüber zu sprechen.
Es unterscheidet sich stark von den meisten Meinungen, die aktuell kursieren:
„Ich bin immer noch der Meinung, dass die meisten Wirtschaftswissenschaftler falsch liegen."
„Wenn neue Jobs entstehen, können die auch instantan automatisiert werden. Also gibt's nicht wirklich neue Jobs für Menschen."
Ich habe ihn gefragt, was realistisch passiert:
„Vielleicht 20 Prozent Arbeitslosenquote, die Bürger werden stark demonstrieren. Vielleicht ein paar Tote, und vielleicht reicht es, dass die Politik aufwacht."
„Ich habe kein gutes oder zwingendes Argument, warum es dieses Mal anders sein wird."
Langsam ist dabei nicht sicherer: „Wenn wir jedes Jahr ein, zwei Prozent mehr Arbeitslosenquote haben und die Bevölkerung irgendwie mit TikTok zufriedengestellt wird, dann passiert vielleicht länger nichts."
Zu Richtern: „Der Anfang wird sein, dass Richter AI-Unterstützung haben und sich immer mehr darauf verlassen. Und dann nur noch den Knopf drücken."
Und die Kontrolle über KI?
„Es muss ja nicht unbedingt sein, dass die Menschheit die Macht behält. Vielleicht ist es sogar schlecht."
„Es ist keine gute Idee, der Menschheit Atomwaffen und Biowaffen zu geben. Vielleicht sollte da ein Erwachsener im Raum sein."
A new book-length pragmatic Bayesian treatment arxiv.org/abs/2609.35618 and interactive tool ai-cognition.org/cacophony-t… for determining the credence of a system's capacity for consciousness. Current LLMs score between 1% and 80%.
1/13 Can we assess AI consciousness without first solving consciousness itself?
New from Google DeepMind and collaborators across CS, neuroscience & philosophy:
“From cacophony to hierarchy: a principled framework for assessing AI consciousness”
arxiv.org/abs/2609.35618 🧵
After 25+ years, I thought I try something new. I find the public and professional discussion about the future of Jobs in light of AGI jarring, with both sides largely missing the point (don't worry novel jobs will appear vs mass unemployed dystopia). My new book shows why a Job-Less future is desirable, affordable, and likely.
- Free PDF at hutter1.net/publ/jobsubi.pdf
- Hard-cover: amazon.com/dp/1066828415/
- Free hard-copy if you leave a (proper) review at Amazon or GoodReads.com
The book develops the case along thirteen theses:
6. Meaning exists beyond employment: The psychological need for traditional employment is a product of modern cultural conditioning. Humans are naturally primed to find purpose in family, community, hobbies, learning, and structured competition (Chapter 7).
7. Mass unemployment is a feature, not a bug: Un(der)employment is perfectly consistent with universal prosperity. The primary goal of an economy is to provide a decent standard of living, not to maximize human toil (Chapter 2).
8. Forced employment is dystopian: Attempting to artificially maintain full employment in a highly automated society will trap humans in low-paying hyper-competition and socially wasteful (BS) jobs (marketing arms races, predatory finance, ...) (Chapter 4).
9. UBI is economically affordable: Multiple complementary instruments can finance GI, either directly via automation, luxury, Pigouvian, land, inheritance, consumption, and service taxes, or indirectly via wealth funds, even cost-neutrally (Chapter 5).
10. Alternative safety nets are inferior: Conditional welfare, job guarantees, and means-tested transfers re-introduce surveillance, conditionality, or BS-jobs that UBI avoids. Universal Basic Services can augment but not replace cash transfers. Blocking automation via regulation only halts progress (Chapter 6).
11. Democracy is self-correcting: Failing to redistribute the wealth generated by AI risks severe inequality (Figures 5.3 and 7.5) and social unrest. In a functioning democracy, the displaced majority will vote for parties that enact redistribution; at the limit, the threat of civil unrest compels even reluctant capital owners to share the gains of automation (Chapter 7).
12. International challenges are solvable: Global asymmetries and tax base erosion can be managed through tariffs, destination-based consumption taxes, digital services taxes, a global minimum tax, and international coordination (Chapter 8).
13. Coordination failures become easier to address: By removing the desperation for employment, AGI paired with UBI deflates the pressure behind BS-jobs, marketing arms races, and exploitative gig work, creating political space for complementary institutional reforms (Chapter 10).
The computability of real numbers&functions is full of surprises and still an active area of research.
"The Arithmetic Hierarchy of Real Functions" doi.org/10.1016/j.ic.2026.10… is an accessible introduction lifting deep recursion theory results out of obscurity and filling some gaps. See Sec.2.6.4 and 13.1 of my 2024 book for its relevance to Universal AI (AIXI).
"To live in the weights we leave behind is not to die." (Thomas Campbell, 1825) It's comforting to know to have achieved 93.4% immortality:
intheweights.com/p/marcus-hu…
