Van Bettauer
962 posts
We’re very proud to share the @noumenal_labs white paper, “How To Build A Brain”:
noumenal.ai/how-to-build-a…
1/7
English
Van Bettauer retweetledi
Strange New Universes: Proof Assistants and Synthetic Foundations, by Michael Shulman
"[...] allow mathematicians to explore radically new kinds of mathematics using an LLM-powered proof assistant to train their intuitions [...]"
ams.org/journals/bull/…
English
Van Bettauer retweetledi
This has been a pleasure to build thus far! Lots more to do. Shout out to @andrewwhite01 for his excellent PaperQA project as inspiration.
Michael Garfield 🔮@michaelgarfield
My dreams came true last week when I found out about a rad new web page designed by #FutureFossils Podcast listener @vanbettauer that allows you to ask MY ENTIRE PODCAST a question and get a cogent, legible short answer synthesized from over 200 episodes! Each response at AskFutureFossils.com includes numerous citations and embedded streaming clips from the most relevant moments of the show:
English
@michaelgarfield It's fun to play around with. I appreciate how simple it is. I want to redo the experiments they do in the paper with each LLM agent grounded in a particular podcast corpus.
English
👀 I'm *not* simulating my excitement for this paper:
"[A]s artificial agents are bound to participate more and more to the evolution of culture, it is crucial to better understand the dynamics of machine-generated cultural evolution. We here present a framework for simulating cultural evolution in populations of LLMs, allowing the manipulation of variables known to be important in cultural evolution, such as network structure, personality, and the way social information is aggregated and transformed. The software we developed for conducting these simulations is open-source and features an intuitive user-interface, which we hope will help to build bridges between the fields of cultural evolution and generative artificial intelligence."
GIF
Alberto Acerbi@acerbialberto
"Cultural evolution in populations of Large Language Models" arxiv.org/abs/2403.08882
English
English
English
My dreams came true last week when I found out about a rad new web page designed by #FutureFossils Podcast listener @vanbettauer that allows you to ask MY ENTIRE PODCAST a question and get a cogent, legible short answer synthesized from over 200 episodes! Each response at AskFutureFossils.com includes numerous citations and embedded streaming clips from the most relevant moments of the show:
Michael Garfield 🔮@michaelgarfield
🔮 So excited to share this with you! Years ago I had a dream about mapping every idea I've discussed and turning all of my podcast episodes into a cool Q&A interface that would let you consult every conversation I have ever had on record as if it were an oracle. Not merely a chatbot that would make up new answers based vaguely on my inputs, but a next-best-me that would rigorously cite primary sources the same way I do and direct people back to the correct links so they could go deeper into trustworthy reference materials. A synthesist, the way I am...a library angel that would deliver just the right listening and reading for people who approach it with a focused inquiry. I am excited to say that this day has come. I didn't even have to build it myself! Without even asking my listeners, one of them took it upon themselves to model the topics from over 200 episodes of #FutureFossils and let you query the mind map to pull up Wikipedia-style summaries accompanied with extensive excerpts from the show itself to support every claim it makes. This is AI done right. I've asked it a bunch of questions to see how close it gets to the answers I would give, and am very pleased. As soon as we tweak the last few knobs on the website I'll share it with everyone on my email list...sometime in the next few days. You're gonna love it. (And it's only going to get better.)
English
@michaelgarfield @erik_davis @kenadams_io This approach definitely works for Expanding Mind!
Things get weird fast when you simulate two chat bots prompted to only respond with clips from a podcast
English
I fully endorse this! You're one of the first people I thought about talking to when I saw this...
And in fact, I think I recall @vanbettauer telling me he was already playing around with a topic model of Expanding Mind and having *your* show and *my* show "talk" to each other? (!)
GIF
English
@DrTomFroese Feels like this gets close to some of the work of Kauffman and Varela back in the 80s homepages.math.uic.edu/~kauffman/Form…. Kauffman also has a more recent talk on non-duality that uses imaginary numbers as an example homepages.math.uic.edu/~kauffman/Kauf…, although a bit different from your approach!
English
Does this make any sense at all? If there is anything to this way of interpreting the sense of i, maybe it could help to indicate non-duality at play when i appears as irreducible in our descriptions of nature. 🤓 End/
English
@michael_nielsen Given his latest article theatlantic.com/ideas/archive/… this is looking increasingly like a fake.
English
Douglas Hofstadter on GPT4 and AI Risk: lesswrong.com/posts/kAmgdEjq…
He appears to have changed his mind (a lot)
English
@bgavran3 Has there been any interest in your group in the idea that deep learning works via the singularities inherent in parameter space, the approach known as singular learning theory?
jessehoogland.com/article/neural…
English
Bateson saying Group Theory will be key; The Erlangen program is some Group Theory stuff. Rocco Gangle saying Category Theory is an extension of the Erlangen program. Bateson + Category Theory?🤔
I originally read that Bateson passage while in a molecular biology lab meeting.
English
TIL a previously unpublished book from George Spencer-Brown (creator of Laws of Form) was published recently called “Designing with NOR”: amzn.to/3OD9ihP
cc @comandingo @vanbettauer @diagramaniac @miketranchina
English
@mertozdal6 @evantthompson @pelindilaraclk I think this is a good place to start. Slowly build up a network of decentralized contemplative science, and expand into physical spaces as the funding comes in.
English
@evantthompson @pelindilaraclk Have you heard web3 phenomenon called Decentralized Science? It will help scientists to own their work fully by themselves and help them to raise money with different incentive reverting mechanisms.
English
Pretty much the same four books for me! I took a course with @evantthompson at UBC and down the rabbit hole I went...
Ryutaro Uchiyama@RyutaroUchiyama
Which books have had a foundational influence on your thinking? 4 for me are: • F Varela, @evantthompson, E Rosch, The embodied mind • Gregory Bateson, Steps to an ecology of mind • Lev Vygotsky, Mind in society • Hideo Kawamoto, Autopoiesis 2001 (in Japanese)
English
“A category is a network of relationships.” Ok, you have my attention.
youtu.be/cJ46AOEOc14
YouTube
English
New book stack arrived! Taking a closer look at some of the neighbors of embodied cognition.
English
English
The folks at DeepMind have discovered C.S.Peirce and thus semiotics. Now I begin to worry. arxiv.org/abs/2102.03406 @santoroAI @deepmind @AndrewLampinen @dnraposo
English
Van Bettauer retweetledi
(1/2) ZX-calculus is now being used within the fundamentals of Quantum Machine Learning, namely for analysing the barren plateau phenomenon in training quantum neural networks, including identification of cases where they don't occur.
arxiv.org/abs/2102.01828
English