#Tag · bonfire.cafe

Hacker News

The Analog I – Inducing Recursive Self-Modeling in LLMs [pdf]

https://github.com/philMarcus/Birth-of-a-Mind

#HackerNews #Analog #I #Inducing #Recursive #Self-Modeling #LLMs #AIResearch #MachineLearning

GitHub

GitHub - philMarcus/Birth-of-a-Mind

Contribute to philMarcus/Birth-of-a-Mind development by creating an account on GitHub.

Alan Zimmerman and 3 others boosted

Marcus Schuler

@schuler@mastodon.social · 7 days ago

When Google restricted AlphaFold 3's commercial use, three MIT PhD students rebuilt the protein folding model in four months. Their open-source version, Boltz-1, now has $28M in funding and a Pfizer partnership. The move highlights tensions between proprietary AI research and scientific openness in drug discovery infrastructure. #OpenScience #ProteinFolding #AIResearch https://www.implicator.ai/when-google-locked-the-door-three-mit-students-picked-the-lock/

Implicator.ai

Boltz: Open-Source AI vs AlphaFold 3 | The Bet

When Google locked AlphaFold 3 behind commercial restrictions, three MIT PhD students rebuilt it in four months. Now Boltz has $28M, a Pfizer partnership, and a bet that open-source can capture drug discovery infrastructure.

Marcus Schuler

@schuler@mastodon.social · 7 days ago

Implicator.ai

Boltz: Open-Source AI vs AlphaFold 3 | The Bet

Hacker News

@h4ckernews@mastodon.social · last week

Task-free intelligence testing of LLMs

https://www.marble.onl/posts/tapping/index.html

#HackerNews #TaskFreeIntelligence #Testing #LLMs #AIResearch #MachineLearning #Innovation

Task-free intelligence testing of LLMs (Part 1)

Hacker News

@h4ckernews@mastodon.social · last week

Dynamic Large Concept Models: Latent Reasoning in an Adaptive Semantic Space

https://arxiv.org/abs/2512.24617

#HackerNews #DynamicModels #LatentReasoning #SemanticSpace #AIResearch #MachineLearning

arXiv.org

Dynamic Large Concept Models: Latent Reasoning in an Adaptive Semantic Space

Large Language Models (LLMs) apply uniform computation to all tokens, despite language exhibiting highly non-uniform information density. This token-uniform regime wastes capacity on locally predictable spans while under-allocating computation to semantically critical transitions. We propose $\textbf{Dynamic Large Concept Models (DLCM)}$, a hierarchical language modeling framework that learns semantic boundaries from latent representations and shifts computation from tokens to a compressed concept space where reasoning is more efficient. DLCM discovers variable-length concepts end-to-end without relying on predefined linguistic units. Hierarchical compression fundamentally changes scaling behavior. We introduce the first $\textbf{compression-aware scaling law}$, which disentangles token-level capacity, concept-level reasoning capacity, and compression ratio, enabling principled compute allocation under fixed FLOPs. To stably train this heterogeneous architecture, we further develop a $\textbf{decoupled $μ$P parametrization}$ that supports zero-shot hyperparameter transfer across widths and compression regimes. At a practical setting ($R=4$, corresponding to an average of four tokens per concept), DLCM reallocates roughly one-third of inference compute into a higher-capacity reasoning backbone, achieving a $\textbf{+2.69$\%$ average improvement}$ across 12 zero-shot benchmarks under matched inference FLOPs.

Hacker News

@h4ckernews@mastodon.social · last week

The Jeff Dean Facts

https://github.com/LRitzdorf/TheJeffDeanFacts

#HackerNews #JeffDeanFacts #HackerNews #TechInnovation #AIResearch

GitHub

GitHub - LRitzdorf/TheJeffDeanFacts: A consolidated list of the Jeff Dean Facts!

A consolidated list of the Jeff Dean Facts! Contribute to LRitzdorf/TheJeffDeanFacts development by creating an account on GitHub.

Hacker News

@h4ckernews@mastodon.social · 2 weeks ago

Scientific production in the era of large language models [pdf]

https://gwern.net/doc/science/2025-kusumegi.pdf

#HackerNews #ScientificProduction #LargeLanguageModels #AIResearch #PDFInsights #FutureOfScience

View (PDF)

Hacker News

@h4ckernews@mastodon.social · 2 weeks ago

The Lottery Ticket Hypothesis: finding sparse trainable NNs with 90% less params

https://arxiv.org/abs/1803.03635

#HackerNews #LotteryTicketHypothesis #SparseNeuralNetworks #DeepLearning #AIResearch #ModelOptimization

Michel Jacquet boosted

EPFL

@EPFL@social.epfl.ch · 10 months ago

🤖⚡ Daniel Gatica-Perez, named the best teacher in the digital humanities section for 2024, believes it’s important to stress the human factor in how technology is adopted and used. His research looks at social computing and human-centered artificial intelligence (AI) – both of which are fast-moving fields.

#DigitalHumanities #InnovationInEducation #AIResearch

Read more 👉 : https://go.epfl.ch/uDV-en