Supervised learning in DNA neural networks Summary This paper demonstrates a DNA-based molecular system that performs supervised learning and subsequent classification entirely in vitro. The authors extend DNA strand-displacement ‘seesaw’…
Training of physical neural networks Summary This article reviews and synthesises recent advances in training “physical neural networks” (PNNs) — systems that use physical media (optical, electronic, mechanical, magnetic, etc.)…
Analog optical computer for AI inference and combinatorial optimization Summary The paper introduces the analog optical computer (AOC), a hybrid opto-electronic platform that runs entirely in the analog domain to…
OpenAI Says Its Business Will Burn $115 Billion Through 2029 Summary OpenAI told investors it expects to burn roughly $115 billion from now through 2029 — about $80 billion more…
Mathematicians Find GPT-5 Makes Critical Errors in Original Proof Generation Summary Researchers at the University of Luxembourg tested GPT-5 on an open mathematical problem: extending a qualitative fourth-moment theorem to…
All IT Work To Involve AI By 2030, Says Gartner Summary Gartner analysts told attendees at their Symposium that by 2030 all work in IT departments will involve some form…
AI chip startup d-Matrix aspires to rack scale with JetStream I/O cards Summary d-Matrix has introduced JetStream, a PCIe 5.0 I/O card (200Gb/s x2 or 400Gb/s single port) built on…
Chip designer SiFive aims to cram more RISC-V cores into AI chips Summary SiFive unveiled its second-generation Intelligence core family at the AI Infra Summit. The update introduces two low-power…
OpenAI Says Its Business Will Burn $115 Billion Through 2029 Summary OpenAI told investors that its business is projected to “burn” about $115 billion in cash from now through 2029…
Mathematicians Find GPT-5 Makes Critical Errors in Original Proof Generation Summary Researchers at the University of Luxembourg tested GPT-5 on an open maths problem: extending a qualitative fourth-moment theorem to…
