Next up: fine-tuning embedding models with Snorkel’s experts. Join us online this Wed, Jan 19 for a live demo and Q&A session : https://
snorkel.ai/webinar/weekly
-demo-with-qa/
…
RESEARCH
-

Snorkel embedding models fine-tuning webinar live demo
By
–
-
AI Research Studies Measure Human Skill Using AI, Not AI Capability
By
–
Every single study that says “AI beats AI plus human at task X” could be rewritten as “Humans using AI well beats humans using AI poorly” Invariably, human researchers set up the AI-only condition—this is humans using AI!
-

SambaNova’s SN40L RDU 5nm Accelerator for Trillion-Parameter AI
By
–
SambaNova accepted at ISSCC! We’re taking the stage tomorrow to share our paper on the SN40L RDU—a 5nm 2.5D accelerator with three memory tiers designed for trillion-parameter AI. #AI #ISSCC2025
-
AGI arrival timeline and predictions discussed
By
–
We’re definitely going to get AGI soon! Hard to explain my full viewpoint on here but will write more about it soon. In the meantime, food for thought:
-
Explicit objective functions limits in AI domains
By
–
Yes for domains and the a single, unchanging, explicit objective function and no epistemic uncertainty. Which is part of every domain of knowledge, but imo will turn out to be a comparatively small part
-
Chess contexts differ fundamentally from other AI domains
By
–
The contexts are extremely different between chess and other domains
-
Comparing Reasoning Models: Call for Transparency Between xAI and OpenAI
By
–
I'm genuinely curious, as I'm currently comparing different reasoning modeling approaches. It would also be nice if we had answers to these questions, making xAI as open as DeepSeek rather than as closed as OpenAI.
-
AI-Generated Content as Junk Food: Cognitive Decline and Human Value
By
–
AI-generated content is like junk food. Consuming this junk food fuels self-referential growth leading to cognitive decline. Humans, with diverse experiences, create unique insights until AI can learn directly from the physical world. Human-generated content will be valued like
-
Foundational approach to book with potential RL bonus material
By
–
Thanks for the comment! I tried to keep the book more "foundational" rather than chasing the latest research (that's what I do in my blog). But yeah, maybe I can add some bonus material to the GitHub repo on implementing RL with verifiable rewards. (I don't want to do just