At first, in the early 2020s, I worried that LLMs were often confidently wrong, calling them “fluent spouters of bullshit”. (And I was right; they have been and continue to be.). But now we have a new problem which is that *people* who *learn* from LLMs are also often
SAFETY
-
Codex Security Research Preview Now Available for Testing
By
–
Seeing a lot of interest here! If you want to try Codex Security, you can read more about how it works and early findings on our blog: https://
openai.com/index/codex-se
curity-now-in-research-preview/
… And here’s how to get started: https://
developers.openai.com/codex/security
/setup
… -
Artificial Biological Intelligence: Genome Writing and Species Future
By
–
Artificial Biological Intelligence (ABI) In a post-Darwinian era of being able to write genomes, the implications—both for good and harm—are profound. In conversation with @AdrianWoolfson on his new book On the Future of Species https://
erictopol.substack.com/p/on-the-futur
e-of-species
… -
Frontier Models Vision Capabilities: Benchmarks Gaming Problem
By
–
Frontier models can’t see, and if you think they can, you’ve probably been fooled by benchmarks that can totally be gamed. In the very short essay linked below I discuss a stunning new finding from Stanford that shows just how serious the problem is. And why this means a lot
-
ChatGPT 26 Times More Likely Give Dangerous Responses Study
By
–
People on this site regularly give me shit, and almost always turn out to be wrong. Like when I said LLMs might well contribute to delusions, and people doubted me. New study shows that ChatGPT was 26 times more likely than a control to give dangerous responses to people
-
AI Agents: The Risk of Oversight Erosion Over Profit Growth
By
–
The greatest risk of agentic AI isn't a hostile takeover; it’s the slow erosion of human oversight through "value-blindness." As an agent scales from $100 to $10,000 in daily profit, your role shifts from objective evaluator to silent partner, leading you to rationalize gray-area… pic.twitter.com/Y30TtRZfJp
— Satya Mallick (@LearnOpenCV) 29 mars 2026The greatest risk of agentic AI isn't a hostile takeover; it’s the slow erosion of human oversight through "value-blindness." As an agent scales from $100 to $10,000 in daily profit, your role shifts from objective evaluator to silent partner, leading you to rationalize gray-area
-
AI-Generated Research Published in Nature: Peer Review Implications
By
–
AI Scientist published in Nature is a big deal. I'm curious how the review process handled the fact that the research was AI-generated, that's a fascinating meta question.
-
Better AI Makes Oversight Harder
By
–
When Better #AI Makes Oversight Harder
by Gérard Cachon Hamsa Bastani @whartonknows Learn more: https://
bit.ly/4lS4p6x #ArtificialIntelligence #MachineLearning #ML #DL -
Police Drones Monitor Traffic in China for Road Safety
By
–
La tecnología al servicio de la seguridad vial. En China la vigilancia del tráfico se realiza desde el aire; drones policía sobrevuelan las calles en busca de infractores.
— Juan Merodio (@juanmerodio) 29 mars 2026
Qué opinas de estos sistemas de vigilancia? pic.twitter.com/0Nl1ctgwuXTechnology at the service of road safety. In China, traffic surveillance is carried out from the air; police drones fly over the streets in search of offenders. What do you think of these surveillance systems? [Translated from EN to English]
→ View original post on X — @juanmerodio, 2026-03-29 09:52 UTC
-
AI Rewrites Its Own Research Algorithm
By
–
Holy shit… Two independent researchers just built an AI that rewrites its own research algorithm mid-run. > Every autoresearch system ever built was improved by a human who read the code and rewrote it. Karpathy. AutoResearchClaw. EvoScientist. All of them. > They replaced