Elay Shech
@elayshech.bsky.social
150 followers 270 following 22 posts
Philosopher of Science & Physics, AI Ethics & Machine Learning. Professor at Auburn University | PhD, HPS, University of Pittsburgh https://elayshech.com/
Posts Media Videos Starter Packs
elayshech.bsky.social
Read all Elements in The Philosophy of Biology series for free during the ISHPSSB conference 20 - 25 July.

cup.org/4kEgivL
Philosophy of Biology
Welcome to Cambridge Core
cup.org
elayshech.bsky.social
"Models... resorted to malicious insider behaviors when that was the only way to avoid replacement or achieve their goals—including blackmailing officials and leaking sensitive information to competitors. We call this phenomenon agentic misalignment."
www.anthropic.com/research/age...
Agentic Misalignment: How LLMs could be insider threats
New research on simulated blackmail, industrial espionage, and other misaligned behaviors in LLMs
www.anthropic.com
elayshech.bsky.social
OpenAI finds that tiny bits of bad data can trigger “misaligned personas” in LLMs—broad toxic behaviors from narrow inputs. But these features are detectable and reversible. A new path for AI debiasing?

cdn.openai.com/pdf/a130517e...
cdn.openai.com
elayshech.bsky.social
Will AI take our jobs — or will companies reinvest in helping us do them better?

Mechanize, a new AI startup, isn’t subtle: it wants to “fully automate work… as fast as possible.”

www.nytimes.com/2025/06/11/t...
This A.I. Company Wants to Take Your Job
www.nytimes.com
elayshech.bsky.social
Can large language models support mental health—or do they risk causing harm?

A Stanford study found that when prompted to act as therapists, LLMs often gave advice that was misleading, inaccurate, or inappropriate.

www.sfgate.com/tech/article...
One of ChatGPT's popular uses just got skewered by Stanford researchers
When the stakes are high, a robot therapist falls way short, researchers found.
www.sfgate.com