Sruthi Kuriakose
sruthik.bsky.social
Sruthi Kuriakose
@sruthik.bsky.social
Graduate, MSc Machine Learning (UCL)
Interests: NeuroAI, mech interp
Reposted by Sruthi Kuriakose
Per OpenAI the model discovered several 0-days in Artifactory, etc. to get internet and break containers. You can be critical of OpenAI's failure to secure the tests while still being concerned about the offensive capabilities of LLMs. How many highly paid red teams could do what the model did?
August 30, 2026 at 5:17 PM
Reposted by Sruthi Kuriakose
Does anyone know of a good counterpoint (eg, the Hugging Face incident was no big deal, it's better than you think, etc) to this? I am genuinely curious to read it to see what the counter-arguments are.
August 30, 2026 at 4:09 PM
Reposted by Sruthi Kuriakose
The URH: representations are shaped by the slice of reality a system can sense, act on, and cares about: its Umwelt.
Alignment is explained by overlap of ecological constraints, not convergence onto a veridical world model. Universality cannot explain both similarity *and* systematic difference.
6/n
August 28, 2026 at 2:59 PM