AI Security Institute – Frontier AI Trends Report

Progress in the cyber domain is significant. Al systems are just beginning to complete expert-level cyber tasks typically requiring 10+ years of experience. Two years ago, they could barely complete tasks requiring one year of cyber expertise.

https://www.aisi.gov.uk/frontier-ai-trends-report

The quote above from this inaugural report by the UK AI Security Institute, demonstrates the rapid rate of progress currently being made by the leading AI system developers. The report outlines the capabilities of the current AI systems that are available and attempts to evaluate what risks they represent. The following excerpts were considered to be of primary interest;

“A human expert baseline was established with PhD holders in relevant biology or chemistry topics. The QA evaluations are designed to be difficult, with absolute scores for PhD holders ranging from approximately 40-50%. Even so, we’ve seen rapid progression in models’ performance up to and beyond this PhD baseline.”
“People are increasingly turning to Al systems for emotional support or social interaction. While many users report positive experiences, recent high-profile cases of harm30 underline the need for research into this area, including the conditions under which harm could occur, and the safeguards that could enable beneficial use.”

Author: BlackHole

Leave a Reply

Your email address will not be published. Required fields are marked *