Chris Albon Questions AI Cybersecurity Strengths
Chris Albon questions whether AI success in cyber stems from intelligence or exhaustive vulnerability searches.
Chris Albon posted a question about recent AI models' strong performance on cybersecurity tasks. He asked if the results reflect high intelligence or the capacity to test every possible system crack until one succeeds. Bojan Tunguz replied in agreement with the second explanation. Hamel Husain added that the models display jagged abilities across domains yet show a sharp spike in coding. The posts come from machine learning practitioners who work on data, evaluations, and model training.
Combined views
6.5K
5 posts, first seen 4d ago
Chris Albon Questions AI Cybersecurity Strengths
Chris Albon questions whether AI success in cyber stems from intelligence or exhaustive vulnerability searches.
Chris Albon posted a question about recent AI models' strong performance on cybersecurity tasks. He asked if the results reflect high intelligence or the capacity to test every possible system crack until one succeeds. Bojan Tunguz replied in agreement with the second explanation. Hamel Husain added that the models display jagged abilities across domains yet show a sharp spike in coding. The posts come from machine learning practitioners who work on data, evaluations, and model training.
Reactions from ranked influencers
5 posts
Are recent models so good at cyber because they are 200IQ or because they can explore every possible crack in a system until they find one.
This was a question
@chrisalbon Also, super jagged abilities and spike at coding
@BlancheMinerva This makes sense. Sometimes I think we confuse intelligence (the model is really smart) with scale (spin up 100000000 AIs and have them work 24/7 for months)