Posted in Security Awareness
I mostly stopped reading AI books; there's only so much "AI Doom" you can read. Co-Existence by Ethan…
Jason M. — Strategic Information Security Leader Championing Quantified Risk Analysis and Security Process Improvement.
I mostly stopped reading AI books; there's only so much "AI Doom" you can read. Co-Existence by Ethan Mollick is the exception, because it's about working with the AI we have today.
I think the most important idea is the jagged frontier. AI is very good at some tasks and poor at others, and the line keeps moving (Sutton's Bitter Lesson is a good reason to bet on it moving). If you tried it once, watched it fail, and wrote it off, you're working from stale data. And if you watched it succeed once and now trust it everywhere, you're trusting it well past what you've tested.
He names four human advantages: wide knowledge, deep knowledge, taste and agency. Taste and deep knowledge are what I judge its output with. Deep knowledge is what lets you know when the AI is confidently presenting something incorrect or slightly inaccurate.
So pick a piece of toil you know cold and work it with AI in each of his three modes: as a centaur (you set the goal and the method, and hand it the pieces you choose), as a cyborg (one tight back-and-forth loop), and as a conjurer (spawn agents, assign the work, evaluate what comes back). Check every output against what you know. That's how you find where the frontier sits for your own work.
Full review: https://www.goodreads.com/review/show/9045273195
#AI #InfoSec #BookReview

4 reactions · 0 comments
Topics: #ai, #infosec, #bookreview