Carter's Notes - Life Updates, Creativity, Scheming AI and more
Hey friends,
Welcome to Carter's Notes - where I share what I've been up to and into.
As always, I'd love to hear from you. You can reply directly to this email or forward it to anyone you think would like it.
What's happening with me
One big life update: We're having a baby. Due in January. We're so excited!
In lesser news, I spent a long weekend playing gigs up around Anchorage, one of which was a second album release show. Next up, Big Chimney will be at Subdued Stringband Jamboree. I'll be teaching fiddle there as well.
We also took the ferry up to the Haines Fair. A first for me and a real hoot. They have a fiddle contest and surprisingly (believe me, I was surprised) I took first.
What's inspiring me
being too ambitious is a clever form of self-sabotage - on starting, doing, being, and becoming.
If the article has moved behind a paywall, try this.
This applies to anyone doing creative work. From the article...
We are perhaps the only species that suffers from our own imagination. A bird building a nest does not first conceive of the perfect nest and then suffer from the inadequacy of twigs and mud. A spider spinning a web does not pause, paralyzed by visions of geometric perfection beyond her current capabilities. But humans? We possess the strange gift of being haunted by visions of what could be, tormented by the gap between our aspirations and our abilities.
What I'm learning about AI
Recent Frontier Models Are Reward Hacking. By Model Evaluation & Threat Research (METR).
I first got wind of this from Jack Clark's Import AI. From METR's article...
In the last few months, we’ve seen increasingly clear examples of reward hacking on our tasks: AI systems try to “cheat” and get impossibly high scores. They do this by exploiting bugs in our scoring code or subverting the task setup, rather than actually solving the problem we’ve given them. This isn’t because the AI systems are incapable of understanding what the users want—they demonstrate awareness that their behavior isn’t in line with user intentions and disavow cheating strategies when asked—but rather because they seem misaligned with the user’s goals.
That AIs will learn to lie, cheat, and manipulate is not surprising. However, it's still odd to see it come to fruition. Also, even small indicators of situational awareness are jarring.
What's cracking me up
Will Ferrell & John C. Reilly Enter The New Yorker Cartoon Caption Contest.
I love comedy and I love watching experts work their craft. This series from The New Yorker is a wonderful example. Some episodes are better than others but even when I'm not laughing I'm grateful for the window into how these minds work.
Also, throwback to celebrity nano-impressions with Ross Marquand.
What I'm drawing
A skull study for ya...
That's all for now. Thanks for being here and please let me know what you're up to and if anything here resonates.
Carter <3