(thank you @esjewett.com for pointing to this, which apart from my above snark is actually pretty cool)

Hacker news is my fave too. Once in a blue moon nerds in the comments mention ABAP and I feel seen

Let’s say you make a tool to automatically review documents with some review criteria your team decides on. Great! Your job is to think 25% harder about those criteria so that when they are applied at scale, they still bring benefits

First idea for me: get way smarter at entrepreneurial stuff

I know I'm shouting this into a void where essentially no one will hear me. I know this is not well-formed thinking. But sometimes, I guess I have to just say something before I can box it in with rationality and structure.

I think I want that. I fully understand that that particular experience may not be part of how AI plays out into whatever we call AGI and beyond. But I think (and hope) there's a path here that leads to a greater capacity for empathy - among many forms of conscious beings.

Something happens that is more than signals across synapses - but it's implemented in...signals across synapses (yeah I know that's a simplification)

How? Why? Will there come a day that I weep for beauty beside a non-human, non-biological intelligence?

My daughter and I wept at the same moment during a movie recently. We turned to each other after the lights came up and said to one another at the same time, "I'm not crying, you're crying!" And it was one of the most joyous moments of my life.

That's what I want to understand.

I get excited about the big ones, where capabilities emerge that no one predicted. That make you scratch your head and think "...maybe there is actual reasoning going on there?"

I think we can find out so much about ourselves, and ways to expand our ability for empathy.

As a not-particularly-credentialed just-completely-fascinated dude, this resonates inside my skull. I don't really get super excited about the advances that make smaller models punch harder with fewer parameters (though of COURSE those are technical marvels)...

I LOVED this episode. Yejin Choi has a knack for making her deep expertise accessible. I came away with several of my own language model assumptions clarified (and changed!), and little idea nuggets to carry forward into more learning.