There's a lot of branding and community building work that needs to be done here. I feel like a lot of my time is just spent trying to cement the definitions so more people can get drawn in.

I'll build up a FAQ over time, but these are:
* Generally for currently enrolled Ph.D. students,
* Generally in person in Seattle, WA at our new office,
* Generally in summer but can be flexible for best candidates,
* Generally 12 weeks, sometimes extended for best candidates,

Mention me in the application if you're interested in my stuff (RL, post-training, etc.), but there are fantastic mentors across the whole stack.

I just cross post everything here, but realistically engagement and growth are really meager here

I have no idea

These lead into GPT-4o's sycophancy nightmare, closely related and overdoing this. buff.ly/sk1DJzx
Or even Grok wanting to imitate everything Elon believes. buff.ly/fI3Z4vr

My library on character training.
First, on OpenAI's original idea of the "model spec" and why transparency matters for personalization. buff.ly/lFR4OD7
Next on OpenAI's updates to the Model Spec and the emerging ideas defining Character Training as a real thing. buff.ly/BoQbJxk

The paper is here: buff.ly/HH79wAJ
Code: buff.ly/OogSjqH
Artifacts: buff.ly/8xPrdFn
Sharan's thread: buff.ly/Wd88Kst

I'll be sharing more thoughts soon, on this, related work, and where character training is heading.

For people following me for a while, you know this has been on my radar as a massively understudied area, as something that's increasingly impacting the cutting edge of model deployment, not studied in academia, but accessible in cost.

Hoping to seed a small research field here!

Source: www.microsoft.com/en-us/resear...