I'm uneducated on how distillation works at more than a basic level so forgive me if this is a stupid question.
Isn't "distillation" of another provider's model exactly how these models got training date in the first place: Massive amounts of the written word + Prompt -> Answer. Why wouldn't distillation produce similar "reasoning" in the new model? It's just inputs and outputs.
> To tell the truth: I do not even know how to use the pen. I have never needed to do this in my life.
I truly do not understand how one could have T1D for 27 years and still be alive with this attitude. Don't get my wrong, I'm happy she is! And maybe that's just my own limited experience of managing my foster daughter's T1D for the past year talking but... Man, this year has been a rude education in how many ways all of these necessary-for-life pieces of equipment and medications can go wrong and how easy it is to get into a life-threatening situation.
Going more than 30 minutes from home without extras of everything (Dexcom, glucometer, fast-acting, long-acting, glucose tabs, glucagon) is completely unthinkable.
> Find an online community that has self-selected for some kind of criterion
I’ll add onto that: find real-life friends/acquaintances who are both not brain dead and read books. Frequently ask them “what are you reading lately?” Not only does this lead to good conversations and deeper friendships, it results in an endless stream of book leads.
Most of the good books I’ve read for the past several years have been curated for me by two friends who are prolific readers and do all the work for me of finding new books. I occasionally find something they haven’t read, but they certainly do most of the heavy lifting.
There's another Voyager episode in which Tuvok is temporarily blinded and he tells the computer to turn on the tactile interface so he can work on a computer console. So even 25 years ago before everyone had a touchscreen in their pockets, it was obvious that touch screens wouldn't be great for all situations.
Isn't "distillation" of another provider's model exactly how these models got training date in the first place: Massive amounts of the written word + Prompt -> Answer. Why wouldn't distillation produce similar "reasoning" in the new model? It's just inputs and outputs.