all there. And after this post, I'll keep them even longer.
Often people don't realize there is only "downwards-compatibility", meaning later technologies can "tunnel" an "older" signal (such as DisplayPort being able to Tunnel DVI/HDMI) when using passive adapters, while the other way it's not possible.
> I like it when new ideas for the text come to me as I’m writing. The writing process itself — the thinking that goes on in the moment — is important to me, not just the result.
This. Applies to me while writing text, writing code and writing data analyses.
Exactly this. You always have to choose what to leave in, what to take out. "Ways to do it" suggest themselves, and you weigh them, "is it the best fit for this project". A lot of what's left out evaporates into the void, but often enough you stumble onto one of these doesn't-fits that's nonetheless a keeper. You know it'll be useful somewhere else, or as the heart of something new.
Sure, you create the product, but it's these other by-the-way ideas that keep one coming back.
Here's my 2 cents: In life, like in economy, everything's about marginal utility. If you work 5 days a week and you reduce by one, you gain 50% (free/family/hobby) time, while reducing 20% of work time. In the lost 20% work time, would you have produced the corresponding amount of output? Or are you even more productive because you avoid unnecessary meetings etc.?
It's clear that being an involved parents doesn't allow you pulling all nighters all the time or work through weekends and doing additional sideprojects. Though I agree that without full involvement and ambition one will not achieve a top position or become a tennis superstar, I think a good mixture of all aspects (family/work/sports/friendships) makes everything much more worthwhile. And in my opinion, all involved parent parts should have their fair share of those aspects.
- Parent with partner both working part time, living in a country that allows for that, equally sharing child care for two that regularly make them crazy, but it's soo worth it.
This reminded me of Gaussian Processes. You start out with n-dimensional unconstrained (but strongly correlated) gaussians. As soon as constraints (data) are added (mathematically it's called conditioning), the thing goes more and more into shape.
Prompting feels a lot like this conditioning phase to me. You start with an LLM in unconstrained mode, basically just a "soup" of knowledge. If you prompt wisely, you immediately condition the LLM into "your space of (domain) knowledge".
What comes out is an extended version of your existing knowledge.
Prompting is conditioning, that is what it is. The visual of a GP (like the thing you get if you google image search “Gaussian process”) is a great metaphor for what prompting an LLM is doing.
The output of the LLM is the logits which is sampled - plucking out tokens from a distribution. The input of the LLM is data which constrains the logits. That is what it is.
That’s also how you know that AI will never “solve” intelligence (the way the boosters say it will) without some general mechanism for this conditioning process. The ultimate mechanism would be embodiment; the crappy mechanism we have now is something like openCLAW.
