crystal_revenge

35 ms·

451.
by crystal_revenge 2y ago
Very refreshing to see some philosophical discussions around LLMs! As someone interested in both philosophy and LLMs it's fascinating how many meta-physical questions dip their toes into the practical without many people realizing it.
452.
by crystal_revenge 2y ago
I suspect, for a lot of math students, Real Analysis is the first introduction to "real math" and is frequently taught assuming such. I found, at least in undergrad, that a fair bit of students that think they are "good at ma
453.
by crystal_revenge 2y ago
Seriously! Working weekends in retail when I was young, one of the hallmarks of a "real, professional job" was not having to work nights/weekends when your routine schedule is during the day. It was a major motivator to get t
454.
by crystal_revenge 2y ago
These are all fantastic points, and I strongly agree that most people don't appreciate linear models nearly enough. Another one I would add that is very important: Human beings, especially in groups, can only reasonably make linear
455.
by crystal_revenge 2y ago
Of course at the extreme you are over tuning for exploitation but in practice it's never completely random. You always have some information about the probably winner, so long as the P(A>B|obs) is not 0.5 Taking a long time to reach
456.
by crystal_revenge 2y ago
> And pulling granular data into a Python environment and fitting a regression is much less efficient than calculating aggregated statistics like mean and variance. This is not true. You almost never need to perform logistic regression o
457.
by crystal_revenge 2y ago
This is the correct approach, but having done A/B testing for many years (and basically moved away from this area of work), nobody in the industry really cares about understanding the problem they care about prompting themselves as exp
458.
by crystal_revenge 2y ago
Strictly speaking you don't need to wait for some arbitrary significance threshold. I don't know why so many people treat website A/B tests as similar to carefully, traditional nhst controlled experiments. Website A/B te
459.
by crystal_revenge 2y ago
> the only thing you can compete on is how many parameters it takes and how cheaply you can serve that to users. The problem with this strategy is that it's really tough to compete with open models in this space over the long run. I
460.
by crystal_revenge 2y ago
I awhile back I used to do work with a major DARPA contractor. If you're familiar with security clearance for these roles, you know that at the higher levels of clearance you eventually need to take a polygraph exam. I was never intere
461.
by crystal_revenge 2y ago
"AI" has always been a marketing term first, there's a great clip of John McCarthy on twitter/X basically pointing out that he invented the term "Artificial Intelligence" for marketing purposes [0]. Don't
462.
by crystal_revenge 2y ago
People take genetic determinism way too seriously, and that original "Why Cilantro Tastes Like Soap" NYT article has been aggressively misused to justify a generation of picky eaters. It's tragic how many people miss out on
463.
by crystal_revenge 2y ago
> The term agent has been used by AI researchers without a formal definition [1] > [1] In traditional AI, agents are defined entities that perceive and act upon their environment, but that definition is less useful in the LLM era — ev
464.
by crystal_revenge 2y ago
While I absolutely love this illustration (and frankly everything Jay Alammar does), it is worth recognizing there is a distinction between visualizing how a transformer (or any model really works) and what the transformer is doing. My