Using LLM embedding cosine similarity to classify products into larger categories (bricks) wasn’t something I would’ve thought of. Last I heard of word vectors was back when word2vec came out, I guess in the back of my mind I knew LLMs have something similar and it makes sense that open weight models reveal that information easily.
I somehow doubt they're going to destroy Amazon anytime soon. The market they're in reminds me more of Blue Apron/Hellofresh than Uber/Lyft. May as well take advantage their funders' bad bet before the river.
Have you looked up small, light items on Amazon and seen that they pack them up packs of large N? They do that to hide the fact that delivery is actually a very large fraction of cost on such items. Food delivery, as mentioned in the OP, is another example.
I wonder whether they've considered the possibility of open-sourcing their apps and backend. Perhaps this is just Pollyannaism, but I tend to believe that workers' rights and open-source are the perfect match in a market like ridesharing. "Just goodenough" open source solutions have difficulty breaking through in consumer software (e.g. the mythical "year of the linux desktop"), but in a market like ridesharing, worker coordination could easily provide the push to critical mass.
Is that a reasonable assumption to make? I don't think we have any evidence of other places we look being "largely the same" with respect to life conduciveness. It took billions of years for our own planet to achieve multicellular life, which is a significant fraction of our best estimate of the lifetime of the entire universe up to this point.
The first link in the article [1], to a blogpost about the etymology of the definition of "shard" that we now think of as the primary meaning of the word, is super interesting.
The release of Ultima Online doesn't really seem too close to any inflection point of the word on Google Ngram, but I'm not sure exactly how close we should expect it to be to 1980 or 2000.
When I think of an investor, I think of someone who picks and choses specific companies or sectors of the market to invest in. I'm curious what the Fed's tools allow it to invest in specific companies or sectors other than the economy as a whole. Can they get more targeted than an individual bank? Or is that the limit of their targeting power?
I have a sudden urge to want to deep fake Steven Hawking's voice. I wonder both how the deep fake tech would handle Stephen Hawking as an input, and I'm just realizing that he just missed the survival threshold for being able to use this tech.