😂
Ok so the “regularly” in my post is doing a bit of lifting. Not too much tho (anchovy is the only other one you could possibly consider frequently used, unless you have a particularly bizarre vocabulary).
My team uses Expensify, and I have to say, I don’t hate it. The website is full featured, the mobile app is actually pretty good, and you can even just email receipts to an email address and it will parse them out properly the vast majority of the time. Management-wise it has all of the approval chains, grouping, etc that you might expect. The company I work for is only about 50 people and my team is only 10 so I couldn’t say how well it scales, but I imagine unless you have some particularly unique requirements it’d do the job.
Believe it or not that initial wave of consolidation brought prices down. A license of SGI Power Animator cost over $30k in the 90s. softImage was not far behind. 3ds Max basically took the fight out of them, at which point Autodesk started going on a buying spree.
It doesn’t come up much except to underscore that in The Culture there is no “normal” with regard to sex or gender because people can be whatever they want and can change on a whim. so without a base gender/sex you can’t have a queer gender/sex.
I believe in the Quantum ClausTM theory - there’s just one guy, and he just makes one present for just one kid (on the nice list, which has at most just one name). But on Christmas Eve he exists in a superposition of states at every child’s house with every possible gift.
Making your own embeddings is for RAG. Most base model providers have standardized on OpenAIs embeddings scheme, but there are many ways. Typically you embed a few tokens worth of data at a time and store that in your vector database. This lets your AI later do some vector math (usually cosine similarity search) to see how similar (related) the embeddings are to each other and to what you asked about. There are fine tuning schemes where you make embeddings before the tuning as well but most people today use whatever fine tuning services their base model provider offers, which usually has some layers of abstraction.
I don’t know about micro but I keep a conventional amount of starter in the fridge and have had it for 5+ years. If I’m out of a bread phase I take it out once every few months, let it come to temperature and feed it. When it gets bubbly and happy again I give it more flour and water till it’s thick and stick it back in the fridge. When my next bread phase kicks in I leave it out of the fridge for a day, feed it again and then use it like normal (once I see it can double in size). Very little waste this way and super-low effort.
I’ve also dehydrated strips of extra-thick starter and have successfully reanimated them years later (just did it recently with 4+ year dehydrated starter in fact).
The easiest option for a layperson is retrieval augmented generation, or RAG. Basically you encode your books and upload them into a special kind of database and then tell a regular base model LLM to check the data when making an answer. I know ChatGPT has a built in UI for this (and maybe anthropic too) but you can also build something out using Langchain or OpenWebUi and the model of your choice.
The next step up from there is fine tuning, where you kinda retrain a base model on your books. This is more complex and time consuming but can give more nuanced answers. It’s often done in combination with RAG for particularly large bodies of information.
I was more just speaking to how it’s many orders of magnitude weaker than the other 3 forces. Though it does work on an infinite scale, so maybe it ought to be a tiny but unbelievably long vaguely dragon-shaped worm thing.
😂 Ok so the “regularly” in my post is doing a bit of lifting. Not too much tho (anchovy is the only other one you could possibly consider frequently used, unless you have a particularly bizarre vocabulary).