This occurs all the time when using LLMs for code due to the variety of versions of each lib in their training data (which is typically years old already)
Some sort of automatic functionality to find deltas in libraries (even just crude function inspection between versions) and detect/remap them (or roll back versions) might solve that and issues like this.
I see your point but at the same time I'm looking for alternatives and guidance isn't really alive and langchain is just... a lot of stuff(arguably bloat..) and I don't see any obvious easy value from it like I see in lmql/guidance.
LMQL seems to be alive and takes some of these concepts even further. It's the project of 1 or 2 PhD students at ETH Zürich so I'm hopeful they'll see it through.
I thought guidance was smart, but LMQL seems brilliant as it merges pythonic constructions with LLMs (I think it may be an outright superset of python with LLM functionalities?)
IRC had a higher barrier to entry and a lot more common expectation of kickbans for various reasons (justified or not).
Plus the concept of +v doesn't really exist on discords. Sometimes you'd be stuck lurking while only a select few could speak.
And frankly being able to embed images/videos degrades the seriousness in various ways as it can become a bit of a meme/comedy competition (and they scroll disproportionately more text off.
And avatars mean less space is actually text..
<[@+]?shortnick>: <text>
is pretty much the optimal format for information density, which is ironic that twitch chat uses it (to mostly spam emotes)
Is it possible to somehow bridge discord and, say issues on GitHub or another forum so that people can use discord but the information is just pulled from other sources and they're redirected there?
Even worse, OpenAI now gives you a moderation warning if your custom prompt tells GPT not to moralize (thus saving you time and them compute). Go figure
I can't figure out if this was legit or American businesses trying to get the price of GPUs down.
China probably doesn't even need them.
The reason we need so many in the West is because of the nature of capitalistic competition. Datacenters are full of GPUs next to each other (or even the same one just time divided) serving companies who are trying to beat the other out and gain customers
If one put even a fraction of all the compute used to make the current landscape into one foundational model (which they don't have to bother to build safety for nor pay alignment tax on) it would probably make gpt4 look like gpt2.
Not to mention they don't have to give the average person access to anything, they can invoke their MaoAI only for governmental purposes and just the rest a simple 6B model or something.