Also the questions must be answered a hundred times by strangers online before.
“But it answers the questions correctly?”
“If the Reddit thread we scraped is accurate and the ‘AI’ doesn’t remix the results, yes.”
This one again?
Still more original than a LLM
Inference is pretty cheap and it makes economically sense in terms of efficient and demand to bundle inference into data centers. Sadly that drives demand for hardware to a point that makes it temporarily unaffordable for inefficient uses like gaming.
But maybe that leads to less demanding computer games? The baby giraffes probably refer to training and although that is a big cost, afterwards inference is pretty affordable and the models scale well.
In terms of hallucinations, that problem has gone down significantly through improved training data and tool use - though its not perfect and probably won’t be anytime soon IMHO.
Honest take, let the hot garbage flow that you claim LLMs produce. ;p
Imagine if we just stored the answers in a database instead of writing a program that makes shit up.
That sounds awesome; like a brand new Library of Alexandria, or WikiHow.
LLMs are literally stored solutions in a relational database. The relation and storage are the weights on each simulated neuron.
They literally aren’t, and you’re literally spouting nonsense.
But they aren’t, they also store all the incorrect solutions. Then, on top of that, they inject randomness into responses to make it non-deterministic, which can cause it to just generate an incorrect solution for no reason. That’s why LLMs can’t do arithmetic and they need to use hybrid architecture that can detect math problems on the front end and steer them away from the large language database.
Yeah fuck them baby giraffes. We only had to feed it eight the other day. See it’s going down.
Sounds like a clanker.
