Quick Introduction We have all been on forums, chats, reddit, discord, youtube, or somewhere and heard “Oh! Model XYZ is AMAZEBALLZ!zomgwtfbbq” then downloaded it (or more likely, some quantized form of it) and said “eww… This sucks!” This post is going to be a rather technical series of experiments to demonstrate the impact of implementation-specific hazards with inference. I will be using the term “reference implementation” to describe the lab that published and offers first-party hosting of ...
Uh huh… That makes you specifically qualified on the topic? Far more than myself, certainly, with my MSc in Machine Learning.
As far as understanding the technology goes? It makes me more qualified than 99% of the AI glazers I encounter, but you may be an exception.
To say that an LLM is “less dumb” is a misnomer tho, because it implies that the LLM is “thinking” rather than “calculating”.
My issue here is the anthropomorphic language being used to intentionally mislead people who don’t have even a tertiary understanding of the underlying mechanics.
Surely you know that under the hood it’s all just algorithms. 1s and 0s run through a fluctuating probability equation.