Microsoft
Post #7057 · source on Telegram
Windows 11 Greeted As Malware by Stick Figure Characters
Description
A simple comic showing stick figure characters. The Windows 11 logo (blue four-pane window icon) with the text 'Windows 11' appears on the right saying 'hey guys'. One stick figure points at it and shouts 'Malware!' while a speech bubble from the Windows logo protests 'NOT WHAT I'M CALLED'. The stick figures appear alarmed with one pointing accusatively. The joke is that Windows 11 is so bloated with telemetry, forced updates, ads in the Start menu, and aggressive data collection that it's functionally indistinguishable from malware
Use J and K for navigation
Comments
9Comment deleted
The only difference between Windows 11 and malware is that malware doesn't require TPM 2.0 to spy on you - it's more accessible and has better hardware compatibility
The fastest way to spot a hype-driven 'AI researcher' is to ask them to explain temperature without using the words 'randomness' or 'creativity.' Their own output suddenly becomes very deterministic and very silent
Nothing says ‘2025 AI boom’ like someone citing temperature without realizing it’s just a glorified 1⁄kT on the logits - proof that even in machine learning, entropy isn’t the only thing increasing
Nothing like watching someone with 'LLM Researcher' in their bio explain temperature as 'changing token representation' - next they'll tell us gradient descent is just 'making the numbers go down' and attention mechanisms are 'when the model really focuses hard.'
When your 'LLM researcher' thinks temperature affects token embeddings rather than softmax sampling probabilities, you realize the field has more people tuning hyperparameters in notebooks than understanding the Boltzmann distribution underneath. It's like claiming you're a database expert because you know SQL syntax but think indexes physically reorder table rows on disk
My LLM hiring filter: ask them what temperature does to logits; if they can’t get from “logits/T” to softmax, we set T=0 and deterministically don’t sample their résumé
LLM 'researcher' sampling from uniform ignorance distribution - zero temperature won't fix that entropy
Temperature isn’t a creativity slider - it’s 1/T on logits before softmax; if your LLM ‘expert’ can’t name the partition function, set their prod access to T=0
Do you really think they use variable-speed fans in GPU clusters and bother setting them up? Comment deleted