AI Hallucination Rates in 2025: Just How Bad IS It?
In some new AI models hallucination rates are improving while in others error rates are inching back up.
Anthropic
Aider: The Terminal Pair Coder I Didn’t Know I Wanted
Aider-chat is a terminal based AI pair programmer that’s easy to install, has some agency–but not too much, and done well–and that can precisely edit your code and also commits changes to Git for you.
Tired of AI Censorship? Open Models for Writing and Smut!
There’s a whole weird world of uncensored text generation out there, and most of it can be found on Hugging Face.
Is the US Monopolizing the AI Industry?
The US funded nearly twice as many AI companies in 2022 as the EU and the UK and over three times as many as China.
Surprise or Not: AI Actually Increases Productivity
The big question: does AI increase productivity? Yes, by as much as 40%, says a recent study.
The Interrogation of Claude 3
Anthropic recently released Claude 3, the latest in the Claude line of “human-centric” chatbots from the company.
Quick Serve 2023-12-21
A study by the Stanford Internet Observatory (SIO) found that the LAION-5B dataset used to train AI image generators contains over 1,000 instances of illegal images of children. The dataset, used to train Stable Diffusion, a popular AI image generator, … Read more
“Universal and Transferable Adversarial Attacks”: Researchers Jailbreak GPT
“Universal and Transferable Adversarial Attacks”: Researchers Jailbreak GPT Researchers with Carnegie Mellon’s Center for AI Safety have published a paper describing a method for developing adversarial attacks (a.k.a. jailbreaks) that was broadly effective across models. The jailbreak in question originates … Read more
What is Context Window Length?
Context windows for LLMs like GPT are the words AROUND the word it is focused on. Context helps the model better understand the target word.
GPT-4 32k & Claude 2 100k: Future Passed
GPT-4-32k and Claude 2 100k were hot stuff a year ago, and now they are footnotes on the road to ever-larger LLM context lengths.