One of the biggest questions right now is, does using copyrighted work to train machine learning models constitute fair use? I think, by the definition set under common law, it does. But it shouldn’t. Let me explain.

We conduct the first large-scale user study examining how users interact with an AI Code assistant to solve a variety of security related tasks across different programming languages. Overall, we find that participants who had access to an AI assistant based on OpenAI's codex-davinci-002 model wrote significantly less secure code than those without access. Additionally, participants with access to an AI assistant were more likely to believe they wrote secure code than those without access to the AI assistant. Furthermore, we find that participants who trusted the AI less and engaged more with the language and format of their prompts (e.g. re-phrasing, adjusting temperature) provided code with fewer security vulnerabilities. Finally, in order to better inform the design of future AI-based Code assistants, we provide an in-depth analysis of participants' language and interaction behavior, as well as release our user interface as an instrument to conduct similar studies in the future.

Searching the Internet for information sucks. We live in an age of information surplus. At any point in the internet there are an unimaginable number of things to read, watch, listen to, and play through. The average person's backlog of entertainment stretches for several lifetimes.

It is impossible to consume all of the information on the Internet. It was impossible even when the Internet was much smaller. Much, much smaller.