Isabelle Levent

We find that models learn just as fast with many prompts that are intentionally irrelevant or even pathologically misleading as they do with instructively “good” prompts. Further, such patterns hold even for models as large as 175 billion parameters (Brown et al., 2020) as well as the recently proposed instruction-tuned models which are trained on

Although the humans involved in the creation of Edmond de Belamy were essentially cut out of the art’s creation narrative, the AI itself was often spoken about as having human-like characteristics.

My lesson from these two examples is that it might be possible to make prompting “invisible” by making it part of the UI, and finetuning output for as much of the writer’s context as possible to make it more useful. Latency matters, and cost matters, which are wonderful because these tend to be “regular engineering” type problems rather than AI

If I have an idea for a novel, but the computer writes most of it, is it still my story? This question will not be answered with numbers about how many words I or the computer wrote. It’s going to be answered culturally; it’s going to be a feeling we have about where authenticity or truth really lies.

With so much focus on creation, few systems consider revision. Revision—this is where the average writer gets the most outside help.

Whatever the size of the space, someone who comes up with a new idea within that thinking-style is being creative in the second, exploratory, sense. If the new idea is surprising not just in itself but as an example of an unexpected general  type , so much the better.

A couple participants found success using the chatbot as a convenient search engine alternative (KL, WT). KL wrote: “It’s kind of great to use the chat interface and treat LaMDA as a thesaurus, quote finder, and general research assistant.”

Indeed, AI is a diffuse term that corresponds to a web of human actors and computational processes interacting in complex ways

Now none of this is meant to say that I think programmers, artists and engineers have no responsibilities when it comes to the outputs of machine learning models. In fact, I think we bear responsibility for everything these models do. (I never, for example, attribute authorship to a program or a model. If I publish the results of a text generator,