Isabelle Levent

New art-making technologies change art in consistent ways, and studying the past helps us understand how things will change in the future.

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

Many people don't consider that when they use the internet, be that making a simple HTML/CSS site, or using a site through a big conglomerate, scrapers are scraping and crawlers are crawling the content unless you've specifically configured  robots.txt  and  no-index  rules to prevent it.

The Lab’s primary focus is on the ways in which artists and designers are adopting, adapting and remaking AI processes, building their own datasets and reaching into the ‘grey box’ of AI technologies. 

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

There are a couple reasons why Wordcraft may have struggled with style and voice... Another reason could have been limitations of the underlying model. LaMDA and other similar language models are trained to be most confident on the kind of text they see most often–typically internet data. However, professional creative writers are usually writing

Imagine an art-lover at an exhibition entitled ‘Dots 2008’. He speaks to two artists, each displaying a painting. In both cases, the art-lover cannot see past the seemingly random arrangement of dots of paint. He mentions this to the first artist, who says: “Oh, no, they’re not randomly placed. Each dot represents a friend of mine. The colour of