On Generative AI

I’ve been playing around with generative AI models and I’m stunned by the power of this technology to change the way humans partner with computers to create and produce original works. Unlike analytical AI that analyzes vast amounts of existing data, generative AI creates new things. You can give it a prompt, and it will generate a paragraph. You can feed it text, and it will create an image—the image at the top of this Insider is the product of a prompt I gave OpenAI’s DALL-E 2: “a human walking amidst the redwoods, digital art”. Entire jobs, industries, and tasks will become computer-assisted in ways that weren’t possible before. Lawyers will use AI to draft contracts, consultants will use AI to assemble qualitative analyses, and students are already using AI to write papers in school, which they then augment with additional citations and further explanation.

What’s most intriguing to me about generative AI are the questions it raises about what is original. Is it more original for us to paint with oil and canvas than it is to collaborate with a computer to create a painting? And if we acknowledge that the inspiration for our oil-and-canvas painting was derived from the millions of inputs from our experience (the art we’ve seen in museums, the books we’ve read, our understanding of beauty), then do we have any claim of originality of the finished product?