From 2023 to 2025, I ran a small business centered on providing hand illustrated designer stationary in India. This involved designing notebook covers, wedding and other invitations, gift cards, reusable planners, etc.. In the beginning, the work was fun and we started getting a decent number of customers, but it changed pretty quickly when in October 2023, Dall-E 3 got released to the wider public and even Canva started adding AI generated graphics to their library. Everybody could now create their own graphics and generate designs. These outputs may not be perfect and in sync with all the design principles but to the general public that was good enough. And it was a cheaper alternative, so I couldn’t blame them. It was a difficult transition for the business and we found it harder and harder to attract and retain customers. Only the people who already supported small businesses / were conscious about AI remained, who were – though a wonderful support – not a pool we could depend on for profits.
These circumstances made me really think about AI usage and artists’ rights. How easy it is for Big Tech to scrape images and artwork from the internet to power their training data sets. How difficult it is for an artist to conceive, conceptualize, and create that very image. Some of the artists’ whose work has powered Generative AI are from US/UK/EU and they have been fighting back and getting their work removed. But many more of these artists belong to the global south, where there is no proper streamlined avenue available to them to fight back and enforce copyright claims. They made me realize the need for a system to check the development and usage of these big LLMs and our collective responsibility to make sure that the system remains fair and non-extractive.
It also made me very conscious of my own LLM usage. So far, I have used it for deep searching the internet, to brainstorm ideas with clear use cases defined, and to provide grammatical feedback on essays. I have never allowed it to explicitly edit the work or fill out documents. I do find that it erases my voice, though it may sound better and be more “correct” technically, it still not a good substitute. My background is in robotics and computer vision. I skipped the NLP course (some regret, yes) but it made me partial to the simpler forms of machine learning and AI solutions. I was especially fond of the probabilistic solutions like the Markov processes that can run on simple math and provide elegant solutions to complex problems. They were the ones that don’t require more than a normal computer or a single GPU to run. In this new arms’ race, I hope the engineering communities and the world at large can keep grounded and not forget that sometimes the simplest solutions are the best ones. If there is no real need for an LLM, let’s not use it. Especially the ones owned and controlled by Big Tech.