Stop Making the ‘Data Scarcity’ Excuse For Your Problems

Data is one of the most important aspects in generative AI, including the likes of DALL.E, Midjourney, Stable Diffusion, and others, alongside large language models like GPT, PaLM, and more, that are trained on tens or hundred billions parameters. Most people in AI believe that increasing the size and quality of datasets is the only way forward, and quibble about increasing the data flywheel for training their AI models. But, it’s time that they started looking beyond data scarcity and LLMs to create intelligent AI systems.  Speaking to Analytics India Magazine, Yoshua Bengio, also agreed that “the bigger, the better” logic for AI is good, but not feasible in the long run. He said that by taking the latest architectures and simply scaling computer power, along with the hopes of increasing data, is a brute force technique and not one that tackles all other problems. Quality versus Quantity For quite some time now, there has been a debate about quality versus quantity