In this article, we explore the new data quality challenges posed by generative AI, and offer some thoughts on how companies can optimize their data for LLM processing.
From the euphoria created by ChatGPT to its adoption (or rather, its transcription) in companies, the initial acceleration has slowed down, but only to better embrace generative AI. Since the launch of OpenAI's solution, managers' expectations of generative Artificial Intelligence (GenAI) have continued to grow, fuelling a breeding ground of new offers and new players. Anxious to stay ahead of the game (or to keep up), companies are seizing on text-based models such as Large Language Models (LLMs) to speed up their internal processes or try to offer innovative services. The question of corporate strategy with regard to generative AI is becoming more pressing than ever.
Problem: most companies forget to focus their attention on the major ingredient of a successful GenAI strategy... data readiness. In a survey of IT and business decision-makers1 we conducted in September 2023, 74% said they didn't know how to integrate data from a generative AI perspective, nor how to manage AI at the heart of their legacy systems and multi-cloud environments.
In this article, Eastech Group teams will attempt to enlighten these managers on how to structure their data assets, by explaining why it is so fundamental to rely on “good” data when it comes to generative AI, and what is really meant by the notion of “good data” in this perspective. The aim is to improve the quality of all the data contributing to GenAI's strategy, and to make the most of AI's transformative capabilities.
Why generative AI reinforces the need for data quality
Experts have been predicting an AI-driven upheaval in the world of work for several years (if not decades) now. Until now, however, the focus has always been on the automation of the most repetitive tasks and on the complementarity between man and machine.
When models such as ChatGPT and DALL-E burst onto the market, it was the creative capacity of these tools that challenged the prophets of tech: faced with the fluidity of certain texts or the evocative power of certain images, some may have believed in the inevitable replacement of man by AI, including for creative tasks.
- Study conducted in the United States and the United Kingdom, published in September 2023.
Artificial intelligence will not replace human intelligence, however creative it may be, they train each other.