20 December 2024

Decoding ChatGPT’s Success: Policy Brief Takeaways for Europe’s AI Future

Alexandre Ferreira Gomes & Maaike Okano-Heijmans

Introduction

The release of ChatGPT 3.5 in November 2022 is among the most important milestones in the development of artificial intelligence (AI). Launched by OpenAI, which is based in San Francisco, ChatGPT was the first widely accessible General-Purpose AI chatbot (see Box 1). It was the outcome of seven years of development since OpenAI’s establishment in December 2015. ChatGPT brought GPAI to the general public and transformed the perception of technology experts and users alike about its potential, akin to how the advent of the internet revolutionised communications in the late 1990s.

The new chatbot attracted over one million users within five days of its launch.3 Users quickly tested the tool, thereby showcasing its diverse capabilities, from essay writing, travel planning and code generation, to step-by-step guides on how to create a business. Shortfalls were also quickly identified, including false academic sources or book names, incorrect answers to simple mathematics problems, or how its safety protocols could initially be bypassed relatively easily through clever prompting.

ChatGPT is the final product of a complex GPAI supply chain that involves all layers of the so-called technology stack, from hard infrastructure (including chips and data centres) to the application layer (software and applications). This GPAI supply chain serves not only large language model (LLM) applications such as ChatGPT, but also image generators like DALL-E, and video generators like Synthesia, which have been launched in parallel as groundbreaking products in the broader GPAI revolution.

No comments: