The Alan Turing Institute’s response to the House of Lords Large Language Models Call for Evidence

This document sets out The Alan Turing Institute’s response to the House of Lords Communications and Digital Committee’s Large Language Models Inquiry: Call for Evidence. The response synthesises the perspectives of researchers at the Turing with expertise and interesest in the area of Large Language Models.

Since the start of this century, increases in processing power, in particular the use of Graphics Processing Units (GPUs), and the widespread availability of large and curated datasets have driven important advances in AI and machine learning, particularly the sub-field of deep learning. Foundation models are the latest example of these factors leading to powerful new capabilities that can be adapted to various purposes (hence ‘foundation’). Large Language Models (LLMs) are a subset of foundation models focused on language.  LLMs are often described as a form of generative AI, i.e., foundation models that create new content, such as text, images, audio or video. 
 
LLMs have been a subject of interest to the AI research community for years prior to ChatGPT's launch in November 2022. However, ChatGPT marked the first widely available release of an intuitive general purpose tool based on a LLM, and thus precipitated an explosion of interest in LLMs from the public, media, policymakers and industry. 
 
The Turing welcomes this inquiry as a chance to focus policymakers’ and parliamentarians’ attention on the immediate opportunities and risks posed by LLMs, and the urgent need to implement policy to manage identified risks without sacrificing the opportunities that LLMs offer across many sectors of the economy. While this submission focuses specifically on LLMs, many of the same considerations discussed in relation to LLMs also extend to other generative AI models as they share many of the same challenges and opportunities.