Chapter 1: The Goldilocks problem
Chapter 2: Future trends
What is a large language model?
Box 1: Key terms
Figure 1: Sample of LLM capabilities and example products
Figure 2: Building, releasing and using a large language model
Figure 3: The scale of open and closed model release
Figure 4: Level of vertical integration in model development and deployment
Chapter 3: Open or closed?
Open and closed models
Figure 5: The structure of UK’s AI ecosystem
Conflicts of interest
Chapter 4: A pro-innovation strategy?
Figure 6: The impact of technology on job creation
Figure 7: Labour exposure to automation by field
Government strategy and evolving priorities
Removing barriers to UK advantage
Figure 8: Affiliation of research teams building notable AI systems
The case for sovereign capabilities
Chapter 5: Risk
What are we talking about?
Table 1: Risk categories
Near-term security risks
Bias and discrimination
Chapter 6: International context and lessons
Lessons for regulation
Chapter 7: Making the White Paper work
Where are the central functions?
Table 2: Indicative staffing overview
Do the regulators have what it takes?
High-risk high-impact testing
Accredited standards and auditing practices
Figure 9: Key actors in the AI assurance ecosystem
Chapter 8: Copyright
Background on data mining
Using rightsholder data
Reviewing the Government’s position
Better licensing options
New powers to assert rights
Summary of conclusions and recommendations
Appendix 1: List of Members and declarations of interest
Appendix 2: List of witnesses
Appendix 3: Call for evidence
Appendix 4: Visits
Evidence is published online at https://committees.parliament.uk/work/7827/large-language-models/publications/ and available for inspection at the Parliamentary Archives (020 7219 3074).
Q in footnotes refers to a question in oral evidence.
© Parliamentary copyright 2024