1、 AI Foundation Models Technical update report 16 April 2024 Crown copyright 2024 You may reuse this information(not including logos)free of charge in any format or medium,under the terms of the Open Government Licence.To view this licence,visit www.nationalarchives.gov.uk/doc/open-government-licence

2、/or write to the Information Policy Team,The National Archives,Kew,London TW9 4DU,or email:psinationalarchives.gsi.gov.uk.2 Contents Page 1.Introduction.4 Three risks to fair,open and effective competition.6 2.Market landscape.8 FM development.9 The number of FMs globally continues to increase.9 Fir

3、ms are developing small but capable FMs.10 Firms are deploying smaller FMs on consumer devices.11 Open-source FMs remain an important force for competition and innovation.13 FMs being released can better process more and different types of data.13 Key inputs for FM development.15 FM developers are a

4、ccessing data through partnerships,licensing deals and synthetic routes.15 FM developers predominantly access compute via the cloud,including through partnerships.17 AI accelerator chip supply remains constrained,but the market is diversifying.17 Demand for and investment in AI expertise remains hig

5、h.19 FM development continues to require large amounts of capital.20 Model release,deployment and integration.20 Developers are releasing FMs on a range of platforms,often hosted by the largest technology firms.20 There has been extensive FM integration in digital markets often involving the largest

6、 technology firms but wider adoption remains at the early stages.23 Fine-tuning services.25 FM developers and deployers are exploring monetisation strategies.26 Value chains.27 Partnerships and investments.30 There are a wide range of partnership structures.32 Regulatory developments.34 Consumer und