The Hartree Centre Trusted Open Models Institute (TOMI) is a national certification capability for artificial intelligence (AI) models and applications. TOMI provides a structured approach through which application developers can deploy their AI application within the Hartree Centre's high performance computing infrastructure to assess its operation. Developed in partnership with experts from Brunel University, GCHQ and STFC Hartree Centre, TOMI takes into account fundamental criteria, such as data security and accountability.
Between the current debate around AI ethics and codification of new rules for data processing linked to General Data Protection Regulation (GDPR), a gap exists to both standardise and make understandable the use of AI via the development of standard models. A lack of standardisation in CM and AI limits understanding of what these technologies do, how they can be assured and safely integrated into existing business and processes reducing trust in any sector. National standards are needed for AI to flourish and for the UK to embrace AI technology in both a safe and secure way.
How TOMI works
The Hartree Centre has a strong track record working with industry to accelerate the adoption of AI. We aim to provide a safe environment for businesses to embrace AI, taking into account ethics and data security. Our powerful high performance computing infrastructure and expertise is perfectly placed to optimise, test and certify AI models for wider industry. We also offer active support across Technology Readiness Levels (TRLs) 4-7, including training courses and workshops.
TOMI operates as a national facility for the testing and classification of AI models and algorithms. These are assessed using a structured approach taking advantage of our available facilities.
Trusted and Open Models Institute (TOMI) is a key element of the support we
give organisations who wish to implement and adopt assured AI. Building on the
wider STFC experience we have in the curation and classification of data, TOMI
is a focus of our collaborations and MOUs in the development of computational
models and AI with partners such as the Alan Turing Institute.”
- Alison Kennedy, Director, STFC Hartree Centre.
The classification of models is presented on STFC servers as open research which enables other organisations to analyse the model before considering adoption. TOMI also acts as a repository for open models developed in collaboration with TOMI partners and specific to different domains.
Open models can be used by industry to develop AI applications. Taking a 4D modelling approach, specially designed transfer learning algorithms are applied. This enables new open models to be created for specific domains that are built on assured existing open models to ensure models are built on quality and comply with standards and regulatory requirements set by TOMI partners.
TOMI is an open institution co-ordinated by the Hartree Centre. New members of TOMI are welcome to use our facilities to test and publish Open Models for specific domains.
Want to use TOMI to test applications that your organisation is using or get involved and play an active role in the project? Contact us.