Early Market Engagement Event for the delivery of Artificial Intelligence (AI) Data Labelling Capabilities

Description

The Defence Artificial Intelligence Centre (DAIC) was established by the Integrated Review in 2021 and first began operations in 2022 and is a part of the Ministry of Defence (MOD). Our purpose is to enable and accelerate ambitious, safe, and responsible Artificial Intelligence (AI) adoption across UK Defence for strategic advantage. We work collaboratively with international partners, academia, industry and across government to:  - Understand and cohere AI initiatives across Defence through a complex network of internal stakeholders. - Enable widespread AI adoption through the provision of central products and services.  - Explore, deliver, integrate and scale AI-enabled solutions at every classification.  - Inform and inspire the development of safe and responsible Defence AI capabilities through national and international partnerships.   We are currently undergoing an exercise to determine the delivery model for an AI Data Labelling capability which may or may not result in a decision to outsource the services in-scope. As part of this exercise, MOD is considering the role and responsibilities it wishes to hold in any future delivery model. Whilst some capability is already outsourced, any new contract(s) would essentially represent 'first generation outsourcing' for this capability i.e. the first time MOD has outsourced the Data Labelling capability at scale. In the context of this potential opportunity, Data Labelling is adding labels/tags to data so that machine learning (ML) models can be trained to identify specific content. Data Labelling adds context to raw data to improve the accuracy of the resulting ML models e.g. labelling objects in an image/video or entities in text. Data Labelling helps to reduce ambiguity and ensure ML models are consistent in their outputs and improves the reliability of future labelling activity. Better Data Labelling capability will allow for improved data driven decision making as well as supporting wider AI capability enhancement. Within MOD, it is likely that the Data Labelling capability will be required by both civilian staff and regular forces across different government security classifications (OFFICIAL, SECRET, TOP SECRET). We are looking to hold an in-person early market engagement event for the Data Labelling capability to help us test the current thinking around delivery models and garner ideas/feedback from the market on any potential procurement, transition and delivery. How to Register MOD will be holding an in-person early market engagement event on Friday 14th February 2025 (10:00 - 14:30). During the event suppliers will have the opportunity to meet the programme and senior management team and find out more about the Data Labelling capability. The objectives of the event are as follows: - Support the development and decision making around the future delivery model for the AI Data Labelling capability within MOD, including in-house/outsource boundaries; - Make suppliers aware of the current strategic direction that MOD is pursuing in relation to Data Labelling capabilities, and how it aligns with the current Defence Artificial Intelligence Strategy (https://www.gov.uk/government/publications/defence-artificial-intelligence-strategy/defence-artificial-intelligence-strategy); and - Provide an opportunity for suppliers to provide feedback on delivery models and pose questions to the Authority. A Request for Information (RFI) exercise related to this opportunity is currently live in the market and started in December 2024; further information can be found within the link here (https://www.contractsfinder.service.gov.uk/Notice/d5033951-77d4-454a-8100-8374dc63b5cf) and we encourage you to respond to this before the deadline of Friday 17th January, 23:45. To note, responding to this RFI is not a pre-requisite for attending this early market engagement event. Suppliers interested in attending the Friday 14th February early market engagement event are invited to register their attendance via this link (https://www.techuk.org/what-we-deliver/events/defence-artificial-intelligence-centre-early-market-engagement-artificial-intelligence-data-labelling-capabilities.html) no later than Tuesday 11th February, 17:00. techUK will be assisting the MOD in running this event and the link provides further information regarding location, time, agenda, number of delegates allowed and further contact details regarding the industry day. Additional information: The DAIC AI Playbook (https://www.gov.uk/government/publications/defence-artificial-intelligence-ai-playbook) provides further useful information for industry to consider when collaborating with Defence on AI deployment.  Data Labelling Current Situation  Currently, the Data Labelling processes within MOD are ineffective and inefficient as capabilities are siloed, time consuming and inconsistent. The workforce has a mixture of Crown Servants/Contractors delivering this capability for specific business areas across varying classification levels. Data Labelling Capability Vision With regards to the Data Labelling capability, MOD's vision is that: "We will establish an Data Labelling capability, empowering the rapid development and deployment of AI solutions across Defence by providing tooling and a trusted network of partners to label, enrich and analyse diverse data modalities at speed and scale. This Data Labelling capability will fuel the development of AI models, delivering faster insights, enhanced situational awareness, and a decisive operational advantage for a safer and more secure future." Data Labelling High-Level Functional/Non-Functional Requirements Work is underway to determine potential user numbers and requirements and MOD may look to procure an enterprise-wide solution including all related agencies and public bodies. MOD is keen to engage with small to medium sized enterprises (SMEs) as part of this process and strongly recommends SMEs provide their feedback. In terms of high-level, initial functional and non-functional requirements for the Data Labelling capability these are provided below. These requirements are not exhaustive nor representative of any final requirements should MOD outsource this capability. Functional requirements The supplier(s) shall be capable of: - providing Data Labelling capabilities across different data types or 'modalities', the main being but not limited to: - Text - Image - Audio - Video - Time Series  - providing Data Labelling capabilities in line with best practice around ethics, safety, bias mitigation and data privacy standards - delivering Service Integration where required with existing data management systems, tools etc. - supporting multiple users across all levels of classification - providing Data Labelling outputs to users in a timely and cost-effective manner Non-Functional requirements - The supplier(s) shall be capable of: - collaborative working within multi skilled teams made of people from several organisations - scaling to handle an increase/decrease in data volumes and number of users - adapting to support changes related to business intent, innovation and new modalities - aligning their solution with the current Defence Artificial Intelligence Strategy (https://www.gov.uk/government/publications/defence-artificial-intelligence-strategy/defence-artificial-intelligence-strategy) providing evidence of quality assurance. - The service must be auditable for compliance purposes. - The service must be delivered by appropriately security cleared personnel (likely SC/DV clearance depending on Data Labelling requirement). - The entirety of the service must be delivered on-shore within the UK including data hosting and personnel provision.