Details

The Scottish Public Sector AI Task Force brings the conversations around AI together and deliver concrete actions to support the safe adoption of AI across government.

Minutes

15 October, 1400 to 1500

Welcome and updates - AI Policy, Scottish Government

• Mr McKee, Minister for Public Finance, was welcomed to the task force meeting.

• The UK Government (UKG) recently announced the launch of their new Regulatory Innovation Office (RIO).

• UKG have announced the launch of a programme of grant funding to support researchers in developing ways to ensure safe AI and data technologies. 

Opening Remarks – Minister for Public Finance

• Responsible for Public Sector Reform and a large part of the digital portfolio

• Using data and AI present an opportunity across the public sector to improve public services

• The Public Sector can leverage the skills and knowledge from the wider AI ecosystem, which includes start-ups, business and academia, to enable the widespread adoption of data and AI projects.

Presentations

Scottish AI Register - AI Policy, Scottish Government

• An AI ‘Library’, accessible to all public sector staff, is available via the administration module of the Register. Containing information on risk management and mitigation, AI policy, models and evaluation.  It is a one stop shop to help keep up with this fast-changing technology.

• Work is ongoing to look at how the Register can be used as part of the procurement process which will ensure our commitment of being transparent about AI use within the Scottish Public Sector.

AI in Geospatial Analysis – Geographic Information Systems (GIS) team, Scottish Government

• The use of AI can be seen as less embedded in day-to-day geospatial analytical work, due to a range of factors including the scarcity and expense of geospatial training data

• However, the use of AI is growing in several areas including in earth observation, where techniques like convolutional neural networks can be combined with satellite imagery to detect objects in higher resolutions and identity features, such as cars in a car park.

• These techniques can be used to aid decision making such as planning regulations and identifying where to target disaster relief.

• Supporting and encouraging the publication and sharing of machine-readable geospatial data is essential to advance the use of AI in geospatial analysis and produce key data sets for policy making.

Automation in Social Security Scotland

• Social Security Scotland worked with the Automation Centre of Excellence team to discover low complexity but high value rules-based processes intelligent automation opportunities.

• Three specific processes were identified. One process used closed machine learning techniques to facilitate testing the ability to read dynamic PDF forms.

• A significant amount of staff engagement was carried out across Social Security Scotland and the team have had positive feedback.

• A ‘human in the loop’ process has been enabled to ensure data accuracy.

• The automated components are reusable and scalable across the public sector with positive results seen in the levels of client satisfaction.

Closing remarks – Minister McKee

•  A key outcome of automation is that it is scalable to make a meaningful impact.

• SG has a key role to play on leading the way on AI best practice and delivering for the people of Scotland.

Working Groups

Health and Social Care

Full spectrum of all AI issues related to health and social care. Possible benefits or services deliverable through using AI, Implications for people’s personal data, can care options be improved through new innovations

Technical

Looking at all aspects of the technical side of AI. New developments or practice in relation to data science and new technical/product development. How we can support those working in this space.

Minutes of Working Group Meetings

Health and Social Care

• Discussion on specific challenges with AI and health and social care setting including governance.

• Discussion on how colleagues can work together to identify shared programmes of work.

Technical

• Discussion on the opportunities and limitations of the use of large language models and associated risks and challenges.

• Discussion on how colleagues can work together to test and deploy AI tools.