Today, we look back on our last panel series = governance and AI. We hosted Katherine Mayes, the Head of Data Analytics, AI and digital ID at TechUK, Ellis Parry, the Information Commissioner’s First Data Ethics Advisor, and Shamus Rae, the CEO of Engine B and previous Head of Innovation and Digital Disruption at KPMG. Our moderator, Lord Clement Jones, is a former chair of the House of Lords Select Committee on AI!
Diving deep into AI and governance, these three experts enabled a discussion on what is AI governance and what are its implications. Indeed, how is AI ‘governed’? How can public policy be used to better the AI national strategy? How can the government encourage global understanding of what responsible AI is? Is it possible to have national, let alone international regulations on AI use?
Governing Data and enabling ethical data use
As Lord Clement Jones put it, adequate governance is the way to operationalise the ethics of AI. To discuss this further, the panel started on discussing the impact of governance on AI innovation.
Shamus explained governance should encourage regulations to make sure only relevant, good quality data is shared and used when developing algorithms. Mores, he explained data must be reviewed with specific tools, which often lack in many countries. Using graphs and crosschecking validity, verifying the good quality and unbiasedness of data must be systematic. Educating all firm members, not just having a few technical specialists within the firm, and making all have the right understanding, technical capabilities and tools for data analysis must happen.
Ellis reinforces points out 6 key recommendations on data treatment that would make the UK align to GDPR objectives and enable a better governance of AI data:
– There must exist meaningful privacy notices to ensure transparency,
– There must be an option of defaulting back to (privacy by design and default),
– We must have in the UK more and bigger data training datasets
– The data output must be at times audited to ensure quality data
– There must be a rationale in the data produced
– Data could be anonymised
According to Ellis, by treating data this way, fairness, transparency, and accountability will be encouraged. He further highlights the importance of being able to hold people accountable, through the data treatment and at the output level.
UK governance of AI – How to ensure UK is a pioneer responsible AI nation- Katherine Holden
With Katherine, the discussion moved specifically to the government’s role in relation to ensuring ethical AI. Indeed, are the policy implications in AI’s development? What is the government’s role in managing and implementing AI nationally? She discusses the need for the furthering in the UK of the AI council roadmap, which ensures AI is well adopted throughout the UK, regions making the best of the available technology, while respecting ethical considerations.
For Katherine, AI is the future, but must be well governed through tailored policy. Change must be enabled to propel UK as a pioneer in responsible AI development and use!
She explains this change must be seen through:
– Better access to good quality data to adequately train algorithms
– Responsible exchange of data from private and public organisations
– The development of a skills and talent office for AI within the government, to ensure UK has a large AI taskforce
– The development of more tailored AI masters, BSc, or conversion courses to make sure UK has AI experts in the future
“UK should brand the AI
experience as a place to
innovate in a ‘responsible
AI environment’, and
emphasise on the
importance of funding AI.”
Katherine here gave us multiple pointers on how the government could better ‘govern’ AI, and how responsible AI could be encouraged and made more prominent in the UK! Most importantly, she explains AI implementation is lagging in the UK due to lack of AI development. By explaining that ‘the D must be put back in R&D’, she sheds light on the need to encourage development of AI through increased funding and awareness. According to Katherine, much academic research is going on, but low amounts of investment and government focus has delayed the actual development of this research into actual innovations!
Shrugging off the sceptics which would argue increased too much governmental intervention in AI development will make UK lag in the global innovation race, Katherine further explains adequate AI governance will on the contrary propel UK as a pioneer responsible AI developer! She says UK should brand the AI experience in the UK the place to innovate in a ‘responsible AI environment’, and emphasise on the importance of funding AI, the industry receiving proportionally far less finances than in other developed nations.
Understanding and educating the wider UK population on AI
During the discussion, the question of how to ensure the general UK population is aware of how AI works frequently came up. Indeed, the government’s role, more so than encouraging the good use of AI, and regulating the present AI data to ensure algorithms produce accurate, unbiased result, is also to ensure the population is aware of the effect of AI in their daily life.
Understanding AI, permits individuals to understand their accountability and responsibility in regards to the treated data. All panellists touched upon the importance for accountability in firms to be taken to the board level, therefore how firm’s board must inevitably be educated on AI principles. Educating the population will also permit people to assess risk (frequency and impact value) associated with AI usages depending on contexts. Educating individuals will also encourage data sharing and incentivise individuals
to give data and agree to new regulations, as they understand what happens to the data they accept to share.
The panellists agreed that more governmental effort must be put into educating the labour force on what is AI, through encouraging AI courses at university to increasing global digital competency levels for all. Furthering this, Shamus shed light of the importance of an AI skilled taskforce as low skilled jobs are increasingly being replaced in the UK by technology. Indeed, re-educating people would help lower unemployment and preserve the labour market while developing new technologies.
Regulations or Guidelines? How to govern AI in the UK
The speakers further discussed how governance should be shaped. Are regulations or guidelines better?
Pointing out the extremely rapid technology developments in the AI sector, Lord Clement Jones discussed how difficult it is to adequately regulate technologies. Regulation in that regards should be ‘technology-agnostic’, centred around data use. Limits of regulations are however seen as discussed due to the important coordination fees stemming from compliance regulations, which as seen, must be countered through more funding of AI firms to help them meet regulation targets. Considering this, the panellists discussed the important use of government guidelines to further govern responsible AI.
Indeed, guidelines permit to educate the firm members, and can be more flexibly adapted to each firm’s strategy with AI. SMEs particularly are more adapted to receiving guidelines rather than regulations as it permits them to divide financial resources accordingly to the importance of the specific guidelines. The issue of SMEs’ responses to guidelines is especially urgent, many firms becoming heavily digitalised these past months in response Covi19, needing increased governmental help, from more education and
funds to be able to meet regulation and guidance targets.
Is Transnational Governance Plausible?
The last major discussion topic of this panel was centred on how data could be used and governed transnationally, in a responsible way. Many problems exist when deploying data internationally, especially when trying to ensure the data keeps its quality throughout. Indeed, many countries have far less regulations on data treatment, and the rights individuals have on their data in the UK may be
infringed as the data is deployed transnationally. Data must be protected and comply to ethical considerations in the UK and internationally as well! Of course, protecting data internationally is very complicated, as national regulations cannot be forced world-wide, and anyhow, very large coordination costs would be seen.
While Elis explained anonymising data and creating large linked datasets (who anonymous and independent – privacy by design and default) could be a solution who are would permit an international deployment potentially less harmful to the data owner due to the anonymity created, he explains this research quality would be decrease with these kinds of datasets. Shamus finished off by offering another vision of international deployment. Arguing knowing where the data comes from is very important, he explains data should not be able to move everywhere with no more initial source. This would decrease accountability.
Overall, the panellists delivered a wonderful and insightful discussion which hopefully educated us on the role of policy, regulation, and government intervention in data use in the UK!
Written by Jeanne Rouot