Foundation model AI could have a range of positive uses in public services, but the outlook has to be tempered by an appreciation of the risks, according to a report from the Ada Lovelace Institute.
It has addressed the outlook for the technology – designed to produce a wide variety of outputs with the best known example being ChatGPT – through a review of relevant literature and two roundtables, participants in which included government product teams.
Titled Foundation models in the public sector, its key findings include that there are causes for optimism in potential purposes such as automating the review of complex contracts and case files, and catching errors and biases in policy drafts.
It could also power real time chatbots for dealing with public enquiries and consolidate knowledge spread databases into memos to make it easier to understand.
Biases, breaches and threats
But the report also identifies significant risks in using the technology, including biases, privacy breaches, security threats, harms to the workforce, unequal access and misinformation. It is vital to mitigate these through monitoring, internal and independent oversight and engaging with people affected by the technology, and small pilots can help to minimise the risks.
It adds that public services should consider counterfactuals before committing to implementations. This means comparing proposed use cases with more mature and tested alternatives that might be more effective or provide better value for money.
Evaluating the alternatives should be guided by the principles of public life.
There is also a need for government to invest in relevant skills and deal with technical dependencies, and it could consider options such as updating procurement guidelines for AI systems.
“As foundation models’ capabilities evolve and market dynamics change, there will be new opportunities for public interest driven innovation, but new risks also need to be anticipated to ensure effective governance” the report adds.
