The UK government is exploring generative AI to improve employee productivity and reimagine citizen experiences. While still early days, there are use cases emerging for generative AI from within the public sector – the topic of discussion at the recent Think Data for Government event.
Swindon Borough Council recently released its first generative AI solution with AWS to make public information more accessible. Sarah Pena, head of emerging technology & business improvement at Swindon Borough Council explained the offering, called Simply Readable.
“It’s developed with a local group of individuals that have learning disability,” she explained. “We wanted to explore social aspects around generative AI for good, and whether we could look at producing easy read information for people with learning disabilities, but also support those with low literacy and cognitive impairment and potentially help those where people don’t have English as a first language.”
Swindon already had a translation solution based on AI. That means that, using generative AI informed by its learning disability group, it can have the right format for easy read produced in less than a minute. It can also use supportive imagery and output that has been informed by them.
“We work at pace, and we work in a proof of concept model that doesn’t take away from that governance and ethics side. So, none of our data is used to train the original base foundational models. We’re not using it with a P2 data at the moment, so no personally identifiable information right now, though we will work to do in the future”
Also on the panel, Barbara Keating, head of architecture at Department for Transport (DfT), said generative AI LLMs [large language models] provide an opportunity to open up archive data.
The gen AI opportunity
“We have lots of data that has been collected and collated over the years and is unstructured, is in image form, is in databases that are closed. So, the opportunity to use LLM to ask and answer questions, have that data without having to turn it into structured data or without having a human go into that space to analyse that data, is quite exciting for us.”
She added that “because the generative AI option allows you to ask and answer questions in an anonymised way, [it] allows us to be able to introduce new questions and new answers to those questions as part of our data driven policy.”
Keating also noted that “we are now no longer expected to speak the language of computers in order to talk to computers, but computers have learned our language and are able to talk to us. That’s where we really see the value in the tech space.”
Don’t innovate in silos
Sue Bateman, chief data officer, CDDO, Cabinet Office, meanwhile, commented on the importance of building a community in the gen AI space to share knowledge and experiences. She said in the past, the government and the wider public sector has innovated, but in silos.
If you liked this content…
“We’ve got a real opportunity to start on the right foot in terms of AI and look for those common areas and not developing using silos,” she said. “Otherwise, we’re potentially fixing the foundational data issues but creating new silos on top.”
Looking to the future, Bateman also questioned, “What does the application of these new technologies mean in terms of the frameworks and standards that we have in the security and risk space? Because there is going to be a lot that we potentially can’t explain. So how do we mitigate against that?”
She also touched on literacy in leadership. “When you’ve got the senior leaders on board, that seems to be the point when things unlock. So why aren’t we setting the ambition for having our data leaders in those decision making positions? At the minute, the career path for data professionals probably culminates in a chief data officer or chief data architect role. I think we need to set our ambitions if we’re going to really maximize generative AI and any of the emerging technologies that catch our excitement over the next couple of years. Where’s our perm sec, who started in a data role? Where’s our chief operating officer who started in data? That’s where we unlock this conundrum that we have about literacy and leadership.”
Investment in data needed for AI success
Deepak Shukla, public sector GTM lead at Amazon Web Services (AWS) said organisations will need their own internal proprietary data to be applied within gen AI applications to give long term sustainable value from their investments in generative AI.
“Because you’re going to differentiate your generative AI applications from any generic application which is out there in the market. Your own data is going to be the differentiator and play a role in that,” he said.
Shukla said AWS is advising customers that if they are investing in generative AI, they need to increase their investment in your data, and in their data strategy to ensure that they have a successful generative AI journey.
“We’ve seen relational data, we’ve seen text, now we are in a space where we’re going to use image video, high volume, high velocity data, which is quite expensive to manage and maintain. And you’re going to use generative AI applications to create even more datasets, which is going to make it even more expensive to operate and manage.
“So, if you’re not revisiting a data strategy to support the government’s AI and generative AI ambition, you’re going to miss the point. And in the next six to eight months, we’ll see a huge noise about ‘oh, this is too expensive’ and ‘this is too difficult to do.’ Or beyond certain use cases, ‘we aren’t able to do much because we don’t have a data foundations right and we haven’t prepared a data foundations for this generative AI journey.’”
Save the date! The next Think Data for Government is on June 5, 2024.
